diff options
Diffstat (limited to 'src/gallium/state_trackers')
49 files changed, 11481 insertions, 0 deletions
diff --git a/src/gallium/state_trackers/README b/src/gallium/state_trackers/README new file mode 100644 index 0000000000..28dd27bbd5 --- /dev/null +++ b/src/gallium/state_trackers/README @@ -0,0 +1,2 @@ +This directory is a placeholder for incubating state-trackers. Mesa's +state-tracker is in src/mesa. diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile new file mode 100644 index 0000000000..4f7a953484 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/Makefile @@ -0,0 +1,18 @@ +TARGET = libg3dvl.a +OBJECTS = vl_display.o vl_screen.o vl_context.o vl_surface.o vl_shader_build.o vl_util.o vl_basic_csc.o \ + vl_r16snorm_mc.o vl_r16snorm_mc_buf.o +GALLIUMDIR = ../.. + +CFLAGS += -g -Wall -fPIC -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary -I${GALLIUMDIR}/winsys/g3dvl + +############################################# + +.PHONY = all clean + +all: ${TARGET} + +${TARGET}: ${OBJECTS} + ar rcs $@ $^ + +clean: + rm -rf ${OBJECTS} ${TARGET} diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c new file mode 100644 index 0000000000..e3b3d03256 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c @@ -0,0 +1,712 @@ +#define VL_INTERNAL +#include "vl_basic_csc.h" +#include <assert.h> +#include <stdlib.h> +#include <pipe/p_context.h> +#include <pipe/p_winsys.h> +#include <pipe/p_state.h> +#include <pipe/p_inlines.h> +#include <tgsi/tgsi_parse.h> +#include <tgsi/tgsi_build.h> +#include "vl_csc.h" +#include "vl_surface.h" +#include "vl_shader_build.h" +#include "vl_types.h" + +struct vlVertexShaderConsts +{ + struct vlVertex4f dst_scale; + struct vlVertex4f dst_trans; + struct vlVertex4f src_scale; + struct vlVertex4f src_trans; +}; + +struct vlFragmentShaderConsts +{ + struct vlVertex4f bias; + float matrix[16]; +}; + +struct vlBasicCSC +{ + struct vlCSC base; + + struct pipe_context *pipe; + struct pipe_viewport_state viewport; + struct pipe_framebuffer_state framebuffer; + struct pipe_texture *framebuffer_tex; + void *sampler; + void *vertex_shader, *fragment_shader; + struct pipe_vertex_buffer vertex_bufs[2]; + struct pipe_vertex_element vertex_elems[2]; + struct pipe_constant_buffer vs_const_buf, fs_const_buf; +}; + +static int vlResizeFrameBuffer +( + struct vlCSC *csc, + unsigned int width, + unsigned int height +) +{ + struct vlBasicCSC *basic_csc; + struct pipe_context *pipe; + struct pipe_texture template; + + assert(csc); + + basic_csc = (struct vlBasicCSC*)csc; + pipe = basic_csc->pipe; + + if (basic_csc->framebuffer.width == width && basic_csc->framebuffer.height == height) + return 0; + + basic_csc->viewport.scale[0] = width; + basic_csc->viewport.scale[1] = height; + basic_csc->viewport.scale[2] = 1; + basic_csc->viewport.scale[3] = 1; + basic_csc->viewport.translate[0] = 0; + basic_csc->viewport.translate[1] = 0; + basic_csc->viewport.translate[2] = 0; + basic_csc->viewport.translate[3] = 0; + + if (basic_csc->framebuffer_tex) + pipe_texture_release(&basic_csc->framebuffer_tex); + + memset(&template, 0, sizeof(struct pipe_texture)); + template.target = PIPE_TEXTURE_2D; + template.format = PIPE_FORMAT_A8R8G8B8_UNORM; + template.last_level = 0; + template.width[0] = width; + template.height[0] = height; + template.depth[0] = 1; + template.compressed = 0; + pf_get_block(template.format, &template.block); + template.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET; + + basic_csc->framebuffer_tex = pipe->screen->texture_create(pipe->screen, &template); + + basic_csc->framebuffer.width = width; + basic_csc->framebuffer.height = height; + basic_csc->framebuffer.cbufs[0] = pipe->screen->get_tex_surface + ( + pipe->screen, + basic_csc->framebuffer_tex, + 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE + ); + + /* Clear to black, in case video doesn't fill the entire window */ + pipe->clear(pipe, basic_csc->framebuffer.cbufs[0], 0); + + return 0; +} + +static int vlBegin +( + struct vlCSC *csc +) +{ + struct vlBasicCSC *basic_csc; + struct pipe_context *pipe; + + assert(csc); + + basic_csc = (struct vlBasicCSC*)csc; + pipe = basic_csc->pipe; + + pipe->set_framebuffer_state(pipe, &basic_csc->framebuffer); + pipe->set_viewport_state(pipe, &basic_csc->viewport); + pipe->bind_sampler_states(pipe, 1, (void**)&basic_csc->sampler); + /* Source texture set in vlPutPictureCSC() */ + pipe->bind_vs_state(pipe, basic_csc->vertex_shader); + pipe->bind_fs_state(pipe, basic_csc->fragment_shader); + pipe->set_vertex_buffers(pipe, 2, basic_csc->vertex_bufs); + pipe->set_vertex_elements(pipe, 2, basic_csc->vertex_elems); + pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &basic_csc->vs_const_buf); + pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &basic_csc->fs_const_buf); + + return 0; +} + +static int vlPutPictureCSC +( + struct vlCSC *csc, + struct vlSurface *surface, + int srcx, + int srcy, + int srcw, + int srch, + int destx, + int desty, + int destw, + int desth, + enum vlPictureType picture_type +) +{ + struct vlBasicCSC *basic_csc; + struct pipe_context *pipe; + struct vlVertexShaderConsts *vs_consts; + + assert(csc); + assert(surface); + + basic_csc = (struct vlBasicCSC*)csc; + pipe = basic_csc->pipe; + + vs_consts = pipe->winsys->buffer_map + ( + pipe->winsys, + basic_csc->vs_const_buf.buffer, + PIPE_BUFFER_USAGE_CPU_WRITE + ); + + vs_consts->dst_scale.x = destw / (float)basic_csc->framebuffer.cbufs[0]->width; + vs_consts->dst_scale.y = desth / (float)basic_csc->framebuffer.cbufs[0]->height; + vs_consts->dst_scale.z = 1; + vs_consts->dst_scale.w = 1; + vs_consts->dst_trans.x = destx / (float)basic_csc->framebuffer.cbufs[0]->width; + vs_consts->dst_trans.y = desty / (float)basic_csc->framebuffer.cbufs[0]->height; + vs_consts->dst_trans.z = 0; + vs_consts->dst_trans.w = 0; + + vs_consts->src_scale.x = srcw / (float)surface->texture->width[0]; + vs_consts->src_scale.y = srch / (float)surface->texture->height[0]; + vs_consts->src_scale.z = 1; + vs_consts->src_scale.w = 1; + vs_consts->src_trans.x = srcx / (float)surface->texture->width[0]; + vs_consts->src_trans.y = srcy / (float)surface->texture->height[0]; + vs_consts->src_trans.z = 0; + vs_consts->src_trans.w = 0; + + pipe->winsys->buffer_unmap(pipe->winsys, basic_csc->vs_const_buf.buffer); + + pipe->set_sampler_textures(pipe, 1, &surface->texture); + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_STRIP, 0, 4); + + return 0; +} + +static int vlEnd +( + struct vlCSC *csc +) +{ + assert(csc); + + return 0; +} + +static struct pipe_surface* vlGetFrameBuffer +( + struct vlCSC *csc +) +{ + struct vlBasicCSC *basic_csc; + + assert(csc); + + basic_csc = (struct vlBasicCSC*)csc; + + return basic_csc->framebuffer.cbufs[0]; +} + +static int vlDestroy +( + struct vlCSC *csc +) +{ + struct vlBasicCSC *basic_csc; + struct pipe_context *pipe; + unsigned int i; + + assert(csc); + + basic_csc = (struct vlBasicCSC*)csc; + pipe = basic_csc->pipe; + + if (basic_csc->framebuffer_tex) + pipe_texture_release(&basic_csc->framebuffer_tex); + + pipe->delete_sampler_state(pipe, basic_csc->sampler); + pipe->delete_vs_state(pipe, basic_csc->vertex_shader); + pipe->delete_fs_state(pipe, basic_csc->fragment_shader); + + for (i = 0; i < 2; ++i) + pipe->winsys->buffer_destroy(pipe->winsys, basic_csc->vertex_bufs[i].buffer); + + pipe->winsys->buffer_destroy(pipe->winsys, basic_csc->vs_const_buf.buffer); + pipe->winsys->buffer_destroy(pipe->winsys, basic_csc->fs_const_buf.buffer); + + free(basic_csc); + + return 0; +} + +/* + * Represents 2 triangles in a strip in normalized coords. + * Used to render the surface onto the frame buffer. + */ +static const struct vlVertex2f surface_verts[4] = +{ + {0.0f, 0.0f}, + {0.0f, 1.0f}, + {1.0f, 0.0f}, + {1.0f, 1.0f} +}; + +/* + * Represents texcoords for the above. We can use the position values directly. + * TODO: Duplicate these in the shader, no need to create a buffer. + */ +static const struct vlVertex2f *surface_texcoords = surface_verts; + +/* + * Identity color conversion constants, for debugging + */ +static const struct vlFragmentShaderConsts identity = +{ + { + 0.0f, 0.0f, 0.0f, 0.0f + }, + { + 1.0f, 0.0f, 0.0f, 0.0f, + 0.0f, 1.0f, 0.0f, 0.0f, + 0.0f, 0.0f, 1.0f, 0.0f, + 0.0f, 0.0f, 0.0f, 1.0f + } +}; + +/* + * Converts ITU-R BT.601 YCbCr pixels to RGB pixels where: + * Y is in [16,235], Cb and Cr are in [16,240] + * R, G, and B are in [16,235] + */ +static const struct vlFragmentShaderConsts bt_601 = +{ + { + 0.0f, 0.501960784f, 0.501960784f, 0.0f + }, + { + 1.0f, 0.0f, 1.371f, 0.0f, + 1.0f, -0.336f, -0.698f, 0.0f, + 1.0f, 1.732f, 0.0f, 0.0f, + 0.0f, 0.0f, 0.0f, 1.0f + } +}; + +/* + * Converts ITU-R BT.601 YCbCr pixels to RGB pixels where: + * Y is in [16,235], Cb and Cr are in [16,240] + * R, G, and B are in [0,255] + */ +static const struct vlFragmentShaderConsts bt_601_full = +{ + { + 0.062745098f, 0.501960784f, 0.501960784f, 0.0f + }, + { + 1.164f, 0.0f, 1.596f, 0.0f, + 1.164f, -0.391f, -0.813f, 0.0f, + 1.164f, 2.018f, 0.0f, 0.0f, + 0.0f, 0.0f, 0.0f, 1.0f + } +}; + +/* + * Converts ITU-R BT.709 YCbCr pixels to RGB pixels where: + * Y is in [16,235], Cb and Cr are in [16,240] + * R, G, and B are in [16,235] + */ +static const struct vlFragmentShaderConsts bt_709 = +{ + { + 0.0f, 0.501960784f, 0.501960784f, 0.0f + }, + { + 1.0f, 0.0f, 1.540f, 0.0f, + 1.0f, -0.183f, -0.459f, 0.0f, + 1.0f, 1.816f, 0.0f, 0.0f, + 0.0f, 0.0f, 0.0f, 1.0f + } +}; + +/* + * Converts ITU-R BT.709 YCbCr pixels to RGB pixels where: + * Y is in [16,235], Cb and Cr are in [16,240] + * R, G, and B are in [0,255] + */ +const struct vlFragmentShaderConsts bt_709_full = +{ + { + 0.062745098f, 0.501960784f, 0.501960784f, 0.0f + }, + { + 1.164f, 0.0f, 1.793f, 0.0f, + 1.164f, -0.213f, -0.534f, 0.0f, + 1.164f, 2.115f, 0.0f, 0.0f, + 0.0f, 0.0f, 0.0f, 1.0f + } +}; + +static int vlCreateVertexShader +( + struct vlBasicCSC *csc +) +{ + const unsigned int max_tokens = 50; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(context); + + pipe = csc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Vertex texcoords + */ + for (i = 0; i < 2; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling vector to scale vertex pos rect to destination size + * decl c1 ; Translation vector to move vertex pos rect into position + * decl c2 ; Scaling vector to scale texcoord rect to source size + * decl c3 ; Translation vector to move texcoord rect into position + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 3); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl o0 ; Vertex pos + * decl o1 ; Vertex texcoords + */ + for (i = 0; i < 2; i++) + { + decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl t0, t1 */ + decl = vl_decl_temps(0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * madd o0, i0, c0, c1 ; Scale and translate unit output rect to destination size and pos + * madd o1, i1, c2, c3 ; Scale and translate unit texcoord rect to source size and pos + */ + for (i = 0; i < 2; ++i) + { + inst = vl_inst4(TGSI_OPCODE_MADD, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i, TGSI_FILE_CONSTANT, i * 2, TGSI_FILE_CONSTANT, i * 2 + 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + csc->vertex_shader = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShader +( + struct vlBasicCSC *csc +) +{ + const unsigned int max_tokens = 50; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(context); + + pipe = csc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* decl i0 ; Texcoords for s0 */ + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, 1, 0, 0, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl c0 ; Bias vector for CSC + * decl c1-c4 ; CSC matrix c1-c4 + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0 */ + decl = vl_decl_temps(0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl s0 ; Sampler for tex containing picture to display */ + decl = vl_decl_samplers(0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* tex2d t0, i0, s0 ; Read src pixel */ + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* sub t0, t0, c0 ; Subtract bias vector from pixel */ + inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * dp4 o0.x, t0, c1 ; Multiply pixel by the color conversion matrix + * dp4 o0.y, t0, c2 + * dp4 o0.z, t0, c3 + */ + for (i = 0; i < 3; ++i) + { + inst = vl_inst3(TGSI_OPCODE_DP4, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 1); + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + csc->fragment_shader = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateDataBufs +( + struct vlBasicCSC *csc +) +{ + struct pipe_context *pipe; + + assert(csc); + + pipe = csc->pipe; + + /* + * Create our vertex buffer and vertex buffer element + * VB contains 4 vertices that render a quad covering the entire window + * to display a rendered surface + * Quad is rendered as a tri strip + */ + csc->vertex_bufs[0].pitch = sizeof(struct vlVertex2f); + csc->vertex_bufs[0].max_index = 3; + csc->vertex_bufs[0].buffer_offset = 0; + csc->vertex_bufs[0].buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(struct vlVertex2f) * 4 + ); + + memcpy + ( + pipe->winsys->buffer_map(pipe->winsys, csc->vertex_bufs[0].buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + surface_verts, + sizeof(struct vlVertex2f) * 4 + ); + + pipe->winsys->buffer_unmap(pipe->winsys, csc->vertex_bufs[0].buffer); + + csc->vertex_elems[0].src_offset = 0; + csc->vertex_elems[0].vertex_buffer_index = 0; + csc->vertex_elems[0].nr_components = 2; + csc->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* + * Create our texcoord buffer and texcoord buffer element + * Texcoord buffer contains the TCs for mapping the rendered surface to the 4 vertices + */ + csc->vertex_bufs[1].pitch = sizeof(struct vlVertex2f); + csc->vertex_bufs[1].max_index = 3; + csc->vertex_bufs[1].buffer_offset = 0; + csc->vertex_bufs[1].buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(struct vlVertex2f) * 4 + ); + + memcpy + ( + pipe->winsys->buffer_map(pipe->winsys, csc->vertex_bufs[1].buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + surface_texcoords, + sizeof(struct vlVertex2f) * 4 + ); + + pipe->winsys->buffer_unmap(pipe->winsys, csc->vertex_bufs[1].buffer); + + csc->vertex_elems[1].src_offset = 0; + csc->vertex_elems[1].vertex_buffer_index = 1; + csc->vertex_elems[1].nr_components = 2; + csc->vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* + * Create our vertex shader's constant buffer + * Const buffer contains scaling and translation vectors + */ + csc->vs_const_buf.size = sizeof(struct vlVertexShaderConsts); + csc->vs_const_buf.buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_CONSTANT, + csc->vs_const_buf.size + ); + + /* + * Create our fragment shader's constant buffer + * Const buffer contains the color conversion matrix and bias vectors + */ + csc->fs_const_buf.size = sizeof(struct vlFragmentShaderConsts); + csc->fs_const_buf.buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_CONSTANT, + csc->fs_const_buf.size + ); + + /* + * TODO: Refactor this into a seperate function, + * allow changing the CSC matrix at runtime to switch between regular & full versions + */ + memcpy + ( + pipe->winsys->buffer_map(pipe->winsys, csc->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + &bt_601, + sizeof(struct vlFragmentShaderConsts) + ); + + pipe->winsys->buffer_unmap(pipe->winsys, csc->fs_const_buf.buffer); + + return 0; +} + +static int vlInit +( + struct vlBasicCSC *csc +) +{ + struct pipe_context *pipe; + struct pipe_sampler_state sampler; + + assert(csc); + + pipe = csc->pipe; + + /* Delay creating the FB until vlPutPictureCSC() so we know window size */ + csc->framebuffer_tex = NULL; + csc->framebuffer.width = 0; + csc->framebuffer.height = 0; + csc->framebuffer.num_cbufs = 1; + csc->framebuffer.cbufs[0] = NULL; + csc->framebuffer.zsbuf = NULL; + + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.compare_mode = PIPE_TEX_COMPARE_NONE; + sampler.compare_func = PIPE_FUNC_ALWAYS; + sampler.normalized_coords = 1; + /*sampler.prefilter = ;*/ + /*sampler.shadow_ambient = ;*/ + /*sampler.lod_bias = ;*/ + /*sampler.min_lod = ;*/ + /*sampler.max_lod = ;*/ + /*sampler.border_color[i] = ;*/ + /*sampler.max_anisotropy = ;*/ + csc->sampler = pipe->create_sampler_state(pipe, &sampler); + + vlCreateVertexShader(csc); + vlCreateFragmentShader(csc); + vlCreateDataBufs(csc); + + return 0; +} + +int vlCreateBasicCSC +( + struct pipe_context *pipe, + struct vlCSC **csc +) +{ + struct vlBasicCSC *basic_csc; + + assert(pipe); + assert(csc); + + basic_csc = calloc(1, sizeof(struct vlBasicCSC)); + + if (!basic_csc) + return 1; + + basic_csc->base.vlResizeFrameBuffer = &vlResizeFrameBuffer; + basic_csc->base.vlBegin = &vlBegin; + basic_csc->base.vlPutPicture = &vlPutPictureCSC; + basic_csc->base.vlEnd = &vlEnd; + basic_csc->base.vlGetFrameBuffer = &vlGetFrameBuffer; + basic_csc->base.vlDestroy = &vlDestroy; + basic_csc->pipe = pipe; + + vlInit(basic_csc); + + *csc = &basic_csc->base; + + return 0; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.h b/src/gallium/state_trackers/g3dvl/vl_basic_csc.h new file mode 100644 index 0000000000..2e17f1d814 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.h @@ -0,0 +1,13 @@ +#ifndef vl_basic_csc_h +#define vl_basic_csc_h + +struct pipe_context; +struct vlCSC; + +int vlCreateBasicCSC +( + struct pipe_context *pipe, + struct vlCSC **csc +); + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c new file mode 100644 index 0000000000..fe107e406d --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_context.c @@ -0,0 +1,210 @@ +#define VL_INTERNAL +#include "vl_context.h" +#include <assert.h> +#include <stdlib.h> +#include <pipe/p_context.h> +#include <pipe/p_state.h> +#include "vl_render.h" +#include "vl_r16snorm_mc.h" +#include "vl_r16snorm_mc_buf.h" +#include "vl_csc.h" +#include "vl_basic_csc.h" + +static int vlInitCommon(struct vlContext *context) +{ + struct pipe_context *pipe; + struct pipe_rasterizer_state rast; + struct pipe_blend_state blend; + struct pipe_depth_stencil_alpha_state dsa; + unsigned int i; + + assert(context); + + pipe = context->pipe; + + rast.flatshade = 1; + rast.flatshade_first = 0; + rast.light_twoside = 0; + rast.front_winding = PIPE_WINDING_CCW; + rast.cull_mode = PIPE_WINDING_CW; + rast.fill_cw = PIPE_POLYGON_MODE_FILL; + rast.fill_ccw = PIPE_POLYGON_MODE_FILL; + rast.offset_cw = 0; + rast.offset_ccw = 0; + rast.scissor = 0; + rast.poly_smooth = 0; + rast.poly_stipple_enable = 0; + rast.point_sprite = 0; + rast.point_size_per_vertex = 0; + rast.multisample = 0; + rast.line_smooth = 0; + rast.line_stipple_enable = 0; + rast.line_stipple_factor = 0; + rast.line_stipple_pattern = 0; + rast.line_last_pixel = 0; + /* Don't need clipping, but viewport mapping done here */ + rast.bypass_clipping = 0; + rast.bypass_vs = 0; + rast.origin_lower_left = 0; + rast.line_width = 1; + rast.point_smooth = 0; + rast.point_size = 1; + rast.offset_units = 1; + rast.offset_scale = 1; + /*rast.sprite_coord_mode[i] = ;*/ + context->raster = pipe->create_rasterizer_state(pipe, &rast); + pipe->bind_rasterizer_state(pipe, context->raster); + + blend.blend_enable = 0; + blend.rgb_func = PIPE_BLEND_ADD; + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_func = PIPE_BLEND_ADD; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ONE; + blend.logicop_enable = 0; + blend.logicop_func = PIPE_LOGICOP_CLEAR; + /* Needed to allow color writes to FB, even if blending disabled */ + blend.colormask = PIPE_MASK_RGBA; + blend.dither = 0; + context->blend = pipe->create_blend_state(pipe, &blend); + pipe->bind_blend_state(pipe, context->blend); + + dsa.depth.enabled = 0; + dsa.depth.writemask = 0; + dsa.depth.func = PIPE_FUNC_ALWAYS; + dsa.depth.occlusion_count = 0; + for (i = 0; i < 2; ++i) + { + dsa.stencil[i].enabled = 0; + dsa.stencil[i].func = PIPE_FUNC_ALWAYS; + dsa.stencil[i].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[i].zpass_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[i].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[i].ref_value = 0; + dsa.stencil[i].value_mask = 0; + dsa.stencil[i].write_mask = 0; + } + dsa.alpha.enabled = 0; + dsa.alpha.func = PIPE_FUNC_ALWAYS; + dsa.alpha.ref = 0; + context->dsa = pipe->create_depth_stencil_alpha_state(pipe, &dsa); + pipe->bind_depth_stencil_alpha_state(pipe, context->dsa); + + return 0; +} + +int vlCreateContext +( + struct vlScreen *screen, + struct pipe_context *pipe, + unsigned int picture_width, + unsigned int picture_height, + enum vlFormat picture_format, + enum vlProfile profile, + enum vlEntryPoint entry_point, + struct vlContext **context +) +{ + struct vlContext *ctx; + + assert(screen); + assert(context); + assert(pipe); + + ctx = calloc(1, sizeof(struct vlContext)); + + if (!ctx) + return 1; + + ctx->screen = screen; + ctx->pipe = pipe; + ctx->picture_width = picture_width; + ctx->picture_height = picture_height; + ctx->picture_format = picture_format; + ctx->profile = profile; + ctx->entry_point = entry_point; + + vlInitCommon(ctx); + + /*vlCreateR16SNormMC(pipe, picture_width, picture_height, picture_format, &ctx->render);*/ + vlCreateR16SNormBufferedMC(pipe, picture_width, picture_height, picture_format, &ctx->render); + vlCreateBasicCSC(pipe, &ctx->csc); + + *context = ctx; + + return 0; +} + +int vlDestroyContext +( + struct vlContext *context +) +{ + assert(context); + + /* XXX: Must unbind shaders before we can delete them for some reason */ + context->pipe->bind_vs_state(context->pipe, NULL); + context->pipe->bind_fs_state(context->pipe, NULL); + + context->render->vlDestroy(context->render); + context->csc->vlDestroy(context->csc); + + context->pipe->delete_blend_state(context->pipe, context->blend); + context->pipe->delete_rasterizer_state(context->pipe, context->raster); + context->pipe->delete_depth_stencil_alpha_state(context->pipe, context->dsa); + + free(context); + + return 0; +} + +struct vlScreen* vlContextGetScreen +( + struct vlContext *context +) +{ + assert(context); + + return context->screen; +} + +struct pipe_context* vlGetPipeContext +( + struct vlContext *context +) +{ + assert(context); + + return context->pipe; +} + +unsigned int vlGetPictureWidth +( + struct vlContext *context +) +{ + assert(context); + + return context->picture_width; +} + +unsigned int vlGetPictureHeight +( + struct vlContext *context +) +{ + assert(context); + + return context->picture_height; +} + +enum vlFormat vlGetPictureFormat +( + struct vlContext *context +) +{ + assert(context); + + return context->picture_format; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_context.h b/src/gallium/state_trackers/g3dvl/vl_context.h new file mode 100644 index 0000000000..3d14634c44 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_context.h @@ -0,0 +1,73 @@ +#ifndef vl_context_h +#define vl_context_h + +#include "vl_types.h" + +struct pipe_context; + +#ifdef VL_INTERNAL +struct vlRender; +struct vlCSC; + +struct vlContext +{ + struct vlScreen *screen; + struct pipe_context *pipe; + unsigned int picture_width; + unsigned int picture_height; + enum vlFormat picture_format; + enum vlProfile profile; + enum vlEntryPoint entry_point; + + void *raster; + void *dsa; + void *blend; + + struct vlRender *render; + struct vlCSC *csc; +}; +#endif + +int vlCreateContext +( + struct vlScreen *screen, + struct pipe_context *pipe, + unsigned int picture_width, + unsigned int picture_height, + enum vlFormat picture_format, + enum vlProfile profile, + enum vlEntryPoint entry_point, + struct vlContext **context +); + +int vlDestroyContext +( + struct vlContext *context +); + +struct vlScreen* vlContextGetScreen +( + struct vlContext *context +); + +struct pipe_context* vlGetPipeContext +( + struct vlContext *context +); + +unsigned int vlGetPictureWidth +( + struct vlContext *context +); + +unsigned int vlGetPictureHeight +( + struct vlContext *context +); + +enum vlFormat vlGetPictureFormat +( + struct vlContext *context +); + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_csc.h b/src/gallium/state_trackers/g3dvl/vl_csc.h new file mode 100644 index 0000000000..36417a2792 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_csc.h @@ -0,0 +1,53 @@ +#ifndef vl_csc_h +#define vl_csc_h + +#include "vl_types.h" + +struct pipe_surface; + +struct vlCSC +{ + int (*vlResizeFrameBuffer) + ( + struct vlCSC *csc, + unsigned int width, + unsigned int height + ); + + int (*vlBegin) + ( + struct vlCSC *csc + ); + + int (*vlPutPicture) + ( + struct vlCSC *csc, + struct vlSurface *surface, + int srcx, + int srcy, + int srcw, + int srch, + int destx, + int desty, + int destw, + int desth, + enum vlPictureType picture_type + ); + + int (*vlEnd) + ( + struct vlCSC *csc + ); + + struct pipe_surface* (*vlGetFrameBuffer) + ( + struct vlCSC *csc + ); + + int (*vlDestroy) + ( + struct vlCSC *csc + ); +}; + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_defs.h b/src/gallium/state_trackers/g3dvl/vl_defs.h new file mode 100644 index 0000000000..d612d02502 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_defs.h @@ -0,0 +1,11 @@ +#ifndef vl_defs_h +#define vl_defs_h + +#define VL_BLOCK_WIDTH 8 +#define VL_BLOCK_HEIGHT 8 +#define VL_BLOCK_SIZE (VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT) +#define VL_MACROBLOCK_WIDTH 16 +#define VL_MACROBLOCK_HEIGHT 16 +#define VL_MACROBLOCK_SIZE (VL_MACROBLOCK_WIDTH * VL_MACROBLOCK_HEIGHT) + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_display.c b/src/gallium/state_trackers/g3dvl/vl_display.c new file mode 100644 index 0000000000..af80faa7f5 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_display.c @@ -0,0 +1,48 @@ +#define VL_INTERNAL +#include "vl_display.h" +#include <assert.h> +#include <stdlib.h> + +int vlCreateDisplay +( + vlNativeDisplay native_display, + struct vlDisplay **display +) +{ + struct vlDisplay *dpy; + + assert(native_display); + assert(display); + + dpy = calloc(1, sizeof(struct vlDisplay)); + + if (!dpy) + return 1; + + dpy->native = native_display; + *display = dpy; + + return 0; +} + +int vlDestroyDisplay +( + struct vlDisplay *display +) +{ + assert(display); + + free(display); + + return 0; +} + +vlNativeDisplay vlGetNativeDisplay +( + struct vlDisplay *display +) +{ + assert(display); + + return display->native; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_display.h b/src/gallium/state_trackers/g3dvl/vl_display.h new file mode 100644 index 0000000000..e11fd40799 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_display.h @@ -0,0 +1,29 @@ +#ifndef vl_display_h +#define vl_display_h + +#include "vl_types.h" + +#ifdef VL_INTERNAL +struct vlDisplay +{ + vlNativeDisplay native; +}; +#endif + +int vlCreateDisplay +( + vlNativeDisplay native_display, + struct vlDisplay **display +); + +int vlDestroyDisplay +( + struct vlDisplay *display +); + +vlNativeDisplay vlGetNativeDisplay +( + struct vlDisplay *display +); + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c new file mode 100644 index 0000000000..3272220ef8 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c @@ -0,0 +1,2344 @@ +#define VL_INTERNAL +#include "vl_r16snorm_mc.h" +#include <assert.h> +#include <stdlib.h> +#include <pipe/p_context.h> +#include <pipe/p_winsys.h> +#include <pipe/p_state.h> +#include <pipe/p_inlines.h> +#include <tgsi/tgsi_parse.h> +#include <tgsi/tgsi_build.h> +#include "vl_render.h" +#include "vl_shader_build.h" +#include "vl_surface.h" +#include "vl_util.h" +#include "vl_types.h" +#include "vl_defs.h" + +#define NUM_BUFS 4 /* Number of rotating buffers to use */ + +struct vlVertexShaderConsts +{ + /*struct vlVertex4f scale; + struct vlVertex4f denorm;*/ + struct vlVertex4f scale; + struct vlVertex4f mb_pos_trans; + struct vlVertex4f denorm; + struct + { + struct vlVertex4f top_field; + struct vlVertex4f bottom_field; + } mb_tc_trans[2]; +}; + +struct vlFragmentShaderConsts +{ + struct vlVertex4f multiplier; + struct vlVertex4f div; +}; + +struct vlR16SnormMC +{ + struct vlRender base; + + unsigned int video_width, video_height; + enum vlFormat video_format; + unsigned int cur_buf; + + struct pipe_context *pipe; + struct pipe_viewport_state viewport; + struct pipe_framebuffer_state render_target; + struct pipe_sampler_state *samplers[5]; + struct pipe_texture *textures[NUM_BUFS][5]; + void *i_vs, *p_vs[2], *b_vs[2]; + void *i_fs, *p_fs[2], *b_fs[2]; + struct pipe_vertex_buffer vertex_bufs[3]; + struct pipe_vertex_element vertex_elems[3]; + struct pipe_constant_buffer vs_const_buf, fs_const_buf; +}; + +static int vlBegin +( + struct vlRender *render +) +{ + struct vlR16SnormMC *mc; + struct pipe_context *pipe; + + assert(render); + + mc = (struct vlR16SnormMC*)render; + pipe = mc->pipe; + + /* Frame buffer set in vlRender*Macroblock() */ + /* Shaders, samplers, textures set in vlRender*Macroblock() */ + pipe->set_vertex_buffers(pipe, 3, mc->vertex_bufs); + pipe->set_vertex_elements(pipe, 3, mc->vertex_elems); + pipe->set_viewport_state(pipe, &mc->viewport); + pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &mc->vs_const_buf); + pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &mc->fs_const_buf); + + return 0; +} + +/*static int vlGrabMacroBlock +( + struct vlR16SnormMC *mc, + struct vlMpeg2MacroBlock *macroblock +) +{ + assert(mc); + assert(macroblock); + + + + return 0; +}*/ + +/*#define DO_IDCT*/ + +#ifdef DO_IDCT +static int vlTransformBlock(short *src, short *dst, short bias) +{ + static const float basis[8][8] = + { + {0.3536, 0.4904, 0.4619, 0.4157, 0.3536, 0.2778, 0.1913, 0.0975}, + {0.3536, 0.4157, 0.1913, -0.0975, -0.3536, -0.4904, -0.4619, -0.2778}, + {0.3536, 0.2778, -0.1913, -0.4904, -0.3536, 0.0975, 0.4619, 0.4157}, + {0.3536, 0.0975, -0.4619, -0.2778, 0.3536, 0.4157, -0.1913, -0.4904}, + {0.3536, -0.0975, -0.4619, 0.2778, 0.3536, -0.4157, -0.1913, 0.4904}, + {0.3536, -0.2778, -0.1913, 0.4904, -0.3536, -0.0975, 0.4619, -0.4157}, + {0.3536, -0.4157, 0.1913, 0.0975, -0.3536, 0.4904, -0.4619, 0.2778}, + {0.3536, -0.4904, 0.4619, -0.4157, 0.3536, -0.2778, 0.1913, -0.0975} + }; + + unsigned int x, y; + short tmp[64]; + + for (y = 0; y < VL_BLOCK_HEIGHT; ++y) + for (x = 0; x < VL_BLOCK_WIDTH; ++x) + tmp[y * VL_BLOCK_WIDTH + x] = (short) + ( + src[y * VL_BLOCK_WIDTH + 0] * basis[x][0] + + src[y * VL_BLOCK_WIDTH + 1] * basis[x][1] + + src[y * VL_BLOCK_WIDTH + 2] * basis[x][2] + + src[y * VL_BLOCK_WIDTH + 3] * basis[x][3] + + src[y * VL_BLOCK_WIDTH + 4] * basis[x][4] + + src[y * VL_BLOCK_WIDTH + 5] * basis[x][5] + + src[y * VL_BLOCK_WIDTH + 6] * basis[x][6] + + src[y * VL_BLOCK_WIDTH + 7] * basis[x][7] + ); + + for (x = 0; x < VL_BLOCK_WIDTH; ++x) + for (y = 0; y < VL_BLOCK_HEIGHT; ++y) + { + dst[y * VL_BLOCK_WIDTH + x] = bias + (short) + ( + tmp[0 * VL_BLOCK_WIDTH + x] * basis[y][0] + + tmp[1 * VL_BLOCK_WIDTH + x] * basis[y][1] + + tmp[2 * VL_BLOCK_WIDTH + x] * basis[y][2] + + tmp[3 * VL_BLOCK_WIDTH + x] * basis[y][3] + + tmp[4 * VL_BLOCK_WIDTH + x] * basis[y][4] + + tmp[5 * VL_BLOCK_WIDTH + x] * basis[y][5] + + tmp[6 * VL_BLOCK_WIDTH + x] * basis[y][6] + + tmp[7 * VL_BLOCK_WIDTH + x] * basis[y][7] + ); + if (dst[y * VL_BLOCK_WIDTH + x] > 255) + dst[y * VL_BLOCK_WIDTH + x] = 255; + else if (bias > 0 && dst[y * VL_BLOCK_WIDTH + x] < 0) + dst[y * VL_BLOCK_WIDTH + x] = 0; + } + return 0; +} +#endif + +static int vlGrabFrameCodedBlock(short *src, short *dst, unsigned int dst_pitch) +{ + unsigned int y; + + for (y = 0; y < VL_BLOCK_HEIGHT; ++y) + memcpy + ( + dst + y * dst_pitch, + src + y * VL_BLOCK_WIDTH, + VL_BLOCK_WIDTH * 2 + ); + + return 0; +} + +static int vlGrabFieldCodedBlock(short *src, short *dst, unsigned int dst_pitch) +{ + unsigned int y; + + for (y = 0; y < VL_BLOCK_HEIGHT / 2; ++y) + memcpy + ( + dst + y * dst_pitch * 2, + src + y * VL_BLOCK_WIDTH, + VL_BLOCK_WIDTH * 2 + ); + + dst += VL_BLOCK_HEIGHT * dst_pitch; + + for (; y < VL_BLOCK_HEIGHT; ++y) + memcpy + ( + dst + y * dst_pitch * 2, + src + y * VL_BLOCK_WIDTH, + VL_BLOCK_WIDTH * 2 + ); + + return 0; +} + +static int vlGrabNoBlock(short *dst, unsigned int dst_pitch) +{ + unsigned int y; + + for (y = 0; y < VL_BLOCK_HEIGHT; ++y) + memset + ( + dst + y * dst_pitch, + 0, + VL_BLOCK_WIDTH * 2 + ); + + return 0; +} + +enum vlSampleType +{ + vlSampleTypeFull, + vlSampleTypeDiff +}; + +static int vlGrabBlocks +( + struct vlR16SnormMC *mc, + unsigned int coded_block_pattern, + enum vlDCTType dct_type, + enum vlSampleType sample_type, + short *blocks +) +{ + struct pipe_surface *tex_surface; + short *texels; + unsigned int tex_pitch; + unsigned int tb, sb = 0; + + assert(mc); + assert(blocks); + + tex_surface = mc->pipe->screen->get_tex_surface + ( + mc->pipe->screen, + mc->textures[mc->cur_buf % NUM_BUFS][0], + 0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE + ); + + texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE); + tex_pitch = tex_surface->stride / tex_surface->block.size; + + for (tb = 0; tb < 4; ++tb) + { + if ((coded_block_pattern >> (5 - tb)) & 1) + { + short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT; + +#ifdef DO_IDCT + vlTransformBlock(cur_block, cur_block, sample_type == vlSampleTypeFull ? 128 : 0); +#endif + + if (dct_type == vlDCTTypeFrameCoded) + vlGrabFrameCodedBlock + ( + cur_block, + texels + tb * tex_pitch * VL_BLOCK_HEIGHT, + tex_pitch + ); + else + vlGrabFieldCodedBlock + ( + cur_block, + texels + (tb % 2) * tex_pitch * VL_BLOCK_HEIGHT + (tb / 2) * tex_pitch, + tex_pitch + ); + + ++sb; + } + else + vlGrabNoBlock(texels + tb * tex_pitch * VL_BLOCK_HEIGHT, tex_pitch); + } + + pipe_surface_unmap(tex_surface); + + /* TODO: Implement 422, 444 */ + for (tb = 0; tb < 2; ++tb) + { + tex_surface = mc->pipe->screen->get_tex_surface + ( + mc->pipe->screen, + mc->textures[mc->cur_buf % NUM_BUFS][tb + 1], + 0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE + ); + + texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE); + tex_pitch = tex_surface->stride / tex_surface->block.size; + + if ((coded_block_pattern >> (1 - tb)) & 1) + { + short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT; + +#ifdef DO_IDCT + vlTransformBlock(cur_block, cur_block, sample_type == vlSampleTypeFull ? 128 : 0); +#endif + + vlGrabFrameCodedBlock + ( + cur_block, + texels, + tex_pitch + ); + + ++sb; + } + else + vlGrabNoBlock(texels, tex_pitch); + + pipe_surface_unmap(tex_surface); + } + + return 0; +} + +static int vlRenderIMacroBlock +( + struct vlR16SnormMC *mc, + enum vlPictureType picture_type, + enum vlFieldOrder field_order, + unsigned int mbx, + unsigned int mby, + unsigned int coded_block_pattern, + enum vlDCTType dct_type, + short *blocks, + struct vlSurface *surface +) +{ + struct pipe_context *pipe; + struct vlVertexShaderConsts *vs_consts; + + assert(blocks); + assert(surface); + + /* TODO: Implement interlaced rendering */ + if (picture_type != vlPictureTypeFrame) + return 0; + + vlGrabBlocks(mc, coded_block_pattern, dct_type, vlSampleTypeFull, blocks); + + pipe = mc->pipe; + + vs_consts = pipe->winsys->buffer_map + ( + pipe->winsys, + mc->vs_const_buf.buffer, + PIPE_BUFFER_USAGE_CPU_WRITE + ); + + vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->texture->width[0]; + vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->texture->height[0]; + vs_consts->scale.z = 1.0f; + vs_consts->scale.w = 1.0f; + vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->texture->width[0]; + vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->texture->height[0]; + vs_consts->mb_pos_trans.z = 0.0f; + vs_consts->mb_pos_trans.w = 0.0f; + + pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer); + + mc->render_target.cbufs[0] = pipe->screen->get_tex_surface + ( + pipe->screen, + surface->texture, + 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE + ); + pipe->set_framebuffer_state(pipe, &mc->render_target); + pipe->set_sampler_textures(pipe, 3, mc->textures[mc->cur_buf % NUM_BUFS]); + pipe->bind_sampler_states(pipe, 3, (void**)mc->samplers); + pipe->bind_vs_state(pipe, mc->i_vs); + pipe->bind_fs_state(pipe, mc->i_fs); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24); + + mc->cur_buf++; + + return 0; +} + +static int vlRenderPMacroBlock +( + struct vlR16SnormMC *mc, + enum vlPictureType picture_type, + enum vlFieldOrder field_order, + unsigned int mbx, + unsigned int mby, + enum vlMotionType mc_type, + short top_x, + short top_y, + short bottom_x, + short bottom_y, + unsigned int coded_block_pattern, + enum vlDCTType dct_type, + short *blocks, + struct vlSurface *ref_surface, + struct vlSurface *surface +) +{ + struct pipe_context *pipe; + struct vlVertexShaderConsts *vs_consts; + + assert(motion_vectors); + assert(blocks); + assert(ref_surface); + assert(surface); + + /* TODO: Implement interlaced rendering */ + if (picture_type != vlPictureTypeFrame) + return 0; + /* TODO: Implement other MC types */ + if (mc_type != vlMotionTypeFrame && mc_type != vlMotionTypeField) + return 0; + + vlGrabBlocks(mc, coded_block_pattern, dct_type, vlSampleTypeDiff, blocks); + + pipe = mc->pipe; + + vs_consts = pipe->winsys->buffer_map + ( + pipe->winsys, + mc->vs_const_buf.buffer, + PIPE_BUFFER_USAGE_CPU_WRITE + ); + + vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->texture->width[0]; + vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->texture->height[0]; + vs_consts->scale.z = 1.0f; + vs_consts->scale.w = 1.0f; + vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->texture->width[0]; + vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->texture->height[0]; + vs_consts->mb_pos_trans.z = 0.0f; + vs_consts->mb_pos_trans.w = 0.0f; + vs_consts->mb_tc_trans[0].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + top_x * 0.5f) / (float)surface->texture->width[0]; + vs_consts->mb_tc_trans[0].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + top_y * 0.5f) / (float)surface->texture->height[0]; + vs_consts->mb_tc_trans[0].top_field.z = 0.0f; + vs_consts->mb_tc_trans[0].top_field.w = 0.0f; + + if (mc_type == vlMotionTypeField) + { + vs_consts->denorm.x = (float)surface->texture->width[0]; + vs_consts->denorm.y = (float)surface->texture->height[0]; + + vs_consts->mb_tc_trans[0].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + bottom_x * 0.5f) / (float)surface->texture->width[0]; + vs_consts->mb_tc_trans[0].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + bottom_y * 0.5f) / (float)surface->texture->height[0]; + vs_consts->mb_tc_trans[0].bottom_field.z = 0.0f; + vs_consts->mb_tc_trans[0].bottom_field.w = 0.0f; + + pipe->bind_vs_state(pipe, mc->p_vs[1]); + pipe->bind_fs_state(pipe, mc->p_fs[1]); + } + else + { + pipe->bind_vs_state(pipe, mc->p_vs[0]); + pipe->bind_fs_state(pipe, mc->p_fs[0]); + } + + pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer); + + mc->render_target.cbufs[0] = pipe->screen->get_tex_surface + ( + pipe->screen, + surface->texture, + 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE + ); + pipe->set_framebuffer_state(pipe, &mc->render_target); + + mc->textures[mc->cur_buf % NUM_BUFS][3] = ref_surface->texture; + pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUFS]); + pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24); + + mc->cur_buf++; + + return 0; +} + +static int vlRenderBMacroBlock +( + struct vlR16SnormMC *mc, + enum vlPictureType picture_type, + enum vlFieldOrder field_order, + unsigned int mbx, + unsigned int mby, + enum vlMotionType mc_type, + short top_past_x, + short top_past_y, + short bottom_past_x, + short bottom_past_y, + short top_future_x, + short top_future_y, + short bottom_future_x, + short bottom_future_y, + unsigned int coded_block_pattern, + enum vlDCTType dct_type, + short *blocks, + struct vlSurface *past_surface, + struct vlSurface *future_surface, + struct vlSurface *surface +) +{ + struct pipe_context *pipe; + struct vlVertexShaderConsts *vs_consts; + + assert(motion_vectors); + assert(blocks); + assert(ref_surface); + assert(surface); + + /* TODO: Implement interlaced rendering */ + if (picture_type != vlPictureTypeFrame) + return 0; + /* TODO: Implement other MC types */ + if (mc_type != vlMotionTypeFrame && mc_type != vlMotionTypeField) + return 0; + + vlGrabBlocks(mc, coded_block_pattern, dct_type, vlSampleTypeDiff, blocks); + + pipe = mc->pipe; + + vs_consts = pipe->winsys->buffer_map + ( + pipe->winsys, + mc->vs_const_buf.buffer, + PIPE_BUFFER_USAGE_CPU_WRITE + ); + + vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->texture->width[0]; + vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->texture->height[0]; + vs_consts->scale.z = 1.0f; + vs_consts->scale.w = 1.0f; + vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->texture->width[0]; + vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->texture->height[0]; + vs_consts->mb_pos_trans.z = 0.0f; + vs_consts->mb_pos_trans.w = 0.0f; + vs_consts->mb_tc_trans[0].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + top_past_x * 0.5f) / (float)surface->texture->width[0]; + vs_consts->mb_tc_trans[0].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + top_past_y * 0.5f) / (float)surface->texture->height[0]; + vs_consts->mb_tc_trans[0].top_field.z = 0.0f; + vs_consts->mb_tc_trans[0].top_field.w = 0.0f; + vs_consts->mb_tc_trans[1].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + top_future_x * 0.5f) / (float)surface->texture->width[0]; + vs_consts->mb_tc_trans[1].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + top_future_y * 0.5f) / (float)surface->texture->height[0]; + vs_consts->mb_tc_trans[1].top_field.z = 0.0f; + vs_consts->mb_tc_trans[1].top_field.w = 0.0f; + + if (mc_type == vlMotionTypeField) + { + vs_consts->denorm.x = (float)surface->texture->width[0]; + vs_consts->denorm.y = (float)surface->texture->height[0]; + + vs_consts->mb_tc_trans[0].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + bottom_past_x * 0.5f) / (float)surface->texture->width[0]; + vs_consts->mb_tc_trans[0].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + bottom_past_y * 0.5f) / (float)surface->texture->height[0]; + vs_consts->mb_tc_trans[0].bottom_field.z = 0.0f; + vs_consts->mb_tc_trans[0].bottom_field.w = 0.0f; + vs_consts->mb_tc_trans[1].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + bottom_future_x * 0.5f) / (float)surface->texture->width[0]; + vs_consts->mb_tc_trans[1].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + bottom_future_y * 0.5f) / (float)surface->texture->height[0]; + vs_consts->mb_tc_trans[1].bottom_field.z = 0.0f; + vs_consts->mb_tc_trans[1].bottom_field.w = 0.0f; + + pipe->bind_vs_state(pipe, mc->b_vs[1]); + pipe->bind_fs_state(pipe, mc->b_fs[1]); + } + else + { + pipe->bind_vs_state(pipe, mc->b_vs[0]); + pipe->bind_fs_state(pipe, mc->b_fs[0]); + } + + pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer); + + mc->render_target.cbufs[0] = pipe->screen->get_tex_surface + ( + pipe->screen, + surface->texture, + 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE + ); + pipe->set_framebuffer_state(pipe, &mc->render_target); + + mc->textures[mc->cur_buf % NUM_BUFS][3] = past_surface->texture; + mc->textures[mc->cur_buf % NUM_BUFS][4] = future_surface->texture; + pipe->set_sampler_textures(pipe, 5, mc->textures[mc->cur_buf % NUM_BUFS]); + pipe->bind_sampler_states(pipe, 5, (void**)mc->samplers); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24); + + mc->cur_buf++; + + return 0; +} + +static int vlRenderMacroBlocksMpeg2R16Snorm +( + struct vlRender *render, + struct vlMpeg2MacroBlockBatch *batch, + struct vlSurface *surface +) +{ + struct vlR16SnormMC *mc; + unsigned int i; + + assert(render); + + mc = (struct vlR16SnormMC*)render; + + /*for (i = 0; i < batch->num_macroblocks; ++i) + vlGrabMacroBlock(batch->macroblocks[i]);*/ + + for (i = 0; i < batch->num_macroblocks; ++i) + { + switch (batch->macroblocks[i].mb_type) + { + case vlMacroBlockTypeIntra: + { + vlRenderIMacroBlock + ( + mc, + batch->picture_type, + batch->field_order, + batch->macroblocks[i].mbx, + batch->macroblocks[i].mby, + batch->macroblocks[i].cbp, + batch->macroblocks[i].dct_type, + batch->macroblocks[i].blocks, + surface + ); + break; + } + case vlMacroBlockTypeFwdPredicted: + { + vlRenderPMacroBlock + ( + mc, + batch->picture_type, + batch->field_order, + batch->macroblocks[i].mbx, + batch->macroblocks[i].mby, + batch->macroblocks[i].mo_type, + batch->macroblocks[i].PMV[0][0][0], + batch->macroblocks[i].PMV[0][0][1], + batch->macroblocks[i].PMV[1][0][0], + batch->macroblocks[i].PMV[1][0][1], + batch->macroblocks[i].cbp, + batch->macroblocks[i].dct_type, + batch->macroblocks[i].blocks, + batch->past_surface, + surface + ); + break; + } + case vlMacroBlockTypeBkwdPredicted: + { + vlRenderPMacroBlock + ( + mc, + batch->picture_type, + batch->field_order, + batch->macroblocks[i].mbx, + batch->macroblocks[i].mby, + batch->macroblocks[i].mo_type, + batch->macroblocks[i].PMV[0][1][0], + batch->macroblocks[i].PMV[0][1][1], + batch->macroblocks[i].PMV[1][1][0], + batch->macroblocks[i].PMV[1][1][1], + batch->macroblocks[i].cbp, + batch->macroblocks[i].dct_type, + batch->macroblocks[i].blocks, + batch->future_surface, + surface + ); + break; + } + case vlMacroBlockTypeBiPredicted: + { + vlRenderBMacroBlock + ( + mc, + batch->picture_type, + batch->field_order, + batch->macroblocks[i].mbx, + batch->macroblocks[i].mby, + batch->macroblocks[i].mo_type, + batch->macroblocks[i].PMV[0][0][0], + batch->macroblocks[i].PMV[0][0][1], + batch->macroblocks[i].PMV[1][0][0], + batch->macroblocks[i].PMV[1][0][1], + batch->macroblocks[i].PMV[0][1][0], + batch->macroblocks[i].PMV[0][1][1], + batch->macroblocks[i].PMV[1][1][0], + batch->macroblocks[i].PMV[1][1][1], + batch->macroblocks[i].cbp, + batch->macroblocks[i].dct_type, + batch->macroblocks[i].blocks, + batch->past_surface, + batch->future_surface, + surface + ); + break; + } + default: + assert(0); + } + } + + return 0; +} + +static int vlEnd +( + struct vlRender *render +) +{ + assert(render); + + return 0; +} + +static int vlFlush +( + struct vlRender *render +) +{ + assert(render); + + return 0; +} + +static int vlDestroy +( + struct vlRender *render +) +{ + struct vlR16SnormMC *mc; + struct pipe_context *pipe; + unsigned int i; + + assert(render); + + mc = (struct vlR16SnormMC*)render; + pipe = mc->pipe; + + for (i = 0; i < 5; ++i) + pipe->delete_sampler_state(pipe, mc->samplers[i]); + + for (i = 0; i < 3; ++i) + pipe->winsys->buffer_destroy(pipe->winsys, mc->vertex_bufs[i].buffer); + + /* Textures 3 & 4 are not created directly, no need to release them here */ + for (i = 0; i < NUM_BUFS; ++i) + { + pipe_texture_release(&mc->textures[i][0]); + pipe_texture_release(&mc->textures[i][1]); + pipe_texture_release(&mc->textures[i][2]); + } + + pipe->delete_vs_state(pipe, mc->i_vs); + pipe->delete_fs_state(pipe, mc->i_fs); + + for (i = 0; i < 2; ++i) + { + pipe->delete_vs_state(pipe, mc->p_vs[i]); + pipe->delete_fs_state(pipe, mc->p_fs[i]); + pipe->delete_vs_state(pipe, mc->b_vs[i]); + pipe->delete_fs_state(pipe, mc->b_fs[i]); + } + + pipe->winsys->buffer_destroy(pipe->winsys, mc->vs_const_buf.buffer); + pipe->winsys->buffer_destroy(pipe->winsys, mc->fs_const_buf.buffer); + + free(mc); + + return 0; +} + +/* + * Represents 8 triangles (4 quads, 1 per block) in noormalized coords + * that render a macroblock. + * Need to be scaled to cover mbW*mbH macroblock pixels and translated into + * position on target surface. + */ +static const struct vlVertex2f macroblock_verts[24] = +{ + {0.0f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.0f}, + {0.5f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.5f}, + + {0.5f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.0f}, + {1.0f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.5f}, + + {0.0f, 0.5f}, {0.0f, 1.0f}, {0.5f, 0.5f}, + {0.5f, 0.5f}, {0.0f, 1.0f}, {0.5f, 1.0f}, + + {0.5f, 0.5f}, {0.5f, 1.0f}, {1.0f, 0.5f}, + {1.0f, 0.5f}, {0.5f, 1.0f}, {1.0f, 1.0f} +}; + +/* + * Represents texcoords for the above for rendering 4 luma blocks arranged + * in a bW*(bH*4) texture. First luma block located at 0,0->bW,bH; second at + * 0,bH->bW,2bH; third at 0,2bH->bW,3bH; fourth at 0,3bH->bW,4bH. + */ +static const struct vlVertex2f macroblock_luma_texcoords[24] = +{ + {0.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.0f}, + {1.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.25f}, + + {0.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.25f}, + {1.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.5f}, + + {0.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.5f}, + {1.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.75f}, + + {0.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 0.75f}, + {1.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 1.0f} +}; + +/* + * Represents texcoords for the above for rendering 1 chroma block. + * Straight forward 0,0->1,1 mapping so we can reuse the MB pos vectors. + */ +static const struct vlVertex2f *macroblock_chroma_420_texcoords = macroblock_verts; + +/* + * Represents texcoords for the above for rendering 2 chroma blocks arranged + * in a bW*(bH*2) texture. First chroma block located at 0,0->bW,bH; second at + * 0,bH->bW,2bH. We can render this with 0,0->1,1 mapping. + * Straight forward 0,0->1,1 mapping so we can reuse MB pos vectors. + */ +static const struct vlVertex2f *macroblock_chroma_422_texcoords = macroblock_verts; + +/* + * Represents texcoords for the above for rendering 4 chroma blocks. + * Same case as 4 luma blocks. + */ +static const struct vlVertex2f *macroblock_chroma_444_texcoords = macroblock_luma_texcoords; + +/* + * Used when rendering P and B macroblocks, multiplier is applied to the A channel, + * which is then added to the L channel, then the bias is subtracted from that to + * get back the differential. The differential is then added to the samples from the + * reference surface(s). + */ +static const struct vlFragmentShaderConsts fs_consts = +{ + {32767.0f / 255.0f, 32767.0f / 255.0f, 32767.0f / 255.0f, 0.0f}, + {0.5f, 2.0f, 0.0f, 0.0f} +}; + +static int vlCreateVertexShaderIMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 50; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma texcoords + */ + for (i = 0; i < 3; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling vector to scale unit rect to macroblock size + * decl c1 ; Translation vector to move macroblock into position + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma texcoords + */ + for (i = 0; i < 3; i++) + { + decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl t0 */ + decl = vl_decl_temps(0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* mul t0, i0, c0 ; Scale unit rect to normalized MB size */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, c1 ; Translate rect into position */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * mov o1, i1 ; Move input luma texcoords to output + * mov o2, i2 ; Move input chroma texcoords to output + */ + for (i = 1; i < 3; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->i_vs = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderIMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Texcoords for s0 + * decl i1 ; Texcoords for s1, s2 + */ + for (i = 0; i < 2; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0, t1 */ + decl = vl_decl_temps(0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + */ + for (i = 0; i < 3; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header,max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i1, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + } + + /* mul o0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->i_fs = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateVertexShaderFramePMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma texcoords + */ + for (i = 0; i < 3; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling vector to scale unit rect to macroblock size + * decl c1 ; Translation vector to move macroblock into position + * decl c2 ; Unused + * decl c3 ; Translation vector to move ref macroblock texcoords into position + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 3); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma texcoords + * decl o3 ; Ref macroblock texcoords + */ + for (i = 0; i < 4; i++) + { + decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl t0 */ + decl = vl_decl_temps(0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* mul t0, i0, c0 ; Scale unit rect to normalized MB size */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, c1 ; Translate rect into position */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * mov o1, i1 ; Move input luma texcoords to output + * mov o2, i2 ; Move input chroma texcoords to output + */ + for (i = 1; i < 3; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* add o3, t0, c3 ; Translate rect into position on ref macroblock */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->p_vs[0] = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateVertexShaderFieldPMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma texcoords + */ + for (i = 0; i < 3; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration + ( + &decl, + &tokens[ti], + header, + max_tokens - ti + ); + } + + /* + * decl c0 ; Scaling vector to scale unit rect to macroblock size + * decl c1 ; Translation vector to move macroblock into position + * decl c2 ; Denorm coefficients + * decl c3 ; Translation vector to move top field ref macroblock texcoords into position + * decl c4 ; Translation vector to move bottom field ref macroblock texcoords into position + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma texcoords + * decl o3 ; Top field ref macroblock texcoords + * decl o4 ; Bottom field ref macroblock texcoords + * decl o5 ; Denormalized vertex pos + */ + for (i = 0; i < 6; i++) + { + decl = vl_decl_output((i == 0 || i == 5) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl t0, t1 */ + decl = vl_decl_temps(0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* mul t0, i0, c0 ; Scale unit rect to normalized MB size */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add t1, t0, c1 ; Translate rect into position */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mov o0, t1 ; Move vertex pos to output */ + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + mov o1, i1 ; Move input luma texcoords to output + mov o2, i2 ; Move input chroma texcoords to output + */ + for (i = 1; i < 3; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* add o3, t0, c3 ; Translate top field rect into position on ref macroblock + add o4, t0, c4 ; Translate bottom field rect into position on ref macroblock */ + for (i = 0; i < 2; ++i) + { + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* mul o5, t1, c2 ; Denorm vertex pos */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 5, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->p_vs[1] = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderFramePMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Texcoords for s0 + * decl i1 ; Texcoords for s1, s2 + * decl i2 ; Texcoords for s3 + */ + for (i = 0; i < 3; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0, t1 */ + decl = vl_decl_temps(0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + * decl s3 ; Sampler for ref surface texture + */ + for (i = 0; i < 4; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i1, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + } + + /* mul t0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* tex2d t1, i2, s3 ; Read texel from ref macroblock */ + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 2, TGSI_FILE_SAMPLER, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, t1 ; Add ref and differential to form final output */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->p_fs[0] = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderFieldPMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 200; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Texcoords for s0 + * decl i1 ; Texcoords for s1, s2 + * decl i2 ; Texcoords for s3 + * decl i3 ; Texcoords for s3 + * decl i4 ; Denormalized vertex pos + */ + for (i = 0; i < 5; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm + * decl c1 ; Constants 1/2 & 2 in .x, .y channels for Y-mod-2 top/bottom field selection + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0-t4 */ + decl = vl_decl_temps(0, 4); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + * decl s3 ; Sampler for ref surface texture + */ + for (i = 0; i < 4; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i1, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + } + + /* mul t0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * tex2d t1, i2, s3 ; Read texel from ref macroblock top field + * tex2d t2, i3, s3 ; Read texel from ref macroblock bottom field + */ + for (i = 0; i < 2; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* XXX: Pos values off by 0.5? */ + /* sub t4, i4.y, c1.x ; Sub 0.5 from denormalized pos */ + inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 4, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mul t3, t4, c1.x ; Multiply pos Y-coord by 1/2 */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* floor t3, t3 ; Get rid of fractional part */ + inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mul t3, t3, c1.y ; Multiply by 2 */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* sub t3, t4, t3 ; Subtract from original Y to get Y % 2 */ + inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* TODO: Move to conditional tex fetch on t3 instead of lerp */ + /* lerp t1, t3, t1, t2 ; Choose between top and bottom fields based on Y % 2 */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, t1 ; Add ref and differential to form final output */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->p_fs[1] = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateVertexShaderFrameBMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma texcoords + */ + for (i = 0; i < 3; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling vector to scale unit rect to macroblock size + * decl c1 ; Translation vector to move macroblock into position + * decl c2 ; Unused + * decl c3 ; Translation vector to move past ref macroblock texcoords into position + * decl c4 ; Unused + * decl c5 ; Translation vector to move future ref macroblock texcoords into position + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 5); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma texcoords + * decl o3 ; Past ref macroblock texcoords + * decl o4 ; Future ref macroblock texcoords + */ + for (i = 0; i < 5; i++) + { + decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl t0 */ + decl = vl_decl_temps(0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* mul t0, i0, c0 ; Scale unit rect to normalized MB size */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, c1 ; Translate rect into position */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * mov o1, i1 ; Move input luma texcoords to output + * mov o2, i2 ; Move input chroma texcoords to output + */ + for (i = 1; i < 3; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* add o3, t0, c3 ; Translate rect into position on past ref macroblock + add o4, t0, c5 ; Translate rect into position on future ref macroblock */ + for (i = 0; i < 2; ++i) + { + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i * 2 + 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->b_vs[0] = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateVertexShaderFieldBMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma texcoords + */ + for (i = 0; i < 3; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling vector to scale unit rect to macroblock size + * decl c1 ; Translation vector to move macroblock into position + * decl c2 ; Denorm coefficients + * decl c3 ; Translation vector to move top field past ref macroblock texcoords into position + * decl c4 ; Translation vector to move bottom field past ref macroblock texcoords into position + * decl c5 ; Translation vector to move top field future ref macroblock texcoords into position + * decl c6 ; Translation vector to move bottom field future ref macroblock texcoords into position + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 6); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma texcoords + * decl o3 ; Top field past ref macroblock texcoords + * decl o4 ; Bottom field past ref macroblock texcoords + * decl o5 ; Top field future ref macroblock texcoords + * decl o6 ; Bottom field future ref macroblock texcoords + * decl o7 ; Denormalized vertex pos + */ + for (i = 0; i < 8; i++) + { + decl = vl_decl_output((i == 0 || i == 7) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl t0, t1 */ + decl = vl_decl_temps(0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* mul t0, i0, c0 ; Scale unit rect to normalized MB size */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add t1, t0, c1 ; Translate rect into position */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mov o0, t1 ; Move vertex pos to output */ + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * mov o1, i1 ; Move input luma texcoords to output + * mov o2, i2 ; Move input chroma texcoords to output + */ + for (i = 1; i < 3; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* + * add o3, t0, c3 ; Translate top field rect into position on past ref macroblock + * add o4, t0, c4 ; Translate bottom field rect into position on past ref macroblock + * add o5, t0, c5 ; Translate top field rect into position on future ref macroblock + * add o6, t0, c6 ; Translate bottom field rect into position on future ref macroblock + */ + for (i = 0; i < 4; ++i) + { + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* mul o7, t1, c2 ; Denorm vertex pos */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 7, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->b_vs[1] = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderFrameBMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Texcoords for s0 + * decl i1 ; Texcoords for s1, s2 + * decl i2 ; Texcoords for s3 + * decl i3 ; Texcoords for s4 + */ + for (i = 0; i < 4; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm + * decl c1 ; Constant 1/2 in .x channel to use as weight to blend past and future texels + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0-t2 */ + decl = vl_decl_temps(0, 2); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + * decl s3 ; Sampler for past ref surface texture + * decl s4 ; Sampler for future ref surface texture + */ + for (i = 0; i < 5; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i1, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + } + + /* mul t0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * tex2d t1, i2, s3 ; Read texel from past ref macroblock + * tex2d t2, i3, s4 ; Read texel from future ref macroblock + */ + for (i = 0; i < 2; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, i + 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* lerp t1, c1.x, t1, t2 ; Blend past and future texels */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, t1 ; Add past/future ref and differential to form final output */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->b_fs[0] = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderFieldBMB +( + struct vlR16SnormMC *mc +) +{ + const unsigned int max_tokens = 200; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Texcoords for s0 + * decl i1 ; Texcoords for s1, s2 + * decl i2 ; Texcoords for s3 + * decl i3 ; Texcoords for s3 + * decl i4 ; Texcoords for s4 + * decl i5 ; Texcoords for s4 + * decl i6 ; Denormalized vertex pos + */ + for (i = 0; i < 7; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm + * decl c1 ; Constants 1/2 & 2 in .x, .y channels to use as weight to blend past and future texels + * ; and for Y-mod-2 top/bottom field selection + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0-t5 */ + decl = vl_decl_temps(0, 5); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + * decl s3 ; Sampler for past ref surface texture + * decl s4 ; Sampler for future ref surface texture + */ + for (i = 0; i < 5; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i1, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + } + + /* mul t0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* XXX: Pos values off by 0.5? */ + /* sub t4, i6.y, c1.x ; Sub 0.5 from denormalized pos */ + inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 6, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mul t3, t4, c1.x ; Multiply pos Y-coord by 1/2 */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* floor t3, t3 ; Get rid of fractional part */ + inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mul t3, t3, c1.y ; Multiply by 2 */ + inst = vl_inst3( TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* sub t3, t4, t3 ; Subtract from original Y to get Y % 2 */ + inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * tex2d t1, i2, s3 ; Read texel from past ref macroblock top field + * tex2d t2, i3, s3 ; Read texel from past ref macroblock bottom field + */ + for (i = 0; i < 2; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* TODO: Move to conditional tex fetch on t3 instead of lerp */ + /* lerp t1, t3, t1, t2 ; Choose between top and bottom fields based on Y % 2 */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * tex2d t4, i4, s4 ; Read texel from future ref macroblock top field + * tex2d t5, i5, s4 ; Read texel from future ref macroblock bottom field + */ + for (i = 0; i < 2; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 4, TGSI_FILE_INPUT, i + 4, TGSI_FILE_SAMPLER, 4); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* TODO: Move to conditional tex fetch on t3 instead of lerp */ + /* lerp t2, t3, t4, t5 ; Choose between top and bottom fields based on Y % 2 */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 2, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 5); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* lerp t1, c1.x, t1, t2 ; Blend past and future texels */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, t1 ; Add past/future ref and differential to form final output */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->b_fs[1] = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateDataBufs +( + struct vlR16SnormMC *mc +) +{ + struct pipe_context *pipe; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + + /* Create our vertex buffer and vertex buffer element */ + mc->vertex_bufs[0].pitch = sizeof(struct vlVertex2f); + mc->vertex_bufs[0].max_index = 23; + mc->vertex_bufs[0].buffer_offset = 0; + mc->vertex_bufs[0].buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(struct vlVertex2f) * 24 + ); + + mc->vertex_elems[0].src_offset = 0; + mc->vertex_elems[0].vertex_buffer_index = 0; + mc->vertex_elems[0].nr_components = 2; + mc->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* Create our texcoord buffers and texcoord buffer elements */ + for (i = 1; i < 3; ++i) + { + mc->vertex_bufs[i].pitch = sizeof(struct vlVertex2f); + mc->vertex_bufs[i].max_index = 23; + mc->vertex_bufs[i].buffer_offset = 0; + mc->vertex_bufs[i].buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(struct vlVertex2f) * 24 + ); + + mc->vertex_elems[i].src_offset = 0; + mc->vertex_elems[i].vertex_buffer_index = i; + mc->vertex_elems[i].nr_components = 2; + mc->vertex_elems[i].src_format = PIPE_FORMAT_R32G32_FLOAT; + } + + /* Fill buffers */ + memcpy + ( + pipe->winsys->buffer_map(pipe->winsys, mc->vertex_bufs[0].buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + macroblock_verts, + sizeof(struct vlVertex2f) * 24 + ); + memcpy + ( + pipe->winsys->buffer_map(pipe->winsys, mc->vertex_bufs[1].buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + macroblock_luma_texcoords, + sizeof(struct vlVertex2f) * 24 + ); + /* TODO: Accomodate 422, 444 */ + memcpy + ( + pipe->winsys->buffer_map(pipe->winsys, mc->vertex_bufs[2].buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + macroblock_chroma_420_texcoords, + sizeof(struct vlVertex2f) * 24 + ); + + for (i = 0; i < 3; ++i) + pipe->winsys->buffer_unmap(pipe->winsys, mc->vertex_bufs[i].buffer); + + /* Create our constant buffer */ + mc->vs_const_buf.size = sizeof(struct vlVertexShaderConsts); + mc->vs_const_buf.buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_CONSTANT, + mc->vs_const_buf.size + ); + + mc->fs_const_buf.size = sizeof(struct vlFragmentShaderConsts); + mc->fs_const_buf.buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_CONSTANT, + mc->fs_const_buf.size + ); + + memcpy + ( + pipe->winsys->buffer_map(pipe->winsys, mc->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + &fs_consts, + sizeof(struct vlFragmentShaderConsts) + ); + + pipe->winsys->buffer_unmap(pipe->winsys, mc->fs_const_buf.buffer); + + return 0; +} + +static int vlInit +( + struct vlR16SnormMC *mc +) +{ + struct pipe_context *pipe; + struct pipe_sampler_state sampler; + struct pipe_texture template; + unsigned int filters[5]; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + + /* For MC we render to textures, which are rounded up to nearest POT */ + mc->viewport.scale[0] = vlRoundUpPOT(mc->video_width); + mc->viewport.scale[1] = vlRoundUpPOT(mc->video_height); + mc->viewport.scale[2] = 1; + mc->viewport.scale[3] = 1; + mc->viewport.translate[0] = 0; + mc->viewport.translate[1] = 0; + mc->viewport.translate[2] = 0; + mc->viewport.translate[3] = 0; + + mc->render_target.width = vlRoundUpPOT(mc->video_width); + mc->render_target.height = vlRoundUpPOT(mc->video_height); + mc->render_target.num_cbufs = 1; + /* FB for MC stage is a vlSurface, set in vlSetRenderSurface() */ + mc->render_target.zsbuf = NULL; + + filters[0] = PIPE_TEX_FILTER_NEAREST; + filters[1] = mc->video_format == vlFormatYCbCr444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR; + filters[2] = mc->video_format == vlFormatYCbCr444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR; + filters[3] = PIPE_TEX_FILTER_LINEAR; + filters[4] = PIPE_TEX_FILTER_LINEAR; + + for (i = 0; i < 5; ++i) + { + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_img_filter = filters[i]; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.mag_img_filter = filters[i]; + sampler.compare_mode = PIPE_TEX_COMPARE_NONE; + sampler.compare_func = PIPE_FUNC_ALWAYS; + sampler.normalized_coords = 1; + /*sampler.prefilter = ;*/ + /*sampler.shadow_ambient = ;*/ + /*sampler.lod_bias = ;*/ + sampler.min_lod = 0; + /*sampler.max_lod = ;*/ + /*sampler.border_color[i] = ;*/ + /*sampler.max_anisotropy = ;*/ + mc->samplers[i] = pipe->create_sampler_state(pipe, &sampler); + } + + memset(&template, 0, sizeof(struct pipe_texture)); + template.target = PIPE_TEXTURE_2D; + template.format = PIPE_FORMAT_R16_SNORM; + template.last_level = 0; + template.width[0] = 8; + template.height[0] = 8 * 4; + template.depth[0] = 1; + template.compressed = 0; + pf_get_block(template.format, &template.block); + + for (i = 0; i < NUM_BUFS; ++i) + mc->textures[i][0] = pipe->screen->texture_create(pipe->screen, &template); + + if (mc->video_format == vlFormatYCbCr420) + template.height[0] = 8; + else if (mc->video_format == vlFormatYCbCr422) + template.height[0] = 8 * 2; + else if (mc->video_format == vlFormatYCbCr444) + template.height[0] = 8 * 4; + else + assert(0); + + for (i = 0; i < NUM_BUFS; ++i) + { + mc->textures[i][1] = pipe->screen->texture_create(pipe->screen, &template); + mc->textures[i][2] = pipe->screen->texture_create(pipe->screen, &template); + } + + /* textures[3] & textures[4] are assigned from vlSurfaces for P and B macroblocks at render time */ + + vlCreateVertexShaderIMB(mc); + vlCreateFragmentShaderIMB(mc); + vlCreateVertexShaderFramePMB(mc); + vlCreateVertexShaderFieldPMB(mc); + vlCreateFragmentShaderFramePMB(mc); + vlCreateFragmentShaderFieldPMB(mc); + vlCreateVertexShaderFrameBMB(mc); + vlCreateVertexShaderFieldBMB(mc); + vlCreateFragmentShaderFrameBMB(mc); + vlCreateFragmentShaderFieldBMB(mc); + vlCreateDataBufs(mc); + + return 0; +} + +int vlCreateR16SNormMC +( + struct pipe_context *pipe, + unsigned int video_width, + unsigned int video_height, + enum vlFormat video_format, + struct vlRender **render +) +{ + struct vlR16SnormMC *mc; + + assert(pipe); + assert(render); + + mc = calloc(1, sizeof(struct vlR16SnormMC)); + + mc->base.vlBegin = &vlBegin; + mc->base.vlRenderMacroBlocksMpeg2 = &vlRenderMacroBlocksMpeg2R16Snorm; + mc->base.vlEnd = &vlEnd; + mc->base.vlFlush = &vlFlush; + mc->base.vlDestroy = &vlDestroy; + mc->pipe = pipe; + mc->video_width = video_width; + mc->video_height = video_height; + mc->cur_buf = 0; + + vlInit(mc); + + *render = &mc->base; + + return 0; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h new file mode 100644 index 0000000000..9842926bf7 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h @@ -0,0 +1,18 @@ +#ifndef vl_r16snorm_mc_h +#define vl_r16snorm_mc_h + +#include "vl_types.h" + +struct pipe_context; +struct vlRender; + +int vlCreateR16SNormMC +( + struct pipe_context *pipe, + unsigned int video_width, + unsigned int video_height, + enum vlFormat video_format, + struct vlRender **render +); + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c new file mode 100644 index 0000000000..650528ed8f --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c @@ -0,0 +1,1147 @@ +#define VL_INTERNAL +#include "vl_r16snorm_mc_buf.h" +#include <assert.h> +#include <stdlib.h> +#include <pipe/p_context.h> +#include <pipe/p_winsys.h> +#include <pipe/p_screen.h> +#include <pipe/p_state.h> +#include <pipe/p_inlines.h> +#include <tgsi/tgsi_parse.h> +#include <tgsi/tgsi_build.h> +#include <util/u_math.h> +#include "vl_render.h" +#include "vl_shader_build.h" +#include "vl_surface.h" +#include "vl_util.h" +#include "vl_types.h" +#include "vl_defs.h" + +/* + * TODO: Dynamically determine number of buf sets to use, based on + * video size and available mem, since we can easily run out of memory + * for high res videos. + * Note: Destroying previous frame's buffers and creating new ones + * doesn't work, since the buffer are not actually destroyed until their + * fence is signalled, and if we render fast enough we will create faster + * than we destroy. + */ +#define NUM_BUF_SETS 4 /* Number of rotating buffer sets to use */ + +enum vlMacroBlockTypeEx +{ + vlMacroBlockExTypeIntra, + vlMacroBlockExTypeFwdPredictedFrame, + vlMacroBlockExTypeFwdPredictedField, + vlMacroBlockExTypeBkwdPredictedFrame, + vlMacroBlockExTypeBkwdPredictedField, + vlMacroBlockExTypeBiPredictedFrame, + vlMacroBlockExTypeBiPredictedField, + + vlNumMacroBlockExTypes +}; + +struct vlVertexShaderConsts +{ + struct vlVertex4f denorm; +}; + +struct vlFragmentShaderConsts +{ + struct vlVertex4f multiplier; + struct vlVertex4f div; +}; + +struct vlR16SnormBufferedMC +{ + struct vlRender base; + + unsigned int picture_width, picture_height; + enum vlFormat picture_format; + + unsigned int cur_buf; + struct vlSurface *buffered_surface; + struct vlSurface *past_surface, *future_surface; + struct vlVertex2f surface_tex_inv_size; + struct vlVertex2f zero_block[3]; + unsigned int num_macroblocks; + struct vlMpeg2MacroBlock *macroblocks; + + struct pipe_context *pipe; + struct pipe_viewport_state viewport; + struct pipe_framebuffer_state render_target; + struct pipe_sampler_state *samplers[5]; + struct pipe_texture *textures[NUM_BUF_SETS][5]; + struct pipe_surface *tex_surface[3]; + short *texels[3]; + void *i_vs, *p_vs[2], *b_vs[2]; + void *i_fs, *p_fs[2], *b_fs[2]; + struct pipe_vertex_buffer vertex_bufs[NUM_BUF_SETS][3]; + struct pipe_vertex_element vertex_elems[8]; + struct pipe_constant_buffer vs_const_buf, fs_const_buf; +}; + +static int vlBegin +( + struct vlRender *render +) +{ + assert(render); + + return 0; +} + +static inline int vlGrabFrameCodedBlock(short *src, short *dst, unsigned int dst_pitch) +{ + unsigned int y; + + for (y = 0; y < VL_BLOCK_HEIGHT; ++y) + memcpy + ( + dst + y * dst_pitch, + src + y * VL_BLOCK_WIDTH, + VL_BLOCK_WIDTH * 2 + ); + + return 0; +} + +static inline int vlGrabFieldCodedBlock(short *src, short *dst, unsigned int dst_pitch) +{ + unsigned int y; + + for (y = 0; y < VL_BLOCK_HEIGHT; ++y) + memcpy + ( + dst + y * dst_pitch * 2, + src + y * VL_BLOCK_WIDTH, + VL_BLOCK_WIDTH * 2 + ); + + return 0; +} + +static inline int vlGrabNoBlock(short *dst, unsigned int dst_pitch) +{ + unsigned int y; + + for (y = 0; y < VL_BLOCK_HEIGHT; ++y) + memset + ( + dst + y * dst_pitch, + 0, + VL_BLOCK_WIDTH * 2 + ); + + return 0; +} + +static inline int vlGrabBlocks +( + struct vlR16SnormBufferedMC *mc, + unsigned int mbx, + unsigned int mby, + enum vlDCTType dct_type, + unsigned int coded_block_pattern, + short *blocks +) +{ + short *texels; + unsigned int tex_pitch; + unsigned int x, y, tb = 0, sb = 0; + unsigned int mbpx = mbx * VL_MACROBLOCK_WIDTH, mbpy = mby * VL_MACROBLOCK_HEIGHT; + + assert(mc); + assert(blocks); + + tex_pitch = mc->tex_surface[0]->stride / mc->tex_surface[0]->block.size; + texels = mc->texels[0] + mbpy * tex_pitch + mbpx; + + for (y = 0; y < 2; ++y) + { + for (x = 0; x < 2; ++x, ++tb) + { + if ((coded_block_pattern >> (5 - tb)) & 1) + { + short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT; + + if (dct_type == vlDCTTypeFrameCoded) + { + vlGrabFrameCodedBlock + ( + cur_block, + texels + y * tex_pitch * VL_BLOCK_HEIGHT + x * VL_BLOCK_WIDTH, + tex_pitch + ); + } + else + { + vlGrabFieldCodedBlock + ( + cur_block, + texels + y * tex_pitch + x * VL_BLOCK_WIDTH, + tex_pitch + ); + } + + ++sb; + } + else if (mc->zero_block[0].x < 0.0f) + { + vlGrabNoBlock(texels + y * tex_pitch * VL_BLOCK_HEIGHT + x * VL_BLOCK_WIDTH, tex_pitch); + + mc->zero_block[0].x = (mbpx + x * 8) * mc->surface_tex_inv_size.x; + mc->zero_block[0].y = (mbpy + y * 8) * mc->surface_tex_inv_size.y; + } + } + } + + /* TODO: Implement 422, 444 */ + mbpx >>= 1; + mbpy >>= 1; + + for (tb = 0; tb < 2; ++tb) + { + tex_pitch = mc->tex_surface[tb + 1]->stride / mc->tex_surface[tb + 1]->block.size; + texels = mc->texels[tb + 1] + mbpy * tex_pitch + mbpx; + + if ((coded_block_pattern >> (1 - tb)) & 1) + { + short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT; + + vlGrabFrameCodedBlock + ( + cur_block, + texels, + tex_pitch + ); + + ++sb; + } + else if (mc->zero_block[tb + 1].x < 0.0f) + { + vlGrabNoBlock(texels, tex_pitch); + + mc->zero_block[tb + 1].x = (mbpx << 1) * mc->surface_tex_inv_size.x; + mc->zero_block[tb + 1].y = (mbpy << 1) * mc->surface_tex_inv_size.y; + } + } + + return 0; +} + +static inline enum vlMacroBlockTypeEx vlGetMacroBlockTypeEx(struct vlMpeg2MacroBlock *mb) +{ + assert(mb); + + switch (mb->mb_type) + { + case vlMacroBlockTypeIntra: + return vlMacroBlockExTypeIntra; + case vlMacroBlockTypeFwdPredicted: + return mb->mo_type == vlMotionTypeFrame ? + vlMacroBlockExTypeFwdPredictedFrame : vlMacroBlockExTypeFwdPredictedField; + case vlMacroBlockTypeBkwdPredicted: + return mb->mo_type == vlMotionTypeFrame ? + vlMacroBlockExTypeBkwdPredictedFrame : vlMacroBlockExTypeBkwdPredictedField; + case vlMacroBlockTypeBiPredicted: + return mb->mo_type == vlMotionTypeFrame ? + vlMacroBlockExTypeBiPredictedFrame : vlMacroBlockExTypeBiPredictedField; + default: + assert(0); + } + + /* Unreachable */ + return -1; +} + +static inline int vlGrabMacroBlock +( + struct vlR16SnormBufferedMC *mc, + struct vlMpeg2MacroBlock *macroblock +) +{ + assert(mc); + assert(macroblock); + + mc->macroblocks[mc->num_macroblocks].mbx = macroblock->mbx; + mc->macroblocks[mc->num_macroblocks].mby = macroblock->mby; + mc->macroblocks[mc->num_macroblocks].mb_type = macroblock->mb_type; + mc->macroblocks[mc->num_macroblocks].mo_type = macroblock->mo_type; + mc->macroblocks[mc->num_macroblocks].dct_type = macroblock->dct_type; + mc->macroblocks[mc->num_macroblocks].PMV[0][0][0] = macroblock->PMV[0][0][0]; + mc->macroblocks[mc->num_macroblocks].PMV[0][0][1] = macroblock->PMV[0][0][1]; + mc->macroblocks[mc->num_macroblocks].PMV[0][1][0] = macroblock->PMV[0][1][0]; + mc->macroblocks[mc->num_macroblocks].PMV[0][1][1] = macroblock->PMV[0][1][1]; + mc->macroblocks[mc->num_macroblocks].PMV[1][0][0] = macroblock->PMV[1][0][0]; + mc->macroblocks[mc->num_macroblocks].PMV[1][0][1] = macroblock->PMV[1][0][1]; + mc->macroblocks[mc->num_macroblocks].PMV[1][1][0] = macroblock->PMV[1][1][0]; + mc->macroblocks[mc->num_macroblocks].PMV[1][1][1] = macroblock->PMV[1][1][1]; + mc->macroblocks[mc->num_macroblocks].cbp = macroblock->cbp; + mc->macroblocks[mc->num_macroblocks].blocks = macroblock->blocks; + + vlGrabBlocks + ( + mc, + macroblock->mbx, + macroblock->mby, + macroblock->dct_type, + macroblock->cbp, + macroblock->blocks + ); + + mc->num_macroblocks++; + + return 0; +} + +#define SET_BLOCK(vb, cbp, mbx, mby, unitx, unity, ofsx, ofsy, hx, hy, lm, cbm, crm, zb) \ + (vb)[0].pos.x = (mbx) * (unitx) + (ofsx); (vb)[0].pos.y = (mby) * (unity) + (ofsy); \ + (vb)[1].pos.x = (mbx) * (unitx) + (ofsx); (vb)[1].pos.y = (mby) * (unity) + (ofsy) + (hy); \ + (vb)[2].pos.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[2].pos.y = (mby) * (unity) + (ofsy); \ + (vb)[3].pos.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[3].pos.y = (mby) * (unity) + (ofsy); \ + (vb)[4].pos.x = (mbx) * (unitx) + (ofsx); (vb)[4].pos.y = (mby) * (unity) + (ofsy) + (hy); \ + (vb)[5].pos.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[5].pos.y = (mby) * (unity) + (ofsy) + (hy); \ + \ + if ((cbp) & (lm)) \ + { \ + (vb)[0].luma_tc.x = (mbx) * (unitx) + (ofsx); (vb)[0].luma_tc.y = (mby) * (unity) + (ofsy); \ + (vb)[1].luma_tc.x = (mbx) * (unitx) + (ofsx); (vb)[1].luma_tc.y = (mby) * (unity) + (ofsy) + (hy); \ + (vb)[2].luma_tc.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[2].luma_tc.y = (mby) * (unity) + (ofsy); \ + (vb)[3].luma_tc.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[3].luma_tc.y = (mby) * (unity) + (ofsy); \ + (vb)[4].luma_tc.x = (mbx) * (unitx) + (ofsx); (vb)[4].luma_tc.y = (mby) * (unity) + (ofsy) + (hy); \ + (vb)[5].luma_tc.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[5].luma_tc.y = (mby) * (unity) + (ofsy) + (hy); \ + } \ + else \ + { \ + (vb)[0].luma_tc.x = (zb)[0].x; (vb)[0].luma_tc.y = (zb)[0].y; \ + (vb)[1].luma_tc.x = (zb)[0].x; (vb)[1].luma_tc.y = (zb)[0].y + (hy); \ + (vb)[2].luma_tc.x = (zb)[0].x + (hx); (vb)[2].luma_tc.y = (zb)[0].y; \ + (vb)[3].luma_tc.x = (zb)[0].x + (hx); (vb)[3].luma_tc.y = (zb)[0].y; \ + (vb)[4].luma_tc.x = (zb)[0].x; (vb)[4].luma_tc.y = (zb)[0].y + (hy); \ + (vb)[5].luma_tc.x = (zb)[0].x + (hx); (vb)[5].luma_tc.y = (zb)[0].y + (hy); \ + } \ + \ + if ((cbp) & (cbm)) \ + { \ + (vb)[0].cb_tc.x = (mbx) * (unitx) + (ofsx); (vb)[0].cb_tc.y = (mby) * (unity) + (ofsy); \ + (vb)[1].cb_tc.x = (mbx) * (unitx) + (ofsx); (vb)[1].cb_tc.y = (mby) * (unity) + (ofsy) + (hy); \ + (vb)[2].cb_tc.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[2].cb_tc.y = (mby) * (unity) + (ofsy); \ + (vb)[3].cb_tc.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[3].cb_tc.y = (mby) * (unity) + (ofsy); \ + (vb)[4].cb_tc.x = (mbx) * (unitx) + (ofsx); (vb)[4].cb_tc.y = (mby) * (unity) + (ofsy) + (hy); \ + (vb)[5].cb_tc.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[5].cb_tc.y = (mby) * (unity) + (ofsy) + (hy); \ + } \ + else \ + { \ + (vb)[0].cb_tc.x = (zb)[1].x; (vb)[0].cb_tc.y = (zb)[1].y; \ + (vb)[1].cb_tc.x = (zb)[1].x; (vb)[1].cb_tc.y = (zb)[1].y + (hy); \ + (vb)[2].cb_tc.x = (zb)[1].x + (hx); (vb)[2].cb_tc.y = (zb)[1].y; \ + (vb)[3].cb_tc.x = (zb)[1].x + (hx); (vb)[3].cb_tc.y = (zb)[1].y; \ + (vb)[4].cb_tc.x = (zb)[1].x; (vb)[4].cb_tc.y = (zb)[1].y + (hy); \ + (vb)[5].cb_tc.x = (zb)[1].x + (hx); (vb)[5].cb_tc.y = (zb)[1].y + (hy); \ + } \ + \ + if ((cbp) & (crm)) \ + { \ + (vb)[0].cr_tc.x = (mbx) * (unitx) + (ofsx); (vb)[0].cr_tc.y = (mby) * (unity) + (ofsy); \ + (vb)[1].cr_tc.x = (mbx) * (unitx) + (ofsx); (vb)[1].cr_tc.y = (mby) * (unity) + (ofsy) + (hy); \ + (vb)[2].cr_tc.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[2].cr_tc.y = (mby) * (unity) + (ofsy); \ + (vb)[3].cr_tc.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[3].cr_tc.y = (mby) * (unity) + (ofsy); \ + (vb)[4].cr_tc.x = (mbx) * (unitx) + (ofsx); (vb)[4].cr_tc.y = (mby) * (unity) + (ofsy) + (hy); \ + (vb)[5].cr_tc.x = (mbx) * (unitx) + (ofsx) + (hx); (vb)[5].cr_tc.y = (mby) * (unity) + (ofsy) + (hy); \ + } \ + else \ + { \ + (vb)[0].cr_tc.x = (zb)[2].x; (vb)[0].cr_tc.y = (zb)[2].y; \ + (vb)[1].cr_tc.x = (zb)[2].x; (vb)[1].cr_tc.y = (zb)[2].y + (hy); \ + (vb)[2].cr_tc.x = (zb)[2].x + (hx); (vb)[2].cr_tc.y = (zb)[2].y; \ + (vb)[3].cr_tc.x = (zb)[2].x + (hx); (vb)[3].cr_tc.y = (zb)[2].y; \ + (vb)[4].cr_tc.x = (zb)[2].x; (vb)[4].cr_tc.y = (zb)[2].y + (hy); \ + (vb)[5].cr_tc.x = (zb)[2].x + (hx); (vb)[5].cr_tc.y = (zb)[2].y + (hy); \ + } + +static inline int vlGrabMacroBlockVB +( + struct vlR16SnormBufferedMC *mc, + struct vlMpeg2MacroBlock *macroblock, + unsigned int pos +) +{ + struct vlVertex2f mo_vec[2]; + unsigned int i; + + assert(mc); + assert(macroblock); + + switch (macroblock->mb_type) + { + case vlMacroBlockTypeBiPredicted: + { + struct vlVertex2f *vb; + + vb = (struct vlVertex2f*)mc->pipe->winsys->buffer_map + ( + mc->pipe->winsys, + mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][2].buffer, + PIPE_BUFFER_USAGE_CPU_WRITE + ) + pos * 2 * 24; + + mo_vec[0].x = macroblock->PMV[0][1][0] * 0.5f * mc->surface_tex_inv_size.x; + mo_vec[0].y = macroblock->PMV[0][1][1] * 0.5f * mc->surface_tex_inv_size.y; + + if (macroblock->mo_type == vlMotionTypeFrame) + { + for (i = 0; i < 24 * 2; i += 2) + { + vb[i].x = mo_vec[0].x; + vb[i].y = mo_vec[0].y; + } + } + else + { + mo_vec[1].x = macroblock->PMV[1][1][0] * 0.5f * mc->surface_tex_inv_size.x; + mo_vec[1].y = macroblock->PMV[1][1][1] * 0.5f * mc->surface_tex_inv_size.y; + + for (i = 0; i < 24 * 2; i += 2) + { + vb[i].x = mo_vec[0].x; + vb[i].y = mo_vec[0].y; + vb[i + 1].x = mo_vec[1].x; + vb[i + 1].y = mo_vec[1].y; + } + } + + mc->pipe->winsys->buffer_unmap(mc->pipe->winsys, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][2].buffer); + + /* fall-through */ + } + case vlMacroBlockTypeFwdPredicted: + case vlMacroBlockTypeBkwdPredicted: + { + struct vlVertex2f *vb; + + vb = (struct vlVertex2f*)mc->pipe->winsys->buffer_map + ( + mc->pipe->winsys, + mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][1].buffer, + PIPE_BUFFER_USAGE_CPU_WRITE + ) + pos * 2 * 24; + + if (macroblock->mb_type == vlMacroBlockTypeBkwdPredicted) + { + mo_vec[0].x = macroblock->PMV[0][1][0] * 0.5f * mc->surface_tex_inv_size.x; + mo_vec[0].y = macroblock->PMV[0][1][1] * 0.5f * mc->surface_tex_inv_size.y; + + if (macroblock->mo_type == vlMotionTypeField) + { + mo_vec[1].x = macroblock->PMV[1][1][0] * 0.5f * mc->surface_tex_inv_size.x; + mo_vec[1].y = macroblock->PMV[1][1][1] * 0.5f * mc->surface_tex_inv_size.y; + } + } + else + { + mo_vec[0].x = macroblock->PMV[0][0][0] * 0.5f * mc->surface_tex_inv_size.x; + mo_vec[0].y = macroblock->PMV[0][0][1] * 0.5f * mc->surface_tex_inv_size.y; + + if (macroblock->mo_type == vlMotionTypeField) + { + mo_vec[1].x = macroblock->PMV[1][0][0] * 0.5f * mc->surface_tex_inv_size.x; + mo_vec[1].y = macroblock->PMV[1][0][1] * 0.5f * mc->surface_tex_inv_size.y; + } + } + + if (macroblock->mo_type == vlMotionTypeFrame) + { + for (i = 0; i < 24 * 2; i += 2) + { + vb[i].x = mo_vec[0].x; + vb[i].y = mo_vec[0].y; + } + } + else + { + for (i = 0; i < 24 * 2; i += 2) + { + vb[i].x = mo_vec[0].x; + vb[i].y = mo_vec[0].y; + vb[i + 1].x = mo_vec[1].x; + vb[i + 1].y = mo_vec[1].y; + } + } + + mc->pipe->winsys->buffer_unmap(mc->pipe->winsys, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][1].buffer); + + /* fall-through */ + } + case vlMacroBlockTypeIntra: + { + const struct vlVertex2f unit = + { + mc->surface_tex_inv_size.x * VL_MACROBLOCK_WIDTH, + mc->surface_tex_inv_size.y * VL_MACROBLOCK_HEIGHT + }; + const struct vlVertex2f half = + { + mc->surface_tex_inv_size.x * (VL_MACROBLOCK_WIDTH / 2), + mc->surface_tex_inv_size.y * (VL_MACROBLOCK_HEIGHT / 2) + }; + + struct vlMacroBlockVertexStream0 + { + struct vlVertex2f pos; + struct vlVertex2f luma_tc; + struct vlVertex2f cb_tc; + struct vlVertex2f cr_tc; + } *vb; + + vb = (struct vlMacroBlockVertexStream0*)mc->pipe->winsys->buffer_map + ( + mc->pipe->winsys, + mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][0].buffer, + PIPE_BUFFER_USAGE_CPU_WRITE + ) + pos * 24; + + SET_BLOCK + ( + vb, + macroblock->cbp, macroblock->mbx, macroblock->mby, + unit.x, unit.y, 0, 0, half.x, half.y, + 32, 2, 1, mc->zero_block + ); + + SET_BLOCK + ( + vb + 6, + macroblock->cbp, macroblock->mbx, macroblock->mby, + unit.x, unit.y, half.x, 0, half.x, half.y, + 16, 2, 1, mc->zero_block + ); + + SET_BLOCK + ( + vb + 12, + macroblock->cbp, macroblock->mbx, macroblock->mby, + unit.x, unit.y, 0, half.y, half.x, half.y, + 8, 2, 1, mc->zero_block + ); + + SET_BLOCK + ( + vb + 18, + macroblock->cbp, macroblock->mbx, macroblock->mby, + unit.x, unit.y, half.x, half.y, half.x, half.y, + 4, 2, 1, mc->zero_block + ); + + mc->pipe->winsys->buffer_unmap(mc->pipe->winsys, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][0].buffer); + + break; + } + default: + assert(0); + } + + return 0; +} + +static int vlFlush +( + struct vlRender *render +) +{ + struct vlR16SnormBufferedMC *mc; + struct pipe_context *pipe; + struct vlVertexShaderConsts *vs_consts; + unsigned int num_macroblocks[vlNumMacroBlockExTypes] = {0}; + unsigned int offset[vlNumMacroBlockExTypes]; + unsigned int vb_start = 0; + unsigned int mbw; + unsigned int mbh; + unsigned int num_mb_per_frame; + unsigned int i; + + assert(render); + + mc = (struct vlR16SnormBufferedMC*)render; + + if (!mc->buffered_surface) + return 0; + + mbw = align(mc->picture_width, VL_MACROBLOCK_WIDTH) / VL_MACROBLOCK_WIDTH; + mbh = align(mc->picture_height, VL_MACROBLOCK_HEIGHT) / VL_MACROBLOCK_HEIGHT; + num_mb_per_frame = mbw * mbh; + + if (mc->num_macroblocks < num_mb_per_frame) + return 0; + + pipe = mc->pipe; + + for (i = 0; i < mc->num_macroblocks; ++i) + { + enum vlMacroBlockTypeEx mb_type_ex = vlGetMacroBlockTypeEx(&mc->macroblocks[i]); + + num_macroblocks[mb_type_ex]++; + } + + offset[0] = 0; + + for (i = 1; i < vlNumMacroBlockExTypes; ++i) + offset[i] = offset[i - 1] + num_macroblocks[i - 1]; + + for (i = 0; i < mc->num_macroblocks; ++i) + { + enum vlMacroBlockTypeEx mb_type_ex = vlGetMacroBlockTypeEx(&mc->macroblocks[i]); + + vlGrabMacroBlockVB(mc, &mc->macroblocks[i], offset[mb_type_ex]); + + offset[mb_type_ex]++; + } + + for (i = 0; i < 3; ++i) + { + pipe_surface_unmap(mc->tex_surface[i]); + mc->pipe->screen->tex_surface_release(mc->pipe->screen, &mc->tex_surface[i]); + } + + mc->render_target.cbufs[0] = pipe->screen->get_tex_surface + ( + pipe->screen, + mc->buffered_surface->texture, + 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE + ); + + pipe->set_framebuffer_state(pipe, &mc->render_target); + pipe->set_viewport_state(pipe, &mc->viewport); + vs_consts = pipe->winsys->buffer_map + ( + pipe->winsys, + mc->vs_const_buf.buffer, + PIPE_BUFFER_USAGE_CPU_WRITE + ); + + vs_consts->denorm.x = mc->buffered_surface->texture->width[0]; + vs_consts->denorm.y = mc->buffered_surface->texture->height[0]; + + pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer); + pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &mc->vs_const_buf); + pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &mc->fs_const_buf); + + if (num_macroblocks[vlMacroBlockExTypeIntra] > 0) + { + pipe->set_vertex_buffers(pipe, 1, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS]); + pipe->set_vertex_elements(pipe, 4, mc->vertex_elems); + pipe->set_sampler_textures(pipe, 3, mc->textures[mc->cur_buf % NUM_BUF_SETS]); + pipe->bind_sampler_states(pipe, 3, (void**)mc->samplers); + pipe->bind_vs_state(pipe, mc->i_vs); + pipe->bind_fs_state(pipe, mc->i_fs); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeIntra] * 24); + vb_start += num_macroblocks[vlMacroBlockExTypeIntra] * 24; + } + + if (num_macroblocks[vlMacroBlockExTypeFwdPredictedFrame] > 0) + { + pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS]); + pipe->set_vertex_elements(pipe, 6, mc->vertex_elems); + mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->past_surface->texture; + pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUF_SETS]); + pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers); + pipe->bind_vs_state(pipe, mc->p_vs[0]); + pipe->bind_fs_state(pipe, mc->p_fs[0]); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeFwdPredictedFrame] * 24); + vb_start += num_macroblocks[vlMacroBlockExTypeFwdPredictedFrame] * 24; + } + + if (num_macroblocks[vlMacroBlockExTypeFwdPredictedField] > 0) + { + pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS]); + pipe->set_vertex_elements(pipe, 6, mc->vertex_elems); + mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->past_surface->texture; + pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUF_SETS]); + pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers); + pipe->bind_vs_state(pipe, mc->p_vs[1]); + pipe->bind_fs_state(pipe, mc->p_fs[1]); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeFwdPredictedField] * 24); + vb_start += num_macroblocks[vlMacroBlockExTypeFwdPredictedField] * 24; + } + + if (num_macroblocks[vlMacroBlockExTypeBkwdPredictedFrame] > 0) + { + pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS]); + pipe->set_vertex_elements(pipe, 6, mc->vertex_elems); + mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->future_surface->texture; + pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUF_SETS]); + pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers); + pipe->bind_vs_state(pipe, mc->p_vs[0]); + pipe->bind_fs_state(pipe, mc->p_fs[0]); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeBkwdPredictedFrame] * 24); + vb_start += num_macroblocks[vlMacroBlockExTypeBkwdPredictedFrame] * 24; + } + + if (num_macroblocks[vlMacroBlockExTypeBkwdPredictedField] > 0) + { + pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS]); + pipe->set_vertex_elements(pipe, 6, mc->vertex_elems); + mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->future_surface->texture; + pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUF_SETS]); + pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers); + pipe->bind_vs_state(pipe, mc->p_vs[1]); + pipe->bind_fs_state(pipe, mc->p_fs[1]); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeBkwdPredictedField] * 24); + vb_start += num_macroblocks[vlMacroBlockExTypeBkwdPredictedField] * 24; + } + + if (num_macroblocks[vlMacroBlockExTypeBiPredictedFrame] > 0) + { + pipe->set_vertex_buffers(pipe, 3, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS]); + pipe->set_vertex_elements(pipe, 8, mc->vertex_elems); + mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->past_surface->texture; + mc->textures[mc->cur_buf % NUM_BUF_SETS][4] = mc->future_surface->texture; + pipe->set_sampler_textures(pipe, 5, mc->textures[mc->cur_buf % NUM_BUF_SETS]); + pipe->bind_sampler_states(pipe, 5, (void**)mc->samplers); + pipe->bind_vs_state(pipe, mc->b_vs[0]); + pipe->bind_fs_state(pipe, mc->b_fs[0]); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeBiPredictedFrame] * 24); + vb_start += num_macroblocks[vlMacroBlockExTypeBiPredictedFrame] * 24; + } + + if (num_macroblocks[vlMacroBlockExTypeBiPredictedField] > 0) + { + pipe->set_vertex_buffers(pipe, 3, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS]); + pipe->set_vertex_elements(pipe, 8, mc->vertex_elems); + mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->past_surface->texture; + mc->textures[mc->cur_buf % NUM_BUF_SETS][4] = mc->future_surface->texture; + pipe->set_sampler_textures(pipe, 5, mc->textures[mc->cur_buf % NUM_BUF_SETS]); + pipe->bind_sampler_states(pipe, 5, (void**)mc->samplers); + pipe->bind_vs_state(pipe, mc->b_vs[1]); + pipe->bind_fs_state(pipe, mc->b_fs[1]); + + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeBiPredictedField] * 24); + vb_start += num_macroblocks[vlMacroBlockExTypeBiPredictedField] * 24; + } + + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, &mc->buffered_surface->render_fence); + + for (i = 0; i < 3; ++i) + mc->zero_block[i].x = -1.0f; + + mc->buffered_surface = NULL; + mc->num_macroblocks = 0; + mc->cur_buf++; + + return 0; +} + +static int vlRenderMacroBlocksMpeg2R16SnormBuffered +( + struct vlRender *render, + struct vlMpeg2MacroBlockBatch *batch, + struct vlSurface *surface +) +{ + struct vlR16SnormBufferedMC *mc; + unsigned int i; + + assert(render); + + mc = (struct vlR16SnormBufferedMC*)render; + + if (mc->buffered_surface) + { + if (mc->buffered_surface != surface) + { + vlFlush(&mc->base); + mc->buffered_surface = surface; + mc->past_surface = batch->past_surface; + mc->future_surface = batch->future_surface; + mc->surface_tex_inv_size.x = 1.0f / surface->texture->width[0]; + mc->surface_tex_inv_size.y = 1.0f / surface->texture->height[0]; + + for (i = 0; i < 3; ++i) + { + mc->tex_surface[i] = mc->pipe->screen->get_tex_surface + ( + mc->pipe->screen, + mc->textures[mc->cur_buf % NUM_BUF_SETS][i], + 0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE + ); + + mc->texels[i] = pipe_surface_map(mc->tex_surface[i], PIPE_BUFFER_USAGE_CPU_WRITE); + } + } + } + else + { + mc->buffered_surface = surface; + mc->past_surface = batch->past_surface; + mc->future_surface = batch->future_surface; + mc->surface_tex_inv_size.x = 1.0f / surface->texture->width[0]; + mc->surface_tex_inv_size.y = 1.0f / surface->texture->height[0]; + + for (i = 0; i < 3; ++i) + { + mc->tex_surface[i] = mc->pipe->screen->get_tex_surface + ( + mc->pipe->screen, + mc->textures[mc->cur_buf % NUM_BUF_SETS][i], + 0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE + ); + + mc->texels[i] = pipe_surface_map(mc->tex_surface[i], PIPE_BUFFER_USAGE_CPU_WRITE); + } + } + + for (i = 0; i < batch->num_macroblocks; ++i) + vlGrabMacroBlock(mc, &batch->macroblocks[i]); + + return 0; +} + +static int vlEnd +( + struct vlRender *render +) +{ + assert(render); + + return 0; +} + +static int vlDestroy +( + struct vlRender *render +) +{ + struct vlR16SnormBufferedMC *mc; + struct pipe_context *pipe; + unsigned int h, i; + + assert(render); + + mc = (struct vlR16SnormBufferedMC*)render; + pipe = mc->pipe; + + for (i = 0; i < 5; ++i) + pipe->delete_sampler_state(pipe, mc->samplers[i]); + + for (h = 0; h < NUM_BUF_SETS; ++h) + for (i = 0; i < 3; ++i) + pipe->winsys->buffer_destroy(pipe->winsys, mc->vertex_bufs[h][i].buffer); + + /* Textures 3 & 4 are not created directly, no need to release them here */ + for (i = 0; i < NUM_BUF_SETS; ++i) + { + pipe_texture_release(&mc->textures[i][0]); + pipe_texture_release(&mc->textures[i][1]); + pipe_texture_release(&mc->textures[i][2]); + } + + pipe->delete_vs_state(pipe, mc->i_vs); + pipe->delete_fs_state(pipe, mc->i_fs); + + for (i = 0; i < 2; ++i) + { + pipe->delete_vs_state(pipe, mc->p_vs[i]); + pipe->delete_fs_state(pipe, mc->p_fs[i]); + pipe->delete_vs_state(pipe, mc->b_vs[i]); + pipe->delete_fs_state(pipe, mc->b_fs[i]); + } + + pipe->winsys->buffer_destroy(pipe->winsys, mc->vs_const_buf.buffer); + pipe->winsys->buffer_destroy(pipe->winsys, mc->fs_const_buf.buffer); + + free(mc->macroblocks); + free(mc); + + return 0; +} + +/* + * Muliplier renormalizes block samples from 16 bits to 12 bits. + * Divider is used when calculating Y % 2 for choosing top or bottom + * field for P or B macroblocks. + * TODO: Use immediates. + */ +static const struct vlFragmentShaderConsts fs_consts = +{ + {32767.0f / 255.0f, 32767.0f / 255.0f, 32767.0f / 255.0f, 0.0f}, + {0.5f, 2.0f, 0.0f, 0.0f} +}; + +#include "vl_r16snorm_mc_buf_shaders.inc" + +static int vlCreateDataBufs +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int mbw = align(mc->picture_width, VL_MACROBLOCK_WIDTH) / VL_MACROBLOCK_WIDTH; + const unsigned int mbh = align(mc->picture_height, VL_MACROBLOCK_HEIGHT) / VL_MACROBLOCK_HEIGHT; + const unsigned int num_mb_per_frame = mbw * mbh; + + struct pipe_context *pipe; + unsigned int h, i; + + assert(mc); + + pipe = mc->pipe; + + /* Create our vertex buffers */ + for (h = 0; h < NUM_BUF_SETS; ++h) + { + mc->vertex_bufs[h][0].pitch = sizeof(struct vlVertex2f) * 4; + mc->vertex_bufs[h][0].max_index = 24 * num_mb_per_frame - 1; + mc->vertex_bufs[h][0].buffer_offset = 0; + mc->vertex_bufs[h][0].buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(struct vlVertex2f) * 4 * 24 * num_mb_per_frame + ); + + for (i = 1; i < 3; ++i) + { + mc->vertex_bufs[h][i].pitch = sizeof(struct vlVertex2f) * 2; + mc->vertex_bufs[h][i].max_index = 24 * num_mb_per_frame - 1; + mc->vertex_bufs[h][i].buffer_offset = 0; + mc->vertex_bufs[h][i].buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(struct vlVertex2f) * 2 * 24 * num_mb_per_frame + ); + } + } + + /* Position element */ + mc->vertex_elems[0].src_offset = 0; + mc->vertex_elems[0].vertex_buffer_index = 0; + mc->vertex_elems[0].nr_components = 2; + mc->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* Luma, texcoord element */ + mc->vertex_elems[1].src_offset = sizeof(struct vlVertex2f); + mc->vertex_elems[1].vertex_buffer_index = 0; + mc->vertex_elems[1].nr_components = 2; + mc->vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* Chroma Cr texcoord element */ + mc->vertex_elems[2].src_offset = sizeof(struct vlVertex2f) * 2; + mc->vertex_elems[2].vertex_buffer_index = 0; + mc->vertex_elems[2].nr_components = 2; + mc->vertex_elems[2].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* Chroma Cb texcoord element */ + mc->vertex_elems[3].src_offset = sizeof(struct vlVertex2f) * 3; + mc->vertex_elems[3].vertex_buffer_index = 0; + mc->vertex_elems[3].nr_components = 2; + mc->vertex_elems[3].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* First ref surface top field texcoord element */ + mc->vertex_elems[4].src_offset = 0; + mc->vertex_elems[4].vertex_buffer_index = 1; + mc->vertex_elems[4].nr_components = 2; + mc->vertex_elems[4].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* First ref surface bottom field texcoord element */ + mc->vertex_elems[5].src_offset = sizeof(struct vlVertex2f); + mc->vertex_elems[5].vertex_buffer_index = 1; + mc->vertex_elems[5].nr_components = 2; + mc->vertex_elems[5].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* Second ref surface top field texcoord element */ + mc->vertex_elems[6].src_offset = 0; + mc->vertex_elems[6].vertex_buffer_index = 2; + mc->vertex_elems[6].nr_components = 2; + mc->vertex_elems[6].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* Second ref surface bottom field texcoord element */ + mc->vertex_elems[7].src_offset = sizeof(struct vlVertex2f); + mc->vertex_elems[7].vertex_buffer_index = 2; + mc->vertex_elems[7].nr_components = 2; + mc->vertex_elems[7].src_format = PIPE_FORMAT_R32G32_FLOAT; + + /* Create our constant buffer */ + mc->vs_const_buf.size = sizeof(struct vlVertexShaderConsts); + mc->vs_const_buf.buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_CONSTANT, + mc->vs_const_buf.size + ); + + mc->fs_const_buf.size = sizeof(struct vlFragmentShaderConsts); + mc->fs_const_buf.buffer = pipe->winsys->buffer_create + ( + pipe->winsys, + 1, + PIPE_BUFFER_USAGE_CONSTANT, + mc->fs_const_buf.size + ); + + memcpy + ( + pipe->winsys->buffer_map(pipe->winsys, mc->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + &fs_consts, + sizeof(struct vlFragmentShaderConsts) + ); + + pipe->winsys->buffer_unmap(pipe->winsys, mc->fs_const_buf.buffer); + + mc->macroblocks = malloc(sizeof(struct vlMpeg2MacroBlock) * num_mb_per_frame); + + return 0; +} + +static int vlInit +( + struct vlR16SnormBufferedMC *mc +) +{ + struct pipe_context *pipe; + struct pipe_sampler_state sampler; + struct pipe_texture template; + unsigned int filters[5]; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + + /* For MC we render to textures, which are rounded up to nearest POT */ + mc->viewport.scale[0] = vlRoundUpPOT(mc->picture_width); + mc->viewport.scale[1] = vlRoundUpPOT(mc->picture_height); + mc->viewport.scale[2] = 1; + mc->viewport.scale[3] = 1; + mc->viewport.translate[0] = 0; + mc->viewport.translate[1] = 0; + mc->viewport.translate[2] = 0; + mc->viewport.translate[3] = 0; + + mc->render_target.width = vlRoundUpPOT(mc->picture_width); + mc->render_target.height = vlRoundUpPOT(mc->picture_height); + mc->render_target.num_cbufs = 1; + /* FB for MC stage is a vlSurface created by the user, set at render time */ + mc->render_target.zsbuf = NULL; + + filters[0] = PIPE_TEX_FILTER_NEAREST; + /* FIXME: Linear causes discoloration around block edges */ + filters[1] = /*mc->picture_format == vlFormatYCbCr444 ?*/ PIPE_TEX_FILTER_NEAREST /*: PIPE_TEX_FILTER_LINEAR*/; + filters[2] = /*mc->picture_format == vlFormatYCbCr444 ?*/ PIPE_TEX_FILTER_NEAREST /*: PIPE_TEX_FILTER_LINEAR*/; + filters[3] = PIPE_TEX_FILTER_LINEAR; + filters[4] = PIPE_TEX_FILTER_LINEAR; + + for (i = 0; i < 5; ++i) + { + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_img_filter = filters[i]; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.mag_img_filter = filters[i]; + sampler.compare_mode = PIPE_TEX_COMPARE_NONE; + sampler.compare_func = PIPE_FUNC_ALWAYS; + sampler.normalized_coords = 1; + /*sampler.prefilter = ;*/ + /*sampler.shadow_ambient = ;*/ + /*sampler.lod_bias = ;*/ + sampler.min_lod = 0; + /*sampler.max_lod = ;*/ + /*sampler.border_color[i] = ;*/ + /*sampler.max_anisotropy = ;*/ + mc->samplers[i] = pipe->create_sampler_state(pipe, &sampler); + } + + memset(&template, 0, sizeof(struct pipe_texture)); + template.target = PIPE_TEXTURE_2D; + template.format = PIPE_FORMAT_R16_SNORM; + template.last_level = 0; + template.width[0] = vlRoundUpPOT(mc->picture_width); + template.height[0] = vlRoundUpPOT(mc->picture_height); + template.depth[0] = 1; + template.compressed = 0; + pf_get_block(template.format, &template.block); + template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; + + for (i = 0; i < NUM_BUF_SETS; ++i) + mc->textures[i][0] = pipe->screen->texture_create(pipe->screen, &template); + + if (mc->picture_format == vlFormatYCbCr420) + { + template.width[0] = vlRoundUpPOT(mc->picture_width / 2); + template.height[0] = vlRoundUpPOT(mc->picture_height / 2); + } + else if (mc->picture_format == vlFormatYCbCr422) + template.height[0] = vlRoundUpPOT(mc->picture_height / 2); + + for (i = 0; i < NUM_BUF_SETS; ++i) + { + mc->textures[i][1] = pipe->screen->texture_create(pipe->screen, &template); + mc->textures[i][2] = pipe->screen->texture_create(pipe->screen, &template); + } + + /* textures[3] & textures[4] are assigned from vlSurfaces for P and B macroblocks at render time */ + + vlCreateVertexShaderIMB(mc); + vlCreateFragmentShaderIMB(mc); + vlCreateVertexShaderFramePMB(mc); + vlCreateVertexShaderFieldPMB(mc); + vlCreateFragmentShaderFramePMB(mc); + vlCreateFragmentShaderFieldPMB(mc); + vlCreateVertexShaderFrameBMB(mc); + vlCreateVertexShaderFieldBMB(mc); + vlCreateFragmentShaderFrameBMB(mc); + vlCreateFragmentShaderFieldBMB(mc); + vlCreateDataBufs(mc); + + return 0; +} + +int vlCreateR16SNormBufferedMC +( + struct pipe_context *pipe, + unsigned int picture_width, + unsigned int picture_height, + enum vlFormat picture_format, + struct vlRender **render +) +{ + struct vlR16SnormBufferedMC *mc; + unsigned int i; + + assert(pipe); + assert(render); + + mc = calloc(1, sizeof(struct vlR16SnormBufferedMC)); + + mc->base.vlBegin = &vlBegin; + mc->base.vlRenderMacroBlocksMpeg2 = &vlRenderMacroBlocksMpeg2R16SnormBuffered; + mc->base.vlEnd = &vlEnd; + mc->base.vlFlush = &vlFlush; + mc->base.vlDestroy = &vlDestroy; + mc->pipe = pipe; + mc->picture_width = picture_width; + mc->picture_height = picture_height; + + mc->cur_buf = 0; + mc->buffered_surface = NULL; + mc->past_surface = NULL; + mc->future_surface = NULL; + for (i = 0; i < 3; ++i) + mc->zero_block[i].x = -1.0f; + mc->num_macroblocks = 0; + + vlInit(mc); + + *render = &mc->base; + + return 0; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h new file mode 100644 index 0000000000..27177d64ca --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h @@ -0,0 +1,18 @@ +#ifndef vl_r16snorm_mc_buf_h +#define vl_r16snorm_mc_buf_h + +#include "vl_types.h" + +struct pipe_context; +struct vlRender; + +int vlCreateR16SNormBufferedMC +( + struct pipe_context *pipe, + unsigned int picture_width, + unsigned int picture_height, + enum vlFormat picture_format, + struct vlRender **render +); + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf_shaders.inc b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf_shaders.inc new file mode 100644 index 0000000000..ef4a4b2add --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf_shaders.inc @@ -0,0 +1,1185 @@ +static int vlCreateVertexShaderIMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 50; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma Cb texcoords + * decl i3 ; Chroma Cr texcoords + */ + for (i = 0; i < 4; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma Cb texcoords + * decl o3 ; Chroma Cr texcoords + */ + for (i = 0; i < 4; i++) + { + decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * mov o0, i0 ; Move input vertex pos to output + * mov o1, i1 ; Move input luma texcoords to output + * mov o2, i2 ; Move input chroma Cb texcoords to output + * mov o3, i3 ; Move input chroma Cr texcoords to output + */ + for (i = 0; i < 4; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->i_vs = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderIMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Luma texcoords + * decl i1 ; Chroma Cb texcoords + * decl i2 ; Chroma Cr texcoords + */ + for (i = 0; i < 3; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0, t1 */ + decl = vl_decl_temps(0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + */ + for (i = 0; i < 3; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header,max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i2, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* mul o0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->i_fs = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateVertexShaderFramePMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma Cb texcoords + * decl i3 ; Chroma Cr texcoords + * decl i4 ; Ref surface top field texcoords + * decl i5 ; Ref surface bottom field texcoords (unused, packed in the same stream) + */ + for (i = 0; i < 6; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma Cb texcoords + * decl o3 ; Chroma Cr texcoords + * decl o4 ; Ref macroblock texcoords + */ + for (i = 0; i < 5; i++) + { + decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * mov o0, i0 ; Move input vertex pos to output + * mov o1, i1 ; Move input luma texcoords to output + * mov o2, i2 ; Move input chroma Cb texcoords to output + * mov o3, i3 ; Move input chroma Cr texcoords to output + */ + for (i = 0; i < 4; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* add o4, i0, i4 ; Translate vertex pos by motion vec to form ref macroblock texcoords */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 4, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, 4); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->p_vs[0] = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateVertexShaderFieldPMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma Cb texcoords + * decl i3 ; Chroma Cr texcoords + * decl i4 ; Ref macroblock top field texcoords + * decl i5 ; Ref macroblock bottom field texcoords + */ + for (i = 0; i < 6; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl c0 ; Render target dimensions */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma Cb texcoords + * decl o3 ; Chroma Cr texcoords + * decl o4 ; Ref macroblock top field texcoords + * decl o5 ; Ref macroblock bottom field texcoords + * decl o6 ; Denormalized vertex pos + */ + for (i = 0; i < 7; i++) + { + decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * mov o0, i0 ; Move input vertex pos to output + * mov o1, i1 ; Move input luma texcoords to output + * mov o2, i2 ; Move input chroma Cb texcoords to output + * mov o3, i3 ; Move input chroma Cr texcoords to output + */ + for (i = 0; i < 4; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* + * add o4, i0, i4 ; Translate vertex pos by motion vec to form top field macroblock texcoords + * add o5, i0, i5 ; Translate vertex pos by motion vec to form bottom field macroblock texcoords + */ + for (i = 0; i < 2; ++i) + { + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 4, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, i + 4); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* mul o6, i0, c0 ; Denorm vertex pos */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 6, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->p_vs[1] = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderFramePMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Luma texcoords + * decl i1 ; Chroma Cb texcoords + * decl i2 ; Chroma Cr texcoords + * decl i3 ; Ref macroblock texcoords + */ + for (i = 0; i < 4; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0, t1 */ + decl = vl_decl_temps(0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + * decl s3 ; Sampler for ref surface texture + */ + for (i = 0; i < 4; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i2, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* mul t0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* tex2d t1, i3, s3 ; Read texel from ref macroblock */ + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 3, TGSI_FILE_SAMPLER, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, t1 ; Add ref and differential to form final output */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->p_fs[0] = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderFieldPMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 200; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Luma texcoords + * decl i1 ; Chroma Cb texcoords + * decl i2 ; Chroma Cr texcoords + * decl i3 ; Ref macroblock top field texcoords + * decl i4 ; Ref macroblock bottom field texcoords + * decl i5 ; Denormalized vertex pos + */ + for (i = 0; i < 6; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm + * decl c1 ; Constants 1/2 & 2 in .x, .y channels for Y-mod-2 top/bottom field selection + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0-t4 */ + decl = vl_decl_temps(0, 4); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + * decl s3 ; Sampler for ref surface texture + */ + for (i = 0; i < 4; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i2, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* mul t0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * tex2d t1, i3, s3 ; Read texel from ref macroblock top field + * tex2d t2, i4, s3 ; Read texel from ref macroblock bottom field + */ + for (i = 0; i < 2; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 3, TGSI_FILE_SAMPLER, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* XXX: Pos values off by 0.5? */ + /* sub t4, i5.y, c1.x ; Sub 0.5 from denormalized pos */ + inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 5, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mul t3, t4, c1.x ; Multiply pos Y-coord by 1/2 */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* floor t3, t3 ; Get rid of fractional part */ + inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mul t3, t3, c1.y ; Multiply by 2 */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* sub t3, t4, t3 ; Subtract from original Y to get Y % 2 */ + inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* TODO: Move to conditional tex fetch on t3 instead of lerp */ + /* lerp t1, t3, t1, t2 ; Choose between top and bottom fields based on Y % 2 */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, t1 ; Add ref and differential to form final output */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->p_fs[1] = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateVertexShaderFrameBMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma Cb texcoords + * decl i3 ; Chroma Cr texcoords + * decl i4 ; First ref macroblock top field texcoords + * decl i5 ; First ref macroblock bottom field texcoords (unused, packed in the same stream) + * decl i6 ; Second ref macroblock top field texcoords + * decl i7 ; Second ref macroblock bottom field texcoords (unused, packed in the same stream) + */ + for (i = 0; i < 8; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma Cb texcoords + * decl o3 ; Chroma Cr texcoords + * decl o4 ; First ref macroblock texcoords + * decl o5 ; Second ref macroblock texcoords + */ + for (i = 0; i < 6; i++) + { + decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * mov o0, i0 ; Move input vertex pos to output + * mov o1, i1 ; Move input luma texcoords to output + * mov o2, i2 ; Move input chroma Cb texcoords to output + * mov o3, i3 ; Move input chroma Cr texcoords to output + */ + for (i = 0; i < 4; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* + * add o4, i0, i4 ; Translate vertex pos by motion vec to form first ref macroblock texcoords + * add o5, i0, i6 ; Translate vertex pos by motion vec to form second ref macroblock texcoords + */ + for (i = 0; i < 2; ++i) + { + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 4, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, (i + 2) * 2); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->b_vs[0] = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateVertexShaderFieldBMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state vs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + /* + * decl i0 ; Vertex pos + * decl i1 ; Luma texcoords + * decl i2 ; Chroma Cb texcoords + * decl i3 ; Chroma Cr texcoords + * decl i4 ; First ref macroblock top field texcoords + * decl i5 ; First ref macroblock bottom field texcoords + * decl i6 ; Second ref macroblock top field texcoords + * decl i7 ; Second ref macroblock bottom field texcoords + */ + for (i = 0; i < 8; i++) + { + decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl c0 ; Render target dimensions */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl o0 ; Vertex pos + * decl o1 ; Luma texcoords + * decl o2 ; Chroma Cb texcoords + * decl o3 ; Chroma Cr texcoords + * decl o4 ; First ref macroblock top field texcoords + * decl o5 ; First ref macroblock Bottom field texcoords + * decl o6 ; Second ref macroblock top field texcoords + * decl o7 ; Second ref macroblock Bottom field texcoords + * decl o8 ; Denormalized vertex pos + */ + for (i = 0; i < 9; i++) + { + decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* decl t0, t1 */ + decl = vl_decl_temps(0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * mov o0, i0 ; Move input vertex pos to output + * mov o1, i1 ; Move input luma texcoords to output + * mov o2, i2 ; Move input chroma Cb texcoords to output + * mov o3, i3 ; Move input chroma Cr texcoords to output + */ + for (i = 0; i < 4; ++i) + { + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* + * add o4, i0, i4 ; Translate vertex pos by motion vec to form first top field macroblock texcoords + * add o5, i0, i5 ; Translate vertex pos by motion vec to form first bottom field macroblock texcoords + * add o6, i0, i6 ; Translate vertex pos by motion vec to form second top field macroblock texcoords + * add o7, i0, i7 ; Translate vertex pos by motion vec to form second bottom field macroblock texcoords + */ + for (i = 0; i < 4; ++i) + { + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 4, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, i + 4); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* mul o8, i0, c0 ; Denorm vertex pos */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 8, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + vs.tokens = tokens; + mc->b_vs[1] = pipe->create_vs_state(pipe, &vs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderFrameBMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 100; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Luma texcoords + * decl i1 ; Chroma Cb texcoords + * decl i2 ; Chroma Cr texcoords + * decl i3 ; First ref macroblock texcoords + * decl i4 ; Second ref macroblock texcoords + */ + for (i = 0; i < 5; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm + * decl c1 ; Constant 1/2 in .x channel to use as weight to blend past and future texels + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0-t2 */ + decl = vl_decl_temps(0, 2); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + * decl s3 ; Sampler for first ref surface texture + * decl s4 ; Sampler for second ref surface texture + */ + for (i = 0; i < 5; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i2, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* mul t0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * tex2d t1, i3, s3 ; Read texel from first ref macroblock + * tex2d t2, i4, s4 ; Read texel from second ref macroblock + */ + for (i = 0; i < 2; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 3, TGSI_FILE_SAMPLER, i + 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* lerp t1, c1.x, t1, t2 ; Blend past and future texels */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, t1 ; Add past/future ref and differential to form final output */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->b_fs[0] = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} + +static int vlCreateFragmentShaderFieldBMB +( + struct vlR16SnormBufferedMC *mc +) +{ + const unsigned int max_tokens = 200; + + struct pipe_context *pipe; + struct pipe_shader_state fs; + struct tgsi_token *tokens; + struct tgsi_header *header; + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + + unsigned int ti; + unsigned int i; + + assert(mc); + + pipe = mc->pipe; + tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token)); + + /* Version */ + *(struct tgsi_version*)&tokens[0] = tgsi_build_version(); + /* Header */ + header = (struct tgsi_header*)&tokens[1]; + *header = tgsi_build_header(); + /* Processor */ + *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header); + + ti = 3; + + /* + * decl i0 ; Luma texcoords + * decl i1 ; Chroma Cb texcoords + * decl i2 ; Chroma Cr texcoords + * decl i3 ; First ref macroblock top field texcoords + * decl i4 ; First ref macroblock bottom field texcoords + * decl i5 ; Second ref macroblock top field texcoords + * decl i6 ; Second ref macroblock bottom field texcoords + * decl i7 ; Denormalized vertex pos + */ + for (i = 0; i < 8; ++i) + { + decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * decl c0 ; Scaling factor, rescales 16-bit snorm to 9-bit snorm + * decl c1 ; Constants 1/2 & 2 in .x, .y channels to use as weight to blend past and future texels + * ; and for Y-mod-2 top/bottom field selection + */ + decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl o0 ; Fragment color */ + decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* decl t0-t5 */ + decl = vl_decl_temps(0, 5); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + + /* + * decl s0 ; Sampler for luma texture + * decl s1 ; Sampler for chroma Cb texture + * decl s2 ; Sampler for chroma Cr texture + * decl s3 ; Sampler for first ref surface texture + * decl s4 ; Sampler for second ref surface texture + */ + for (i = 0; i < 5; ++i) + { + decl = vl_decl_samplers(i, i); + ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti); + } + + /* + * tex2d t1, i0, s0 ; Read texel from luma texture + * mov t0.x, t1.x ; Move luma sample into .x component + * tex2d t1, i1, s1 ; Read texel from chroma Cb texture + * mov t0.y, t1.x ; Move Cb sample into .y component + * tex2d t1, i2, s2 ; Read texel from chroma Cr texture + * mov t0.z, t1.x ; Move Cr sample into .z component + */ + for (i = 0; i < 3; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* mul t0, t0, c0 ; Rescale texel to correct range */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* XXX: Pos values off by 0.5? */ + /* sub t4, i7.y, c1.x ; Sub 0.5 from denormalized pos */ + inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 7, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mul t3, t4, c1.x ; Multiply pos Y-coord by 1/2 */ + inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* floor t3, t3 ; Get rid of fractional part */ + inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* mul t3, t3, c1.y ; Multiply by 2 */ + inst = vl_inst3( TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1); + inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* sub t3, t4, t3 ; Subtract from original Y to get Y % 2 */ + inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * tex2d t1, i3, s3 ; Read texel from past ref macroblock top field + * tex2d t2, i4, s3 ; Read texel from past ref macroblock bottom field + */ + for (i = 0; i < 2; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 3, TGSI_FILE_SAMPLER, 3); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* TODO: Move to conditional tex fetch on t3 instead of lerp */ + /* lerp t1, t3, t1, t2 ; Choose between top and bottom fields based on Y % 2 */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* + * tex2d t4, i5, s4 ; Read texel from future ref macroblock top field + * tex2d t5, i6, s4 ; Read texel from future ref macroblock bottom field + */ + for (i = 0; i < 2; ++i) + { + inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 4, TGSI_FILE_INPUT, i + 5, TGSI_FILE_SAMPLER, 4); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + } + + /* TODO: Move to conditional tex fetch on t3 instead of lerp */ + /* lerp t2, t3, t4, t5 ; Choose between top and bottom fields based on Y % 2 */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 2, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 5); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* lerp t1, c1.x, t1, t2 ; Blend past and future texels */ + inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2); + inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X; + inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* add o0, t0, t1 ; Add past/future ref and differential to form final output */ + inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + /* end */ + inst = vl_end(); + ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti); + + fs.tokens = tokens; + mc->b_fs[1] = pipe->create_fs_state(pipe, &fs); + free(tokens); + + return 0; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_render.h b/src/gallium/state_trackers/g3dvl/vl_render.h new file mode 100644 index 0000000000..166030b498 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_render.h @@ -0,0 +1,38 @@ +#ifndef vl_render_h +#define vl_render_h + +#include "vl_types.h" + +struct pipe_surface; + +struct vlRender +{ + int (*vlBegin) + ( + struct vlRender *render + ); + + int (*vlRenderMacroBlocksMpeg2) + ( + struct vlRender *render, + struct vlMpeg2MacroBlockBatch *batch, + struct vlSurface *surface + ); + + int (*vlEnd) + ( + struct vlRender *render + ); + + int (*vlFlush) + ( + struct vlRender *render + ); + + int (*vlDestroy) + ( + struct vlRender *render + ); +}; + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_screen.c b/src/gallium/state_trackers/g3dvl/vl_screen.c new file mode 100644 index 0000000000..484f63b0d4 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_screen.c @@ -0,0 +1,115 @@ +#define VL_INTERNAL +#include "vl_screen.h" +#include <assert.h> +#include <stdlib.h> + +int vlCreateScreen +( + struct vlDisplay *display, + int screen, + struct pipe_screen *pscreen, + struct vlScreen **vl_screen +) +{ + struct vlScreen *scrn; + + assert(display); + assert(pscreen); + assert(vl_screen); + + scrn = calloc(1, sizeof(struct vlScreen)); + + if (!scrn) + return 1; + + scrn->display = display; + scrn->ordinal = screen; + scrn->pscreen = pscreen; + *vl_screen = scrn; + + return 0; +} + +int vlDestroyScreen +( + struct vlScreen *screen +) +{ + assert(screen); + + free(screen); + + return 0; +} + +struct vlDisplay* vlGetDisplay +( + struct vlScreen *screen +) +{ + assert(screen); + + return screen->display; +} + +struct pipe_screen* vlGetPipeScreen +( + struct vlScreen *screen +) +{ + assert(screen); + + return screen->pscreen; +} + +unsigned int vlGetMaxProfiles +( + struct vlScreen *screen +) +{ + assert(screen); + + return vlProfileCount; +} + +int vlQueryProfiles +( + struct vlScreen *screen, + enum vlProfile *profiles +) +{ + assert(screen); + assert(profiles); + + profiles[0] = vlProfileMpeg2Simple; + profiles[1] = vlProfileMpeg2Main; + + return 0; +} + +unsigned int vlGetMaxEntryPoints +( + struct vlScreen *screen +) +{ + assert(screen); + + return vlEntryPointCount; +} + +int vlQueryEntryPoints +( + struct vlScreen *screen, + enum vlProfile profile, + enum vlEntryPoint *entry_points +) +{ + assert(screen); + assert(entry_points); + + entry_points[0] = vlEntryPointIDCT; + entry_points[1] = vlEntryPointMC; + entry_points[2] = vlEntryPointCSC; + + return 0; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_screen.h b/src/gallium/state_trackers/g3dvl/vl_screen.h new file mode 100644 index 0000000000..98f3d429b6 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_screen.h @@ -0,0 +1,63 @@ +#ifndef vl_screen_h +#define vl_screen_h + +#include "vl_types.h" + +struct pipe_screen; + +#ifdef VL_INTERNAL +struct vlScreen +{ + struct vlDisplay *display; + unsigned int ordinal; + struct pipe_screen *pscreen; +}; +#endif + +int vlCreateScreen +( + struct vlDisplay *display, + int screen, + struct pipe_screen *pscreen, + struct vlScreen **vl_screen +); + +int vlDestroyScreen +( + struct vlScreen *screen +); + +struct vlDisplay* vlGetDisplay +( + struct vlScreen *screen +); + +struct pipe_screen* vlGetPipeScreen +( + struct vlScreen *screen +); + +unsigned int vlGetMaxProfiles +( + struct vlScreen *screen +); + +int vlQueryProfiles +( + struct vlScreen *screen, + enum vlProfile *profiles +); + +unsigned int vlGetMaxEntryPoints +( + struct vlScreen *screen +); + +int vlQueryEntryPoints +( + struct vlScreen *screen, + enum vlProfile profile, + enum vlEntryPoint *entry_points +); + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.c b/src/gallium/state_trackers/g3dvl/vl_shader_build.c new file mode 100644 index 0000000000..51f1721a33 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.c @@ -0,0 +1,215 @@ +#include "vl_shader_build.h" +#include <assert.h> +#include <tgsi/tgsi_parse.h> +#include <tgsi/tgsi_build.h> + +struct tgsi_full_declaration vl_decl_input(unsigned int name, unsigned int index, unsigned int first, unsigned int last) +{ + struct tgsi_full_declaration decl = tgsi_default_full_declaration(); + + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = name; + decl.Semantic.SemanticIndex = index; + decl.DeclarationRange.First = first; + decl.DeclarationRange.Last = last; + + return decl; +} + +struct tgsi_full_declaration vl_decl_interpolated_input +( + unsigned int name, + unsigned int index, + unsigned int first, + unsigned int last, + int interpolation +) +{ + struct tgsi_full_declaration decl = tgsi_default_full_declaration(); + + assert + ( + interpolation == TGSI_INTERPOLATE_CONSTANT || + interpolation == TGSI_INTERPOLATE_LINEAR || + interpolation == TGSI_INTERPOLATE_PERSPECTIVE + ); + + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = name; + decl.Semantic.SemanticIndex = index; + decl.Declaration.Interpolate = interpolation;; + decl.DeclarationRange.First = first; + decl.DeclarationRange.Last = last; + + return decl; +} + +struct tgsi_full_declaration vl_decl_constants(unsigned int name, unsigned int index, unsigned int first, unsigned int last) +{ + struct tgsi_full_declaration decl = tgsi_default_full_declaration(); + + decl.Declaration.File = TGSI_FILE_CONSTANT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = name; + decl.Semantic.SemanticIndex = index; + decl.DeclarationRange.First = first; + decl.DeclarationRange.Last = last; + + return decl; +} + +struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int index, unsigned int first, unsigned int last) +{ + struct tgsi_full_declaration decl = tgsi_default_full_declaration(); + + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = name; + decl.Semantic.SemanticIndex = index; + decl.DeclarationRange.First = first; + decl.DeclarationRange.Last = last; + + return decl; +} + +struct tgsi_full_declaration vl_decl_temps(unsigned int first, unsigned int last) +{ + struct tgsi_full_declaration decl = tgsi_default_full_declaration(); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.DeclarationRange.First = first; + decl.DeclarationRange.Last = last; + + return decl; +} + +struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int last) +{ + struct tgsi_full_declaration decl = tgsi_default_full_declaration(); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_SAMPLER; + decl.DeclarationRange.First = first; + decl.DeclarationRange.Last = last; + + return decl; +} + +struct tgsi_full_instruction vl_inst2 +( + int opcode, + enum tgsi_file_type dst_file, + unsigned int dst_index, + enum tgsi_file_type src_file, + unsigned int src_index +) +{ + struct tgsi_full_instruction inst = tgsi_default_full_instruction(); + + inst.Instruction.Opcode = opcode; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = dst_file; + inst.FullDstRegisters[0].DstRegister.Index = dst_index; + inst.Instruction.NumSrcRegs = 1; + inst.FullSrcRegisters[0].SrcRegister.File = src_file; + inst.FullSrcRegisters[0].SrcRegister.Index = src_index; + + return inst; +} + +struct tgsi_full_instruction vl_inst3 +( + int opcode, + enum tgsi_file_type dst_file, + unsigned int dst_index, + enum tgsi_file_type src1_file, + unsigned int src1_index, + enum tgsi_file_type src2_file, + unsigned int src2_index +) +{ + struct tgsi_full_instruction inst = tgsi_default_full_instruction(); + + inst.Instruction.Opcode = opcode; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = dst_file; + inst.FullDstRegisters[0].DstRegister.Index = dst_index; + inst.Instruction.NumSrcRegs = 2; + inst.FullSrcRegisters[0].SrcRegister.File = src1_file; + inst.FullSrcRegisters[0].SrcRegister.Index = src1_index; + inst.FullSrcRegisters[1].SrcRegister.File = src2_file; + inst.FullSrcRegisters[1].SrcRegister.Index = src2_index; + + return inst; +} + +struct tgsi_full_instruction vl_tex +( + int tex, + enum tgsi_file_type dst_file, + unsigned int dst_index, + enum tgsi_file_type src1_file, + unsigned int src1_index, + enum tgsi_file_type src2_file, + unsigned int src2_index +) +{ + struct tgsi_full_instruction inst = tgsi_default_full_instruction(); + + inst.Instruction.Opcode = TGSI_OPCODE_TEX; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = dst_file; + inst.FullDstRegisters[0].DstRegister.Index = dst_index; + inst.Instruction.NumSrcRegs = 2; + inst.InstructionExtTexture.Texture = tex; + inst.FullSrcRegisters[0].SrcRegister.File = src1_file; + inst.FullSrcRegisters[0].SrcRegister.Index = src1_index; + inst.FullSrcRegisters[1].SrcRegister.File = src2_file; + inst.FullSrcRegisters[1].SrcRegister.Index = src2_index; + + return inst; +} + +struct tgsi_full_instruction vl_inst4 +( + int opcode, + enum tgsi_file_type dst_file, + unsigned int dst_index, + enum tgsi_file_type src1_file, + unsigned int src1_index, + enum tgsi_file_type src2_file, + unsigned int src2_index, + enum tgsi_file_type src3_file, + unsigned int src3_index +) +{ + struct tgsi_full_instruction inst = tgsi_default_full_instruction(); + + inst.Instruction.Opcode = opcode; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = dst_file; + inst.FullDstRegisters[0].DstRegister.Index = dst_index; + inst.Instruction.NumSrcRegs = 3; + inst.FullSrcRegisters[0].SrcRegister.File = src1_file; + inst.FullSrcRegisters[0].SrcRegister.Index = src1_index; + inst.FullSrcRegisters[1].SrcRegister.File = src2_file; + inst.FullSrcRegisters[1].SrcRegister.Index = src2_index; + inst.FullSrcRegisters[2].SrcRegister.File = src3_file; + inst.FullSrcRegisters[2].SrcRegister.Index = src3_index; + + return inst; +} + +struct tgsi_full_instruction vl_end(void) +{ + struct tgsi_full_instruction inst = tgsi_default_full_instruction(); + + inst.Instruction.Opcode = TGSI_OPCODE_END; + inst.Instruction.NumDstRegs = 0; + inst.Instruction.NumSrcRegs = 0; + + return inst; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.h b/src/gallium/state_trackers/g3dvl/vl_shader_build.h new file mode 100644 index 0000000000..dc615cb156 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.h @@ -0,0 +1,61 @@ +#ifndef vl_shader_build_h +#define vl_shader_build_h + +#include <pipe/p_shader_tokens.h> + +struct tgsi_full_declaration vl_decl_input(unsigned int name, unsigned int index, unsigned int first, unsigned int last); +struct tgsi_full_declaration vl_decl_interpolated_input +( + unsigned int name, + unsigned int index, + unsigned int first, + unsigned int last, + int interpolation +); +struct tgsi_full_declaration vl_decl_constants(unsigned int name, unsigned int index, unsigned int first, unsigned int last); +struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int index, unsigned int first, unsigned int last); +struct tgsi_full_declaration vl_decl_temps(unsigned int first, unsigned int last); +struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int last); +struct tgsi_full_instruction vl_inst2 +( + int opcode, + enum tgsi_file_type dst_file, + unsigned int dst_index, + enum tgsi_file_type src_file, + unsigned int src_index +); +struct tgsi_full_instruction vl_inst3 +( + int opcode, + enum tgsi_file_type dst_file, + unsigned int dst_index, + enum tgsi_file_type src1_file, + unsigned int src1_index, + enum tgsi_file_type src2_file, + unsigned int src2_index +); +struct tgsi_full_instruction vl_tex +( + int tex, + enum tgsi_file_type dst_file, + unsigned int dst_index, + enum tgsi_file_type src1_file, + unsigned int src1_index, + enum tgsi_file_type src2_file, + unsigned int src2_index +); +struct tgsi_full_instruction vl_inst4 +( + int opcode, + enum tgsi_file_type dst_file, + unsigned int dst_index, + enum tgsi_file_type src1_file, + unsigned int src1_index, + enum tgsi_file_type src2_file, + unsigned int src2_index, + enum tgsi_file_type src3_file, + unsigned int src3_index +); +struct tgsi_full_instruction vl_end(void); + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c new file mode 100644 index 0000000000..076bd40d41 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_surface.c @@ -0,0 +1,237 @@ +#define VL_INTERNAL +#include "vl_surface.h" +#include <assert.h> +#include <stdlib.h> +#include <string.h> +#include <pipe/p_screen.h> +#include <pipe/p_state.h> +#include <pipe/p_inlines.h> +#include <vl_winsys.h> +#include "vl_screen.h" +#include "vl_context.h" +#include "vl_render.h" +#include "vl_csc.h" +#include "vl_util.h" + +int vlCreateSurface +( + struct vlScreen *screen, + unsigned int width, + unsigned int height, + enum vlFormat format, + struct vlSurface **surface +) +{ + struct vlSurface *sfc; + struct pipe_texture template; + + assert(screen); + assert(surface); + + sfc = calloc(1, sizeof(struct vlSurface)); + + if (!sfc) + return 1; + + sfc->screen = screen; + sfc->width = width; + sfc->height = height; + sfc->format = format; + + memset(&template, 0, sizeof(struct pipe_texture)); + template.target = PIPE_TEXTURE_2D; + template.format = PIPE_FORMAT_A8R8G8B8_UNORM; + template.last_level = 0; + template.width[0] = vlRoundUpPOT(sfc->width); + template.height[0] = vlRoundUpPOT(sfc->height); + template.depth[0] = 1; + template.compressed = 0; + pf_get_block(template.format, &template.block); + template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_RENDER_TARGET; + + sfc->texture = vlGetPipeScreen(screen)->texture_create(vlGetPipeScreen(screen), &template); + + *surface = sfc; + + return 0; +} + +int vlDestroySurface +( + struct vlSurface *surface +) +{ + assert(surface); + + pipe_texture_release(&surface->texture); + free(surface); + + return 0; +} + +int vlRenderMacroBlocksMpeg2 +( + struct vlMpeg2MacroBlockBatch *batch, + struct vlSurface *surface +) +{ + assert(batch); + assert(surface); + assert(surface->context); + + surface->context->render->vlBegin(surface->context->render); + + surface->context->render->vlRenderMacroBlocksMpeg2 + ( + surface->context->render, + batch, + surface + ); + + surface->context->render->vlEnd(surface->context->render); + + return 0; +} + +int vlPutPicture +( + struct vlSurface *surface, + vlNativeDrawable drawable, + int srcx, + int srcy, + int srcw, + int srch, + int destx, + int desty, + int destw, + int desth, + int drawable_w, + int drawable_h, + enum vlPictureType picture_type +) +{ + struct vlCSC *csc; + struct pipe_context *pipe; + + assert(surface); + assert(surface->context); + + surface->context->render->vlFlush(surface->context->render); + + csc = surface->context->csc; + pipe = surface->context->pipe; + + csc->vlResizeFrameBuffer(csc, drawable_w, drawable_h); + + csc->vlBegin(csc); + + csc->vlPutPicture + ( + csc, + surface, + srcx, + srcy, + srcw, + srch, + destx, + desty, + destw, + desth, + picture_type + ); + + csc->vlEnd(csc); + + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, &surface->disp_fence); + + bind_pipe_drawable(pipe, drawable); + + pipe->winsys->flush_frontbuffer + ( + pipe->winsys, + csc->vlGetFrameBuffer(csc), + pipe->priv + ); + + return 0; +} + +int vlSurfaceGetStatus +( + struct vlSurface *surface, + enum vlResourceStatus *status +) +{ + assert(surface); + assert(surface->context); + assert(status); + + if (surface->render_fence && !surface->context->pipe->winsys->fence_signalled(surface->context->pipe->winsys, surface->render_fence, 0)) + { + *status = vlResourceStatusRendering; + return 0; + } + + if (surface->disp_fence && !surface->context->pipe->winsys->fence_signalled(surface->context->pipe->winsys, surface->disp_fence, 0)) + { + *status = vlResourceStatusDisplaying; + return 0; + } + + *status = vlResourceStatusFree; + + return 0; +} + +int vlSurfaceFlush +( + struct vlSurface *surface +) +{ + assert(surface); + assert(surface->context); + + surface->context->render->vlFlush(surface->context->render); + + return 0; +} + +int vlSurfaceSync +( + struct vlSurface *surface +) +{ + assert(surface); + assert(surface->context); + assert(surface->render_fence); + + surface->context->pipe->winsys->fence_finish(surface->context->pipe->winsys, surface->render_fence, 0); + + return 0; +} + +struct vlScreen* vlSurfaceGetScreen +( + struct vlSurface *surface +) +{ + assert(surface); + + return surface->screen; +} + +struct vlContext* vlBindToContext +( + struct vlSurface *surface, + struct vlContext *context +) +{ + struct vlContext *old; + + assert(surface); + + old = surface->context; + surface->context = context; + + return old; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.h b/src/gallium/state_trackers/g3dvl/vl_surface.h new file mode 100644 index 0000000000..133e1515ef --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_surface.h @@ -0,0 +1,86 @@ +#ifndef vl_surface_h +#define vl_surface_h + +#include "vl_types.h" + +#ifdef VL_INTERNAL +struct pipe_texture; + +struct vlSurface +{ + struct vlScreen *screen; + struct vlContext *context; + unsigned int width; + unsigned int height; + enum vlFormat format; + struct pipe_texture *texture; + struct pipe_fence_handle *render_fence; + struct pipe_fence_handle *disp_fence; +}; +#endif + +int vlCreateSurface +( + struct vlScreen *screen, + unsigned int width, + unsigned int height, + enum vlFormat format, + struct vlSurface **surface +); + +int vlDestroySurface +( + struct vlSurface *surface +); + +int vlRenderMacroBlocksMpeg2 +( + struct vlMpeg2MacroBlockBatch *batch, + struct vlSurface *surface +); + +int vlPutPicture +( + struct vlSurface *surface, + vlNativeDrawable drawable, + int srcx, + int srcy, + int srcw, + int srch, + int destx, + int desty, + int destw, + int desth, + int drawable_w, + int drawable_h, + enum vlPictureType picture_type +); + +int vlSurfaceGetStatus +( + struct vlSurface *surface, + enum vlResourceStatus *status +); + +int vlSurfaceFlush +( + struct vlSurface *surface +); + +int vlSurfaceSync +( + struct vlSurface *surface +); + +struct vlScreen* vlSurfaceGetScreen +( + struct vlSurface *surface +); + +struct vlContext* vlBindToContext +( + struct vlSurface *surface, + struct vlContext *context +); + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_types.h b/src/gallium/state_trackers/g3dvl/vl_types.h new file mode 100644 index 0000000000..274e1f7437 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_types.h @@ -0,0 +1,115 @@ +#ifndef vl_types_h +#define vl_types_h + +#if 1 /*#ifdef X11*/ +#include <X11/Xlib.h> + +typedef Display* vlNativeDisplay; +typedef Drawable vlNativeDrawable; +#endif + +struct vlDisplay; +struct vlScreen; +struct vlContext; +struct vlSurface; + +enum vlResourceStatus +{ + vlResourceStatusFree, + vlResourceStatusRendering, + vlResourceStatusDisplaying +}; + +enum vlProfile +{ + vlProfileMpeg2Simple, + vlProfileMpeg2Main, + + vlProfileCount +}; + +enum vlEntryPoint +{ + vlEntryPointIDCT, + vlEntryPointMC, + vlEntryPointCSC, + + vlEntryPointCount +}; + +enum vlFormat +{ + vlFormatYCbCr420, + vlFormatYCbCr422, + vlFormatYCbCr444 +}; + +enum vlPictureType +{ + vlPictureTypeTopField, + vlPictureTypeBottomField, + vlPictureTypeFrame +}; + +enum vlMotionType +{ + vlMotionTypeField, + vlMotionTypeFrame, + vlMotionTypeDualPrime, + vlMotionType16x8 +}; + +enum vlFieldOrder +{ + vlFieldOrderFirst, + vlFieldOrderSecond +}; + +enum vlDCTType +{ + vlDCTTypeFrameCoded, + vlDCTTypeFieldCoded +}; + +struct vlVertex2f +{ + float x, y; +}; + +struct vlVertex4f +{ + float x, y, z, w; +}; + +enum vlMacroBlockType +{ + vlMacroBlockTypeIntra, + vlMacroBlockTypeFwdPredicted, + vlMacroBlockTypeBkwdPredicted, + vlMacroBlockTypeBiPredicted, + + vlNumMacroBlockTypes +}; + +struct vlMpeg2MacroBlock +{ + unsigned int mbx, mby; + enum vlMacroBlockType mb_type; + enum vlMotionType mo_type; + enum vlDCTType dct_type; + int PMV[2][2][2]; + unsigned int cbp; + short *blocks; +}; + +struct vlMpeg2MacroBlockBatch +{ + struct vlSurface *past_surface; + struct vlSurface *future_surface; + enum vlPictureType picture_type; + enum vlFieldOrder field_order; + unsigned int num_macroblocks; + struct vlMpeg2MacroBlock *macroblocks; +}; + +#endif diff --git a/src/gallium/state_trackers/g3dvl/vl_util.c b/src/gallium/state_trackers/g3dvl/vl_util.c new file mode 100644 index 0000000000..50aa9af66f --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_util.c @@ -0,0 +1,16 @@ +#include "vl_util.h" +#include <assert.h> + +unsigned int vlRoundUpPOT(unsigned int x) +{ + unsigned int i; + + assert(x > 0); + + --x; + + for (i = 1; i < sizeof(unsigned int) * 8; i <<= 1) + x |= x >> i; + + return x + 1; +} diff --git a/src/gallium/state_trackers/g3dvl/vl_util.h b/src/gallium/state_trackers/g3dvl/vl_util.h new file mode 100644 index 0000000000..bc98e79df4 --- /dev/null +++ b/src/gallium/state_trackers/g3dvl/vl_util.h @@ -0,0 +1,6 @@ +#ifndef vl_util_h +#define vl_util_h + +unsigned int vlRoundUpPOT(unsigned int x); + +#endif diff --git a/src/gallium/state_trackers/python/README b/src/gallium/state_trackers/python/README new file mode 100644 index 0000000000..8f45fb6d1b --- /dev/null +++ b/src/gallium/state_trackers/python/README @@ -0,0 +1,33 @@ +This directory contains Python bindings to Gallium3D. It looks like a state +tracker from the pipe driver perspective, and it looks like a pipe driver from +the python script perspective. + + +To build you'll need: +* Python (with development packages) +* SCons +* SWIG +* Python Imaging Library with TK support (for the samples) + +Invoke scons on the top dir as + + scons statetrackers=python + +To use do + + export PYTHONPATH=build/XXXX-XXXX-XXXX/gallium/state_trackers/python + +and then try running + + python src/gallium/state_trackers/python/samples/tri.py + +which should show a triangle. + + +This is still work in progress: +- errors are not handled properly and almost always result in crash +- state atoms with array members are awkward to set +- there no efficient way to view images + +-- +Jose Fonseca <jrfonseca@tungstengraphics.com> diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript new file mode 100644 index 0000000000..1581182aec --- /dev/null +++ b/src/gallium/state_trackers/python/SConscript @@ -0,0 +1,34 @@ +import sys +import os.path + +Import('*') + +if 'python' in env['statetrackers']: + + env = env.Clone() + + env.Tool('python') + + env.Tool('swig') + env.Append(SWIGPATH = ['#src/gallium/include', '#src/gallium/include/pipe']) + env.Append(SWIGFLAGS = ['-python', '-keyword']) + + env.Append(CPPPATH = '.') + + pyst = env.ConvenienceLibrary( + target = 'pyst', + source = [ + 'gallium.i', + 'st_device.c', + 'st_sample.c', + 'st_softpipe_winsys.c', + ], + ) + + env.SharedLibrary( + target = '_gallium', + source = [ + 'st_hardpipe_winsys.c', + ], + LIBS = [pyst, softpipe, trace] + auxiliaries + env['LIBS'], + ) diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i new file mode 100644 index 0000000000..f4c4b36ea7 --- /dev/null +++ b/src/gallium/state_trackers/python/gallium.i @@ -0,0 +1,103 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca <jrfonseca@tungstengraphics.com> + */ + +%module gallium; + +%{ + +#include <stdio.h> +#include <Python.h> + +#include "pipe/p_screen.h" +#include "pipe/p_context.h" +#include "pipe/p_inlines.h" +#include "util/u_memory.h" +#include "pipe/p_shader_tokens.h" +#include "cso_cache/cso_context.h" +#include "util/u_draw_quad.h" +#include "util/u_tile.h" +#include "tgsi/tgsi_text.h" +#include "tgsi/tgsi_dump.h" + +#include "st_device.h" +#include "st_sample.h" + +%} + +%include "typemaps.i" + +%include "cstring.i" + +%include "carrays.i" +%array_class(unsigned char, ByteArray); +%array_class(int, IntArray); +%array_class(unsigned, UnsignedArray); +%array_class(float, FloatArray); + + +%rename(Device) st_device; +%rename(Context) st_context; +%rename(Texture) pipe_texture; +%rename(Surface) pipe_surface; +%rename(Buffer) st_buffer; + +%rename(BlendColor) pipe_blend_color; +%rename(Blend) pipe_blend_state; +%rename(Clip) pipe_clip_state; +%rename(ConstantBuffer) pipe_constant_buffer; +%rename(Depth) pipe_depth_state; +%rename(Stencil) pipe_stencil_state; +%rename(Alpha) pipe_alpha_state; +%rename(DepthStencilAlpha) pipe_depth_stencil_alpha_state; +%rename(FormatBlock) pipe_format_block; +%rename(Framebuffer) pipe_framebuffer_state; +%rename(PolyStipple) pipe_poly_stipple; +%rename(Rasterizer) pipe_rasterizer_state; +%rename(Sampler) pipe_sampler_state; +%rename(Scissor) pipe_scissor_state; +%rename(Shader) pipe_shader_state; +%rename(VertexBuffer) pipe_vertex_buffer; +%rename(VertexElement) pipe_vertex_element; +%rename(Viewport) pipe_viewport_state; + + +%include "p_compiler.i" +%include "pipe/p_defines.h"; +%include "p_format.i" + +%include "p_device.i" +%include "p_context.i" +%include "p_texture.i" +%include "p_state.i" + diff --git a/src/gallium/state_trackers/python/p_compiler.i b/src/gallium/state_trackers/python/p_compiler.i new file mode 100644 index 0000000000..15f6ba5b9d --- /dev/null +++ b/src/gallium/state_trackers/python/p_compiler.i @@ -0,0 +1,29 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +typedef unsigned char ubyte; diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i new file mode 100644 index 0000000000..1fdcec639f --- /dev/null +++ b/src/gallium/state_trackers/python/p_context.i @@ -0,0 +1,289 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca <jrfonseca@tungstengraphics.com> + */ + +%nodefaultctor st_context; +%nodefaultdtor st_context; + +struct st_context { +}; + +%extend st_context { + + ~st_context() { + st_context_destroy($self); + } + + /* + * State functions (create/bind/destroy state objects) + */ + + void set_blend( const struct pipe_blend_state *state ) { + cso_set_blend($self->cso, state); + } + + void set_sampler( unsigned index, const struct pipe_sampler_state *state ) { + cso_single_sampler($self->cso, index, state); + cso_single_sampler_done($self->cso); + } + + void set_rasterizer( const struct pipe_rasterizer_state *state ) { + cso_set_rasterizer($self->cso, state); + } + + void set_depth_stencil_alpha(const struct pipe_depth_stencil_alpha_state *state) { + cso_set_depth_stencil_alpha($self->cso, state); + } + + void set_fragment_shader( const struct pipe_shader_state *state ) { + void *fs; + + if(!state) { + cso_set_fragment_shader_handle($self->cso, NULL); + return; + } + + fs = $self->pipe->create_fs_state($self->pipe, state); + if(!fs) + return; + + if(cso_set_fragment_shader_handle($self->cso, fs) != PIPE_OK) + return; + + cso_delete_fragment_shader($self->cso, $self->fs); + $self->fs = fs; + } + + void set_vertex_shader( const struct pipe_shader_state *state ) { + void *vs; + + if(!state) { + cso_set_vertex_shader_handle($self->cso, NULL); + return; + } + + vs = $self->pipe->create_vs_state($self->pipe, state); + if(!vs) + return; + + if(cso_set_vertex_shader_handle($self->cso, vs) != PIPE_OK) + return; + + cso_delete_vertex_shader($self->cso, $self->vs); + $self->vs = vs; + } + + /* + * Parameter-like state (or properties) + */ + + void set_blend_color(const struct pipe_blend_color *state ) { + cso_set_blend_color($self->cso, state); + } + + void set_clip(const struct pipe_clip_state *state ) { + $self->pipe->set_clip_state($self->pipe, state); + } + + void set_constant_buffer(unsigned shader, unsigned index, + struct st_buffer *buffer ) + { + struct pipe_constant_buffer state; + memset(&state, 0, sizeof(state)); + state.buffer = buffer ? buffer->buffer : NULL; + state.size = buffer->buffer->size; + $self->pipe->set_constant_buffer($self->pipe, shader, index, &state); + } + + void set_framebuffer(const struct pipe_framebuffer_state *state ) { + cso_set_framebuffer($self->cso, state); + } + + void set_polygon_stipple(const struct pipe_poly_stipple *state ) { + $self->pipe->set_polygon_stipple($self->pipe, state); + } + + void set_scissor(const struct pipe_scissor_state *state ) { + $self->pipe->set_scissor_state($self->pipe, state); + } + + void set_viewport(const struct pipe_viewport_state *state) { + cso_set_viewport($self->cso, state); + } + + void set_sampler_texture(unsigned index, + struct pipe_texture *texture) { + if(!texture) + texture = $self->default_texture; + pipe_texture_reference(&$self->sampler_textures[index], texture); + $self->pipe->set_sampler_textures($self->pipe, + PIPE_MAX_SAMPLERS, + $self->sampler_textures); + } + + void set_vertex_buffer(unsigned index, + unsigned pitch, + unsigned max_index, + unsigned buffer_offset, + struct st_buffer *buffer) + { + unsigned i; + struct pipe_vertex_buffer state; + + memset(&state, 0, sizeof(state)); + state.pitch = pitch; + state.max_index = max_index; + state.buffer_offset = buffer_offset; + state.buffer = buffer ? buffer->buffer : NULL; + + memcpy(&$self->vertex_buffers[index], &state, sizeof(state)); + + for(i = 0; i < PIPE_MAX_ATTRIBS; ++i) + if(self->vertex_buffers[i].buffer) + $self->num_vertex_buffers = i + 1; + + $self->pipe->set_vertex_buffers($self->pipe, + $self->num_vertex_buffers, + $self->vertex_buffers); + } + + void set_vertex_element(unsigned index, + const struct pipe_vertex_element *element) + { + memcpy(&$self->vertex_elements[index], element, sizeof(*element)); + } + + void set_vertex_elements(unsigned num) + { + $self->num_vertex_elements = num; + $self->pipe->set_vertex_elements($self->pipe, + $self->num_vertex_elements, + $self->vertex_elements); + } + + /* + * Draw functions + */ + + void draw_arrays(unsigned mode, unsigned start, unsigned count) { + $self->pipe->draw_arrays($self->pipe, mode, start, count); + } + + void draw_elements( struct st_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) + { + $self->pipe->draw_elements($self->pipe, + indexBuffer->buffer, + indexSize, + mode, start, count); + } + + void draw_range_elements( struct st_buffer *indexBuffer, + unsigned indexSize, unsigned minIndex, unsigned maxIndex, + unsigned mode, unsigned start, unsigned count) + { + $self->pipe->draw_range_elements($self->pipe, + indexBuffer->buffer, + indexSize, minIndex, maxIndex, + mode, start, count); + } + + void draw_vertices(unsigned prim, + unsigned num_verts, + unsigned num_attribs, + const float *vertices) + { + struct pipe_context *pipe = $self->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_buffer *vbuf; + float *map; + unsigned size; + + size = num_verts * num_attribs * 4 * sizeof(float); + + vbuf = pipe_buffer_create(screen, + 32, + PIPE_BUFFER_USAGE_VERTEX, + size); + if(!vbuf) + goto error1; + + map = pipe_buffer_map(screen, vbuf, PIPE_BUFFER_USAGE_CPU_WRITE); + if (!map) + goto error2; + memcpy(map, vertices, size); + pipe_buffer_unmap(screen, vbuf); + + util_draw_vertex_buffer(pipe, vbuf, prim, num_verts, num_attribs); + +error2: + pipe_buffer_reference(screen, &vbuf, NULL); +error1: + ; + } + + void + flush(unsigned flags = 0) { + struct pipe_fence_handle *fence = NULL; + $self->pipe->flush($self->pipe, flags | PIPE_FLUSH_RENDER_CACHE, &fence); + /* TODO: allow asynchronous operation */ + $self->pipe->winsys->fence_finish( $self->pipe->winsys, fence, 0 ); + $self->pipe->winsys->fence_reference( $self->pipe->winsys, &fence, NULL ); + } + + /* + * Surface functions + */ + + void surface_copy(int do_flip, + struct pipe_surface *dest, + unsigned destx, unsigned desty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) { + $self->pipe->surface_copy($self->pipe, do_flip, dest, destx, desty, src, srcx, srcy, width, height); + } + + void surface_fill(struct pipe_surface *dst, + unsigned x, unsigned y, + unsigned width, unsigned height, + unsigned value) { + $self->pipe->surface_fill($self->pipe, dst, x, y, width, height, value); + } + + void surface_clear(struct pipe_surface *surface, unsigned value = 0) { + $self->pipe->clear($self->pipe, surface, value); + } + +}; diff --git a/src/gallium/state_trackers/python/p_device.i b/src/gallium/state_trackers/python/p_device.i new file mode 100644 index 0000000000..84fd2e4349 --- /dev/null +++ b/src/gallium/state_trackers/python/p_device.i @@ -0,0 +1,130 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca <jrfonseca@tungstengraphics.com> + */ + + +%nodefaultctor st_device; +%nodefaultdtor st_device; + + +struct st_device { +}; + +%newobject st_device::texture_create; +%newobject st_device::context_create; +%newobject st_device::buffer_create; + +%extend st_device { + + st_device(int hardware = 1) { + return st_device_create(hardware ? TRUE : FALSE); + } + + ~st_device() { + st_device_destroy($self); + } + + const char * get_name( void ) { + return $self->screen->get_name($self->screen); + } + + const char * get_vendor( void ) { + return $self->screen->get_vendor($self->screen); + } + + /** + * Query an integer-valued capability/parameter/limit + * \param param one of PIPE_CAP_x + */ + int get_param( int param ) { + return $self->screen->get_param($self->screen, param); + } + + /** + * Query a float-valued capability/parameter/limit + * \param param one of PIPE_CAP_x + */ + float get_paramf( int param ) { + return $self->screen->get_paramf($self->screen, param); + } + + /** + * Check if the given pipe_format is supported as a texture or + * drawing surface. + * \param type one of PIPE_TEXTURE, PIPE_SURFACE + */ + int is_format_supported( enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags ) { + return $self->screen->is_format_supported( $self->screen, + format, + target, + tex_usage, + geom_flags ); + } + + struct st_context * + context_create(void) { + return st_context_create($self); + } + + struct pipe_texture * + texture_create( + enum pipe_format format, + unsigned width, + unsigned height, + unsigned depth = 1, + unsigned last_level = 0, + enum pipe_texture_target target = PIPE_TEXTURE_2D, + unsigned tex_usage = 0 + ) { + struct pipe_texture templat; + memset(&templat, 0, sizeof(templat)); + templat.format = format; + pf_get_block(templat.format, &templat.block); + templat.width[0] = width; + templat.height[0] = height; + templat.depth[0] = depth; + templat.last_level = last_level; + templat.target = target; + templat.tex_usage = tex_usage; + return $self->screen->texture_create($self->screen, &templat); + } + + struct st_buffer * + buffer_create(unsigned size, unsigned alignment = 0, unsigned usage = 0) { + return st_buffer_create($self, alignment, usage, size); + } + +}; diff --git a/src/gallium/state_trackers/python/p_format.i b/src/gallium/state_trackers/python/p_format.i new file mode 100644 index 0000000000..26fb12b387 --- /dev/null +++ b/src/gallium/state_trackers/python/p_format.i @@ -0,0 +1,162 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright (c) 2008 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * XXX: SWIG can't parse p_format.h, so we need to duplicate the relevant + * declarations here + */ + +%{ +#include "pipe/p_format.h" +%} + +enum pipe_format { + PIPE_FORMAT_NONE, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_X8R8G8B8_UNORM, + PIPE_FORMAT_B8G8R8A8_UNORM, + PIPE_FORMAT_B8G8R8X8_UNORM, + PIPE_FORMAT_A1R5G5B5_UNORM, + PIPE_FORMAT_A4R4G4B4_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_A2B10G10R10_UNORM, + PIPE_FORMAT_L8_UNORM, + PIPE_FORMAT_A8_UNORM, + PIPE_FORMAT_I8_UNORM, + PIPE_FORMAT_A8L8_UNORM, + PIPE_FORMAT_L16_UNORM, + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + PIPE_FORMAT_Z16_UNORM, + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_Z32_FLOAT, + PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_Z24S8_UNORM, + PIPE_FORMAT_X8Z24_UNORM, + PIPE_FORMAT_Z24X8_UNORM, + PIPE_FORMAT_S8_UNORM, + PIPE_FORMAT_R64_FLOAT, + PIPE_FORMAT_R64G64_FLOAT, + PIPE_FORMAT_R64G64B64_FLOAT, + PIPE_FORMAT_R64G64B64A64_FLOAT, + PIPE_FORMAT_R32_FLOAT, + PIPE_FORMAT_R32G32_FLOAT, + PIPE_FORMAT_R32G32B32_FLOAT, + PIPE_FORMAT_R32G32B32A32_FLOAT, + PIPE_FORMAT_R32_UNORM, + PIPE_FORMAT_R32G32_UNORM, + PIPE_FORMAT_R32G32B32_UNORM, + PIPE_FORMAT_R32G32B32A32_UNORM, + PIPE_FORMAT_R32_USCALED, + PIPE_FORMAT_R32G32_USCALED, + PIPE_FORMAT_R32G32B32_USCALED, + PIPE_FORMAT_R32G32B32A32_USCALED, + PIPE_FORMAT_R32_SNORM, + PIPE_FORMAT_R32G32_SNORM, + PIPE_FORMAT_R32G32B32_SNORM, + PIPE_FORMAT_R32G32B32A32_SNORM, + PIPE_FORMAT_R32_SSCALED, + PIPE_FORMAT_R32G32_SSCALED, + PIPE_FORMAT_R32G32B32_SSCALED, + PIPE_FORMAT_R32G32B32A32_SSCALED, + PIPE_FORMAT_R16_UNORM, + PIPE_FORMAT_R16G16_UNORM, + PIPE_FORMAT_R16G16B16_UNORM, + PIPE_FORMAT_R16G16B16A16_UNORM, + PIPE_FORMAT_R16_USCALED, + PIPE_FORMAT_R16G16_USCALED, + PIPE_FORMAT_R16G16B16_USCALED, + PIPE_FORMAT_R16G16B16A16_USCALED, + PIPE_FORMAT_R16_SNORM, + PIPE_FORMAT_R16G16_SNORM, + PIPE_FORMAT_R16G16B16_SNORM, + PIPE_FORMAT_R16G16B16A16_SNORM, + PIPE_FORMAT_R16_SSCALED, + PIPE_FORMAT_R16G16_SSCALED, + PIPE_FORMAT_R16G16B16_SSCALED, + PIPE_FORMAT_R16G16B16A16_SSCALED, + PIPE_FORMAT_R8_UNORM, + PIPE_FORMAT_R8G8_UNORM, + PIPE_FORMAT_R8G8B8_UNORM, + PIPE_FORMAT_R8G8B8A8_UNORM, + PIPE_FORMAT_R8G8B8X8_UNORM, + PIPE_FORMAT_R8_USCALED, + PIPE_FORMAT_R8G8_USCALED, + PIPE_FORMAT_R8G8B8_USCALED, + PIPE_FORMAT_R8G8B8A8_USCALED, + PIPE_FORMAT_R8G8B8X8_USCALED, + PIPE_FORMAT_R8_SNORM, + PIPE_FORMAT_R8G8_SNORM, + PIPE_FORMAT_R8G8B8_SNORM, + PIPE_FORMAT_R8G8B8A8_SNORM, + PIPE_FORMAT_R8G8B8X8_SNORM, + PIPE_FORMAT_B6G5R5_SNORM, + PIPE_FORMAT_A8B8G8R8_SNORM, + PIPE_FORMAT_X8B8G8R8_SNORM, + PIPE_FORMAT_R8_SSCALED, + PIPE_FORMAT_R8G8_SSCALED, + PIPE_FORMAT_R8G8B8_SSCALED, + PIPE_FORMAT_R8G8B8A8_SSCALED, + PIPE_FORMAT_R8G8B8X8_SSCALED, + PIPE_FORMAT_R32_FIXED, + PIPE_FORMAT_R32G32_FIXED, + PIPE_FORMAT_R32G32B32_FIXED, + PIPE_FORMAT_R32G32B32A32_FIXED, + + PIPE_FORMAT_L8_SRGB, + PIPE_FORMAT_A8L8_SRGB, + PIPE_FORMAT_R8G8B8_SRGB, + PIPE_FORMAT_R8G8B8A8_SRGB, + PIPE_FORMAT_R8G8B8X8_SRGB, + PIPE_FORMAT_A8R8G8B8_SRGB, + PIPE_FORMAT_X8R8G8B8_SRGB, + PIPE_FORMAT_B8G8R8A8_SRGB, + PIPE_FORMAT_B8G8R8X8_SRGB, + + PIPE_FORMAT_X8UB8UG8SR8S_NORM, + PIPE_FORMAT_B6UG5SR5S_NORM, + + PIPE_FORMAT_DXT1_RGB, + PIPE_FORMAT_DXT1_RGBA, + PIPE_FORMAT_DXT3_RGBA, + PIPE_FORMAT_DXT5_RGBA, + + PIPE_FORMAT_DXT1_SRGB, + PIPE_FORMAT_DXT1_SRGBA, + PIPE_FORMAT_DXT3_SRGBA, + PIPE_FORMAT_DXT5_SRGBA, +}; + + +struct pipe_format_block +{ + unsigned size; + unsigned width; + unsigned height; +}; + diff --git a/src/gallium/state_trackers/python/p_state.i b/src/gallium/state_trackers/python/p_state.i new file mode 100644 index 0000000000..7f5760b3b6 --- /dev/null +++ b/src/gallium/state_trackers/python/p_state.i @@ -0,0 +1,109 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca <jrfonseca@tungstengraphics.com> + */ + +%module gallium; + +%ignore winsys; +%ignore pipe_vertex_buffer::buffer; + +%include "pipe/p_state.h"; + + +%array_class(struct pipe_stencil_state, StencilArray); + + +%extend pipe_framebuffer_state { + + pipe_framebuffer_state(void) { + return CALLOC_STRUCT(pipe_framebuffer_state); + } + + ~pipe_framebuffer_state() { + unsigned index; + for(index = 0; index < PIPE_MAX_COLOR_BUFS; ++index) + pipe_surface_reference(&$self->cbufs[index], NULL); + pipe_surface_reference(&$self->zsbuf, NULL); + FREE($self); + } + + void + set_cbuf(unsigned index, struct pipe_surface *surface) { + pipe_surface_reference(&$self->cbufs[index], surface); + } + + void + set_zsbuf(struct pipe_surface *surface) { + pipe_surface_reference(&$self->zsbuf, surface); + } + +}; + + +%extend pipe_shader_state { + + pipe_shader_state(const char *text, unsigned num_tokens = 1024) { + struct tgsi_token *tokens; + struct pipe_shader_state *shader; + + tokens = MALLOC(num_tokens * sizeof(struct tgsi_token)); + if(!tokens) + goto error1; + + if(tgsi_text_translate(text, tokens, num_tokens ) != TRUE) + goto error2; + + shader = CALLOC_STRUCT(pipe_shader_state); + if(!shader) + goto error3; + + shader->tokens = tokens; + + return shader; + +error3: +error2: + FREE(tokens); +error1: + return NULL; + } + + ~pipe_shader_state() { + FREE((void*)$self->tokens); + FREE($self); + } + + void dump(unsigned flags = 0) { + tgsi_dump($self->tokens, flags); + } +} diff --git a/src/gallium/state_trackers/python/p_texture.i b/src/gallium/state_trackers/python/p_texture.i new file mode 100644 index 0000000000..08ba0ebe4d --- /dev/null +++ b/src/gallium/state_trackers/python/p_texture.i @@ -0,0 +1,232 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca <jrfonseca@tungstengraphics.com> + */ + + +%nodefaultctor pipe_texture; +%nodefaultctor pipe_surface; +%nodefaultctor st_buffer; + +%nodefaultdtor pipe_texture; +%nodefaultdtor pipe_surface; +%nodefaultdtor st_buffer; + +%ignore pipe_texture::screen; + +%ignore pipe_surface::winsys; +%immutable pipe_surface::texture; +%immutable pipe_surface::buffer; + +%newobject pipe_texture::get_surface; + + +%extend pipe_texture { + + ~pipe_texture() { + struct pipe_texture *ptr = $self; + pipe_texture_reference(&ptr, NULL); + } + + unsigned get_width(unsigned level=0) { + return $self->width[level]; + } + + unsigned get_height(unsigned level=0) { + return $self->height[level]; + } + + unsigned get_depth(unsigned level=0) { + return $self->depth[level]; + } + + unsigned get_nblocksx(unsigned level=0) { + return $self->nblocksx[level]; + } + + unsigned get_nblocksy(unsigned level=0) { + return $self->nblocksy[level]; + } + + /** Get a surface which is a "view" into a texture */ + struct pipe_surface * + get_surface(unsigned face=0, unsigned level=0, unsigned zslice=0, unsigned usage=0 ) + { + struct pipe_screen *screen = $self->screen; + return screen->get_tex_surface(screen, $self, face, level, zslice, usage); + } + +}; + + +%extend pipe_surface { + + ~pipe_surface() { + struct pipe_surface *ptr = $self; + pipe_surface_reference(&ptr, NULL); + } + + // gets mapped to pipe_surface_map automatically + void * map( unsigned flags ); + + // gets mapped to pipe_surface_unmap automatically + void unmap( void ); + + void + get_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, char *raw, unsigned stride) { + pipe_get_tile_raw($self, x, y, w, h, raw, stride); + } + + void + put_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, const char *raw, unsigned stride) { + pipe_put_tile_raw($self, x, y, w, h, raw, stride); + } + + void + get_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, float *rgba) { + pipe_get_tile_rgba($self, x, y, w, h, rgba); + } + + void + put_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba) { + pipe_put_tile_rgba($self, x, y, w, h, rgba); + } + + void + get_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, unsigned *z) { + pipe_get_tile_z($self, x, y, w, h, z); + } + + void + put_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned *z) { + pipe_put_tile_z($self, x, y, w, h, z); + } + + void + sample_rgba(float *rgba) { + st_sample_surface($self, rgba); + } + + unsigned + compare_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba, float tol = 0.0) + { + float *rgba2; + const float *p1; + const float *p2; + unsigned i, j, n; + + rgba2 = MALLOC(h*w*4*sizeof(float)); + if(!rgba2) + return ~0; + + pipe_get_tile_rgba($self, x, y, w, h, rgba2); + + p1 = rgba; + p2 = rgba2; + n = 0; + for(i = h*w; i; --i) { + unsigned differs = 0; + for(j = 4; j; --j) { + float delta = *p2++ - *p1++; + if (delta < -tol || delta > tol) + differs = 1; + } + n += differs; + } + + FREE(rgba2); + + return n; + } + +}; + +struct st_buffer { +}; + +%extend st_buffer { + + ~st_buffer() { + st_buffer_destroy($self); + } + + unsigned __len__(void) + { + assert($self->buffer->refcount); + return $self->buffer->size; + } + + %cstring_output_allocate_size(char **STRING, int *LENGTH, free(*$1)); + void read(char **STRING, int *LENGTH) + { + struct pipe_screen *screen = $self->st_dev->screen; + const char *map; + + assert($self->buffer->refcount); + + *LENGTH = $self->buffer->size; + *STRING = (char *) malloc($self->buffer->size); + if(!*STRING) + return; + + map = pipe_buffer_map(screen, $self->buffer, PIPE_BUFFER_USAGE_CPU_READ); + if(map) { + memcpy(*STRING, map, $self->buffer->size); + pipe_buffer_unmap(screen, $self->buffer); + } + } + + %cstring_input_binary(const char *STRING, unsigned LENGTH); + void write(const char *STRING, unsigned LENGTH, unsigned offset = 0) + { + struct pipe_screen *screen = $self->st_dev->screen; + char *map; + + assert($self->buffer->refcount); + + if(offset > $self->buffer->size) { + PyErr_SetString(PyExc_ValueError, "offset must be smaller than buffer size"); + return; + } + + if(offset + LENGTH > $self->buffer->size) { + PyErr_SetString(PyExc_ValueError, "data length must fit inside the buffer"); + return; + } + + map = pipe_buffer_map(screen, $self->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); + if(map) { + memcpy(map + offset, STRING, LENGTH); + pipe_buffer_unmap(screen, $self->buffer); + } + } +}; diff --git a/src/gallium/state_trackers/python/retrace/README b/src/gallium/state_trackers/python/retrace/README new file mode 100644 index 0000000000..822cd11404 --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/README @@ -0,0 +1,17 @@ +This is an application written in python to replay the traces captured by the + trace pipe driver. + + +To use it follow the instructions in src/gallium/drivers/trace/README and +src/gallium/state_trackers/python/README, and then do + + python src/gallium/state_trackers/python/samples/retrace/interpreter.py filename.trace + + +This is still work in progress: +- not everything is captured/replayed + - surface/textures contents +- any tiny error will result in a crash + +-- +Jose Fonseca <jrfonseca@tungstengraphics.com> diff --git a/src/gallium/state_trackers/python/retrace/format.py b/src/gallium/state_trackers/python/retrace/format.py new file mode 100755 index 0000000000..0bf6baf0b9 --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/format.py @@ -0,0 +1,100 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +class Formatter: + '''Plain formatter''' + + def __init__(self, stream): + self.stream = stream + + def text(self, text): + self.stream.write(text) + + def newline(self): + self.text('\n') + + def function(self, name): + self.text(name) + + def variable(self, name): + self.text(name) + + def literal(self, value): + self.text(str(value)) + + def address(self, addr): + self.text(str(addr)) + + +class AnsiFormatter(Formatter): + '''Formatter for plain-text files which outputs ANSI escape codes. See + http://en.wikipedia.org/wiki/ANSI_escape_code for more information + concerning ANSI escape codes. + ''' + + _csi = '\33[' + + _normal = '0m' + _bold = '1m' + _italic = '3m' + _red = '31m' + _green = '32m' + _blue = '34m' + + def _escape(self, code): + self.text(self._csi + code) + + def function(self, name): + self._escape(self._bold) + Formatter.function(self, name) + self._escape(self._normal) + + def variable(self, name): + self._escape(self._italic) + Formatter.variable(self, name) + self._escape(self._normal) + + def literal(self, value): + self._escape(self._blue) + Formatter.literal(self, value) + self._escape(self._normal) + + def address(self, value): + self._escape(self._green) + Formatter.address(self, value) + self._escape(self._normal) + + + +def DefaultFormatter(stream): + if stream.isatty(): + return AnsiFormatter(stream) + else: + return Formatter(stream) + diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py new file mode 100755 index 0000000000..f418f80d7b --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -0,0 +1,578 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +import sys +import struct + +import gallium +import model +import parser + + +def make_image(surface): + pixels = gallium.FloatArray(surface.height*surface.width*4) + surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) + + import Image + outimage = Image.new( + mode='RGB', + size=(surface.width, surface.height), + color=(0,0,0)) + outpixels = outimage.load() + for y in range(0, surface.height): + for x in range(0, surface.width): + offset = (y*surface.width + x)*4 + r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] + outpixels[x, y] = r, g, b + return outimage + +def save_image(filename, surface): + outimage = make_image(surface) + outimage.save(filename, "PNG") + +def show_image(surface): + outimage = make_image(surface) + + import Tkinter as tk + from PIL import Image, ImageTk + root = tk.Tk() + + root.title('background image') + + image1 = ImageTk.PhotoImage(outimage) + w = image1.width() + h = image1.height() + x = 100 + y = 100 + root.geometry("%dx%d+%d+%d" % (w, h, x, y)) + panel1 = tk.Label(root, image=image1) + panel1.pack(side='top', fill='both', expand='yes') + panel1.image = image1 + root.mainloop() + + + + +class Struct: + """C-like struct""" + + # A basic Python class can pass as a C-like structure + pass + + +struct_factories = { + "pipe_blend_color": gallium.BlendColor, + "pipe_blend_state": gallium.Blend, + #"pipe_clip_state": gallium.Clip, + #"pipe_constant_buffer": gallium.ConstantBuffer, + "pipe_depth_state": gallium.Depth, + "pipe_stencil_state": gallium.Stencil, + "pipe_alpha_state": gallium.Alpha, + "pipe_depth_stencil_alpha_state": gallium.DepthStencilAlpha, + "pipe_format_block": gallium.FormatBlock, + #"pipe_framebuffer_state": gallium.Framebuffer, + "pipe_poly_stipple": gallium.PolyStipple, + "pipe_rasterizer_state": gallium.Rasterizer, + "pipe_sampler_state": gallium.Sampler, + "pipe_scissor_state": gallium.Scissor, + #"pipe_shader_state": gallium.Shader, + #"pipe_vertex_buffer": gallium.VertexBuffer, + "pipe_vertex_element": gallium.VertexElement, + "pipe_viewport_state": gallium.Viewport, + #"pipe_texture": gallium.Texture, +} + + +member_array_factories = { + "pipe_rasterizer_state": {"sprite_coord_mode": gallium.ByteArray}, + "pipe_poly_stipple": {"stipple": gallium.UnsignedArray}, + "pipe_viewport_state": {"scale": gallium.FloatArray, "translate": gallium.FloatArray}, + #"pipe_clip_state": {"ucp": gallium.FloatArray}, + "pipe_depth_stencil_alpha_state": {"stencil": gallium.StencilArray}, + "pipe_blend_color": {"color": gallium.FloatArray}, + "pipe_sampler_state": {"border_color": gallium.FloatArray}, +} + + +class Translator(model.Visitor): + """Translate model arguments into regular Python objects""" + + def __init__(self, interpreter): + self.interpreter = interpreter + self.result = None + + def visit(self, node): + self.result = None + node.visit(self) + return self.result + + def visit_literal(self, node): + self.result = node.value + + def visit_named_constant(self, node): + # lookup the named constant in the gallium module + self.result = getattr(gallium, node.name) + + def visit_array(self, node): + array = [] + for element in node.elements: + array.append(self.visit(element)) + self.result = array + + def visit_struct(self, node): + struct_factory = struct_factories.get(node.name, Struct) + struct = struct_factory() + for member_name, member_node in node.members: + member_value = self.visit(member_node) + try: + array_factory = member_array_factories[node.name][member_name] + except KeyError: + pass + else: + assert isinstance(member_value, list) + array = array_factory(len(member_value)) + for i in range(len(member_value)): + array[i] = member_value[i] + member_value = array + #print node.name, member_name, member_value + assert isinstance(struct, Struct) or hasattr(struct, member_name) + setattr(struct, member_name, member_value) + self.result = struct + + def visit_pointer(self, node): + self.result = self.interpreter.lookup_object(node.address) + + +class Object: + + def __init__(self, interpreter, real): + self.interpreter = interpreter + self.real = real + + +class Global(Object): + + def __init__(self, interpreter, real): + self.interpreter = interpreter + self.real = real + + def pipe_winsys_create(self): + return Winsys(self.interpreter, gallium.Device()) + + def pipe_screen_create(self, winsys): + return Screen(self.interpreter, winsys.real) + + def pipe_context_create(self, screen): + context = screen.real.context_create() + return Context(self.interpreter, context) + + +class Winsys(Object): + + def __init__(self, interpreter, real): + self.interpreter = interpreter + self.real = real + + def get_name(self): + pass + + def user_buffer_create(self, data, size): + # We don't really care to distinguish between user and regular buffers + buffer = self.real.buffer_create(size, + 4, + gallium.PIPE_BUFFER_USAGE_CPU_READ | + gallium.PIPE_BUFFER_USAGE_CPU_WRITE ) + assert size == len(data) + buffer.write(data) + return buffer + + def buffer_create(self, alignment, usage, size): + return self.real.buffer_create(size, alignment, usage) + + def buffer_destroy(self, buffer): + pass + + def buffer_write(self, buffer, data, size): + assert size == len(data) + buffer.write(data) + + def fence_finish(self, fence, flags): + pass + + def fence_reference(self, dst, src): + pass + + def flush_frontbuffer(self, surface): + pass + + def surface_alloc(self): + return None + + def surface_release(self, surface): + pass + + +class Screen(Object): + + def destroy(self): + pass + + def get_name(self): + pass + + def get_vendor(self): + pass + + def get_param(self, param): + pass + + def get_paramf(self, param): + pass + + def is_format_supported(self, format, target, tex_usage, geom_flags): + return self.real.is_format_supported(format, target, tex_usage, geom_flags) + + def texture_create(self, template): + return self.real.texture_create( + format = template.format, + width = template.width[0], + height = template.height[0], + depth = template.depth[0], + last_level = template.last_level, + target = template.target, + tex_usage = template.tex_usage, + ) + + def texture_destroy(self, texture): + self.interpreter.unregister_object(texture) + + def texture_release(self, surface): + pass + + def get_tex_surface(self, texture, face, level, zslice, usage): + return texture.get_surface(face, level, zslice, usage) + + def tex_surface_destroy(self, surface): + self.interpreter.unregister_object(surface) + + def tex_surface_release(self, surface): + pass + + def surface_write(self, surface, data, stride, size): + assert surface.nblocksy * stride == size + surface.put_tile_raw(0, 0, surface.width, surface.height, data, stride) + + +class Context(Object): + + def __init__(self, interpreter, real): + Object.__init__(self, interpreter, real) + self.cbufs = [] + self.zsbuf = None + self.vbufs = [] + self.velems = [] + + def destroy(self): + pass + + def create_blend_state(self, state): + return state + + def bind_blend_state(self, state): + if state is not None: + self.real.set_blend(state) + + def delete_blend_state(self, state): + pass + + def create_sampler_state(self, state): + return state + + def delete_sampler_state(self, state): + pass + + def bind_sampler_states(self, n, states): + for i in range(n): + self.real.set_sampler(i, states[i]) + + def create_rasterizer_state(self, state): + return state + + def bind_rasterizer_state(self, state): + if state is not None: + self.real.set_rasterizer(state) + + def delete_rasterizer_state(self, state): + pass + + def create_depth_stencil_alpha_state(self, state): + return state + + def bind_depth_stencil_alpha_state(self, state): + if state is not None: + self.real.set_depth_stencil_alpha(state) + + def delete_depth_stencil_alpha_state(self, state): + pass + + def create_fs_state(self, state): + tokens = str(state.tokens) + shader = gallium.Shader(tokens) + return shader + + create_vs_state = create_fs_state + + def bind_fs_state(self, state): + self.real.set_fragment_shader(state) + + def bind_vs_state(self, state): + self.real.set_vertex_shader(state) + + def delete_fs_state(self, state): + pass + + delete_vs_state = delete_fs_state + + def set_blend_color(self, state): + self.real.set_blend_color(state) + + def set_clip_state(self, state): + _state = gallium.Clip() + _state.nr = state.nr + if state.nr: + # FIXME + ucp = gallium.FloatArray(gallium.PIPE_MAX_CLIP_PLANES*4) + for i in range(len(state.ucp)): + for j in range(len(state.ucp[i])): + ucp[i*4 + j] = state.ucp[i][j] + _state.ucp = ucp + self.real.set_clip(_state) + + def set_constant_buffer(self, shader, index, state): + if state is not None: + self.real.set_constant_buffer(shader, index, state.buffer) + + if 1: + data = state.buffer.read() + format = '4f' + index = 0 + for offset in range(0, len(data), struct.calcsize(format)): + x, y, z, w = struct.unpack_from(format, data, offset) + sys.stdout.write('\tCONST[%2u] = {%10.4f, %10.4f, %10.4f, %10.4f}\n' % (index, x, y, z, w)) + index += 1 + + def set_framebuffer_state(self, state): + _state = gallium.Framebuffer() + _state.width = state.width + _state.height = state.height + _state.num_cbufs = state.num_cbufs + for i in range(len(state.cbufs)): + _state.set_cbuf(i, state.cbufs[i]) + _state.set_zsbuf(state.zsbuf) + self.real.set_framebuffer(_state) + + self.cbufs = state.cbufs + self.zsbuf = state.zsbuf + + def set_polygon_stipple(self, state): + self.real.set_polygon_stipple(state) + + def set_scissor_state(self, state): + self.real.set_scissor(state) + + def set_viewport_state(self, state): + self.real.set_viewport(state) + + def set_sampler_textures(self, n, textures): + for i in range(n): + self.real.set_sampler_texture(i, textures[i]) + + def set_vertex_buffers(self, n, vbufs): + self.vbufs = vbufs[0:n] + for i in range(n): + vbuf = vbufs[i] + self.real.set_vertex_buffer( + i, + pitch = vbuf.pitch, + max_index = vbuf.max_index, + buffer_offset = vbuf.buffer_offset, + buffer = vbuf.buffer, + ) + + def set_vertex_elements(self, n, elements): + self.velems = elements[0:n] + for i in range(n): + self.real.set_vertex_element(i, elements[i]) + self.real.set_vertex_elements(n) + + def set_edgeflags(self, bitfield): + # FIXME + pass + + def dump_vertices(self, start, count): + for index in range(start, start + count): + if index >= start + 16: + sys.stdout.write('\t...\n') + break + sys.stdout.write('\t{\n') + for velem in self.velems: + vbuf = self.vbufs[velem.vertex_buffer_index] + + offset = vbuf.buffer_offset + velem.src_offset + vbuf.pitch*index + format = { + gallium.PIPE_FORMAT_R32_FLOAT: 'f', + gallium.PIPE_FORMAT_R32G32_FLOAT: '2f', + gallium.PIPE_FORMAT_R32G32B32_FLOAT: '3f', + gallium.PIPE_FORMAT_R32G32B32A32_FLOAT: '4f', + gallium.PIPE_FORMAT_B8G8R8A8_UNORM: '4B', + }[velem.src_format] + + data = vbuf.buffer.read() + values = struct.unpack_from(format, data, offset) + sys.stdout.write('\t\t{' + ', '.join(map(str, values)) + '},\n') + assert len(values) == velem.nr_components + sys.stdout.write('\t},\n') + + def dump_indices(self, ibuf, isize, start, count): + format = { + 1: 'B', + 2: 'H', + 4: 'I', + }[isize] + + assert struct.calcsize(format) == isize + + data = ibuf.read() + maxindex, minindex = 0, 0xffffffff + + sys.stdout.write('\t{\n') + for i in range(start, start + count): + if i >= start + 16: + sys.stdout.write('\t...\n') + break + offset = i*isize + index, = struct.unpack_from(format, data, offset) + sys.stdout.write('\t\t%u,\n' % index) + minindex = min(minindex, index) + maxindex = max(maxindex, index) + sys.stdout.write('\t},\n') + + return minindex, maxindex + + def draw_arrays(self, mode, start, count): + self.dump_vertices(start, count) + + self.real.draw_arrays(mode, start, count) + + def draw_elements(self, indexBuffer, indexSize, mode, start, count): + minindex, maxindex = self.dump_indices(indexBuffer, indexSize, start, count) + self.dump_vertices(minindex, maxindex - minindex) + + self.real.draw_elements(indexBuffer, indexSize, mode, start, count) + + def draw_range_elements(self, indexBuffer, indexSize, minIndex, maxIndex, mode, start, count): + minindex, maxindex = self.dump_indices(indexBuffer, indexSize, start, count) + minindex = min(minindex, minIndex) + maxindex = min(maxindex, maxIndex) + self.dump_vertices(minindex, maxindex - minindex) + + self.real.draw_range_elements(indexBuffer, indexSize, minIndex, maxIndex, mode, start, count) + + def flush(self, flags): + self.real.flush(flags) + if flags & gallium.PIPE_FLUSH_FRAME: + self._update() + return None + + def clear(self, surface, value): + self.real.surface_clear(surface, value) + + def _update(self): + self.real.flush() + + if self.cbufs and self.cbufs[0]: + show_image(self.cbufs[0]) + + +class Interpreter(parser.TraceDumper): + + ignore_calls = set(( + ('pipe_screen', 'is_format_supported'), + ('pipe_screen', 'get_param'), + ('pipe_screen', 'get_paramf'), + )) + + def __init__(self, stream): + parser.TraceDumper.__init__(self, stream) + self.objects = {} + self.result = None + self.globl = Global(self, None) + + def register_object(self, address, object): + self.objects[address] = object + + def unregister_object(self, object): + # FIXME: + pass + + def lookup_object(self, address): + return self.objects[address] + + def interpret(self, trace): + for call in trace.calls: + self.interpret_call(call) + + def handle_call(self, call): + + if (call.klass, call.method) in self.ignore_calls: + return + + parser.TraceDumper.handle_call(self, call) + + args = [self.interpret_arg(arg) for name, arg in call.args] + + if call.klass: + obj = args[0] + args = args[1:] + else: + obj = self.globl + + method = getattr(obj, call.method) + ret = method(*args) + + if call.ret and isinstance(call.ret, model.Pointer): + self.register_object(call.ret.address, ret) + + def interpret_arg(self, node): + translator = Translator(self) + return translator.visit(node) + + +if __name__ == '__main__': + parser.main(Interpreter) diff --git a/src/gallium/state_trackers/python/retrace/model.py b/src/gallium/state_trackers/python/retrace/model.py new file mode 100755 index 0000000000..ae0f4327d7 --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/model.py @@ -0,0 +1,211 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +'''Trace data model.''' + + +import sys +import string +import format + +try: + from cStringIO import StringIO +except ImportError: + from StringIO import StringIO + + +class Node: + + def visit(self, visitor): + raise NotImplementedError + + def __str__(self): + stream = StringIO() + formatter = format.DefaultFormatter(stream) + pretty_printer = PrettyPrinter(formatter) + self.visit(pretty_printer) + return stream.getvalue() + + +class Literal(Node): + + def __init__(self, value): + self.value = value + + def visit(self, visitor): + visitor.visit_literal(self) + + +class NamedConstant(Node): + + def __init__(self, name): + self.name = name + + def visit(self, visitor): + visitor.visit_named_constant(self) + + +class Array(Node): + + def __init__(self, elements): + self.elements = elements + + def visit(self, visitor): + visitor.visit_array(self) + + +class Struct(Node): + + def __init__(self, name, members): + self.name = name + self.members = members + + def visit(self, visitor): + visitor.visit_struct(self) + + +class Pointer(Node): + + def __init__(self, address): + self.address = address + + def visit(self, visitor): + visitor.visit_pointer(self) + + +class Call: + + def __init__(self, klass, method, args, ret): + self.klass = klass + self.method = method + self.args = args + self.ret = ret + + def visit(self, visitor): + visitor.visit_call(self) + + +class Trace: + + def __init__(self, calls): + self.calls = calls + + def visit(self, visitor): + visitor.visit_trace(self) + + +class Visitor: + + def visit_literal(self, node): + raise NotImplementedError + + def visit_named_constant(self, node): + raise NotImplementedError + + def visit_array(self, node): + raise NotImplementedError + + def visit_struct(self, node): + raise NotImplementedError + + def visit_pointer(self, node): + raise NotImplementedError + + def visit_call(self, node): + raise NotImplementedError + + def visit_trace(self, node): + raise NotImplementedError + + +class PrettyPrinter: + + def __init__(self, formatter): + self.formatter = formatter + + def visit_literal(self, node): + if isinstance(node.value, basestring): + if len(node.value) >= 4096 or node.value.strip(string.printable): + self.formatter.text('...') + return + + self.formatter.literal('"' + node.value + '"') + return + + self.formatter.literal(repr(node.value)) + + def visit_named_constant(self, node): + self.formatter.literal(node.name) + + def visit_array(self, node): + self.formatter.text('{') + sep = '' + for value in node.elements: + self.formatter.text(sep) + value.visit(self) + sep = ', ' + self.formatter.text('}') + + def visit_struct(self, node): + self.formatter.text('{') + sep = '' + for name, value in node.members: + self.formatter.text(sep) + self.formatter.variable(name) + self.formatter.text(' = ') + value.visit(self) + sep = ', ' + self.formatter.text('}') + + def visit_pointer(self, node): + self.formatter.address(node.address) + + def visit_call(self, node): + if node.klass is not None: + self.formatter.function(node.klass + '::' + node.method) + else: + self.formatter.function(node.method) + self.formatter.text('(') + sep = '' + for name, value in node.args: + self.formatter.text(sep) + self.formatter.variable(name) + self.formatter.text(' = ') + value.visit(self) + sep = ', ' + self.formatter.text(')') + if node.ret is not None: + self.formatter.text(' = ') + node.ret.visit(self) + + def visit_trace(self, node): + for call in node.calls: + call.visit(self) + self.formatter.newline() + diff --git a/src/gallium/state_trackers/python/retrace/parser.py b/src/gallium/state_trackers/python/retrace/parser.py new file mode 100755 index 0000000000..5205f2d8dd --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/parser.py @@ -0,0 +1,357 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +import sys +import xml.parsers.expat +import binascii + +from model import * + + +ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF = range(4) + + +class XmlToken: + + def __init__(self, type, name_or_data, attrs = None, line = None, column = None): + assert type in (ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF) + self.type = type + self.name_or_data = name_or_data + self.attrs = attrs + self.line = line + self.column = column + + def __str__(self): + if self.type == ELEMENT_START: + return '<' + self.name_or_data + ' ...>' + if self.type == ELEMENT_END: + return '</' + self.name_or_data + '>' + if self.type == CHARACTER_DATA: + return self.name_or_data + if self.type == EOF: + return 'end of file' + assert 0 + + +class XmlTokenizer: + """Expat based XML tokenizer.""" + + def __init__(self, fp, skip_ws = True): + self.fp = fp + self.tokens = [] + self.index = 0 + self.final = False + self.skip_ws = skip_ws + + self.character_pos = 0, 0 + self.character_data = '' + + self.parser = xml.parsers.expat.ParserCreate() + self.parser.StartElementHandler = self.handle_element_start + self.parser.EndElementHandler = self.handle_element_end + self.parser.CharacterDataHandler = self.handle_character_data + + def handle_element_start(self, name, attributes): + self.finish_character_data() + line, column = self.pos() + token = XmlToken(ELEMENT_START, name, attributes, line, column) + self.tokens.append(token) + + def handle_element_end(self, name): + self.finish_character_data() + line, column = self.pos() + token = XmlToken(ELEMENT_END, name, None, line, column) + self.tokens.append(token) + + def handle_character_data(self, data): + if not self.character_data: + self.character_pos = self.pos() + self.character_data += data + + def finish_character_data(self): + if self.character_data: + if not self.skip_ws or not self.character_data.isspace(): + line, column = self.character_pos + token = XmlToken(CHARACTER_DATA, self.character_data, None, line, column) + self.tokens.append(token) + self.character_data = '' + + def next(self): + size = 16*1024 + while self.index >= len(self.tokens) and not self.final: + self.tokens = [] + self.index = 0 + data = self.fp.read(size) + self.final = len(data) < size + data = data.rstrip('\0') + try: + self.parser.Parse(data, self.final) + except xml.parsers.expat.ExpatError, e: + #if e.code == xml.parsers.expat.errors.XML_ERROR_NO_ELEMENTS: + if e.code == 3: + pass + else: + raise e + if self.index >= len(self.tokens): + line, column = self.pos() + token = XmlToken(EOF, None, None, line, column) + else: + token = self.tokens[self.index] + self.index += 1 + return token + + def pos(self): + return self.parser.CurrentLineNumber, self.parser.CurrentColumnNumber + + +class TokenMismatch(Exception): + + def __init__(self, expected, found): + self.expected = expected + self.found = found + + def __str__(self): + return '%u:%u: %s expected, %s found' % (self.found.line, self.found.column, str(self.expected), str(self.found)) + + + +class XmlParser: + """Base XML document parser.""" + + def __init__(self, fp): + self.tokenizer = XmlTokenizer(fp) + self.consume() + + def consume(self): + self.token = self.tokenizer.next() + + def match_element_start(self, name): + return self.token.type == ELEMENT_START and self.token.name_or_data == name + + def match_element_end(self, name): + return self.token.type == ELEMENT_END and self.token.name_or_data == name + + def element_start(self, name): + while self.token.type == CHARACTER_DATA: + self.consume() + if self.token.type != ELEMENT_START: + raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token) + if self.token.name_or_data != name: + raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token) + attrs = self.token.attrs + self.consume() + return attrs + + def element_end(self, name): + while self.token.type == CHARACTER_DATA: + self.consume() + if self.token.type != ELEMENT_END: + raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token) + if self.token.name_or_data != name: + raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token) + self.consume() + + def character_data(self, strip = True): + data = '' + while self.token.type == CHARACTER_DATA: + data += self.token.name_or_data + self.consume() + if strip: + data = data.strip() + return data + + +class TraceParser(XmlParser): + + def parse(self): + self.element_start('trace') + while self.token.type not in (ELEMENT_END, EOF): + call = self.parse_call() + self.handle_call(call) + if self.token.type != EOF: + self.element_end('trace') + + def parse_call(self): + attrs = self.element_start('call') + klass = attrs['class'] + method = attrs['method'] + args = [] + ret = None + while self.token.type == ELEMENT_START: + if self.token.name_or_data == 'arg': + arg = self.parse_arg() + args.append(arg) + elif self.token.name_or_data == 'ret': + ret = self.parse_ret() + elif self.token.name_or_data == 'call': + # ignore nested function calls + self.parse_call() + else: + raise TokenMismatch("<arg ...> or <ret ...>", self.token) + self.element_end('call') + + return Call(klass, method, args, ret) + + def parse_arg(self): + attrs = self.element_start('arg') + name = attrs['name'] + value = self.parse_value() + self.element_end('arg') + + return name, value + + def parse_ret(self): + attrs = self.element_start('ret') + value = self.parse_value() + self.element_end('ret') + + return value + + def parse_value(self): + expected_tokens = ('null', 'bool', 'int', 'uint', 'float', 'string', 'enum', 'array', 'struct', 'ptr', 'bytes') + if self.token.type == ELEMENT_START: + if self.token.name_or_data in expected_tokens: + method = getattr(self, 'parse_' + self.token.name_or_data) + return method() + raise TokenMismatch(" or " .join(expected_tokens), self.token) + + def parse_null(self): + self.element_start('null') + self.element_end('null') + return Literal(None) + + def parse_bool(self): + self.element_start('bool') + value = int(self.character_data()) + self.element_end('bool') + return Literal(value) + + def parse_int(self): + self.element_start('int') + value = int(self.character_data()) + self.element_end('int') + return Literal(value) + + def parse_uint(self): + self.element_start('uint') + value = int(self.character_data()) + self.element_end('uint') + return Literal(value) + + def parse_float(self): + self.element_start('float') + value = float(self.character_data()) + self.element_end('float') + return Literal(value) + + def parse_enum(self): + self.element_start('enum') + name = self.character_data() + self.element_end('enum') + return NamedConstant(name) + + def parse_string(self): + self.element_start('string') + value = self.character_data() + self.element_end('string') + return Literal(value) + + def parse_bytes(self): + self.element_start('bytes') + value = binascii.a2b_hex(self.character_data()) + self.element_end('bytes') + return Literal(value) + + def parse_array(self): + self.element_start('array') + elems = [] + while self.token.type != ELEMENT_END: + elems.append(self.parse_elem()) + self.element_end('array') + return Array(elems) + + def parse_elem(self): + self.element_start('elem') + value = self.parse_value() + self.element_end('elem') + return value + + def parse_struct(self): + attrs = self.element_start('struct') + name = attrs['name'] + members = [] + while self.token.type != ELEMENT_END: + members.append(self.parse_member()) + self.element_end('struct') + return Struct(name, members) + + def parse_member(self): + attrs = self.element_start('member') + name = attrs['name'] + value = self.parse_value() + self.element_end('member') + + return name, value + + def parse_ptr(self): + self.element_start('ptr') + address = self.character_data() + self.element_end('ptr') + + return Pointer(address) + + def handle_call(self, call): + pass + + +class TraceDumper(TraceParser): + + def __init__(self, fp): + TraceParser.__init__(self, fp) + self.formatter = format.DefaultFormatter(sys.stdout) + self.pretty_printer = PrettyPrinter(self.formatter) + + def handle_call(self, call): + call.visit(self.pretty_printer) + self.formatter.newline() + + +def main(ParserFactory): + for arg in sys.argv[1:]: + if arg.endswith('.gz'): + import gzip + stream = gzip.GzipFile(arg, 'rt') + else: + stream = open(arg, 'rt') + parser = ParserFactory(stream) + parser.parse() + + +if __name__ == '__main__': + main(TraceDumper) diff --git a/src/gallium/state_trackers/python/samples/tri.py b/src/gallium/state_trackers/python/samples/tri.py new file mode 100644 index 0000000000..193479f7d6 --- /dev/null +++ b/src/gallium/state_trackers/python/samples/tri.py @@ -0,0 +1,229 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +from gallium import * + + +def make_image(surface): + pixels = FloatArray(surface.height*surface.width*4) + surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) + + import Image + outimage = Image.new( + mode='RGB', + size=(surface.width, surface.height), + color=(0,0,0)) + outpixels = outimage.load() + for y in range(0, surface.height): + for x in range(0, surface.width): + offset = (y*surface.width + x)*4 + r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] + outpixels[x, y] = r, g, b + return outimage + +def save_image(filename, surface): + outimage = make_image(surface) + outimage.save(filename, "PNG") + +def show_image(surface): + outimage = make_image(surface) + + import Tkinter as tk + from PIL import Image, ImageTk + root = tk.Tk() + + root.title('background image') + + image1 = ImageTk.PhotoImage(outimage) + w = image1.width() + h = image1.height() + x = 100 + y = 100 + root.geometry("%dx%d+%d+%d" % (w, h, x, y)) + panel1 = tk.Label(root, image=image1) + panel1.pack(side='top', fill='both', expand='yes') + panel1.image = image1 + root.mainloop() + + +def test(dev): + ctx = dev.context_create() + + width = 255 + height = 255 + + # disabled blending/masking + blend = Blend() + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.colormask = PIPE_MASK_RGBA + ctx.set_blend(blend) + + # no-op depth/stencil/alpha + depth_stencil_alpha = DepthStencilAlpha() + ctx.set_depth_stencil_alpha(depth_stencil_alpha) + + # rasterizer + rasterizer = Rasterizer() + rasterizer.front_winding = PIPE_WINDING_CW + rasterizer.cull_mode = PIPE_WINDING_NONE + rasterizer.bypass_clipping = 1 + rasterizer.scissor = 1 + #rasterizer.bypass_vs = 1 + ctx.set_rasterizer(rasterizer) + + # viewport (identity, we setup vertices in wincoords) + viewport = Viewport() + scale = FloatArray(4) + scale[0] = 1.0 + scale[1] = 1.0 + scale[2] = 1.0 + scale[3] = 1.0 + viewport.scale = scale + translate = FloatArray(4) + translate[0] = 0.0 + translate[1] = 0.0 + translate[2] = 0.0 + translate[3] = 0.0 + viewport.translate = translate + ctx.set_viewport(viewport) + + # samplers + sampler = Sampler() + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.normalized_coords = 1 + ctx.set_sampler(0, sampler) + + # scissor + scissor = Scissor() + scissor.minx = 0 + scissor.miny = 0 + scissor.maxx = width + scissor.maxy = height + ctx.set_scissor(scissor) + + clip = Clip() + clip.nr = 0 + ctx.set_clip(clip) + + # framebuffer + cbuf = dev.texture_create( + PIPE_FORMAT_X8R8G8B8_UNORM, + width, height, + tex_usage=PIPE_TEXTURE_USAGE_DISPLAY_TARGET, + ) + _cbuf = cbuf.get_surface(usage = PIPE_BUFFER_USAGE_GPU_READ|PIPE_BUFFER_USAGE_GPU_WRITE) + fb = Framebuffer() + fb.width = width + fb.height = height + fb.num_cbufs = 1 + fb.set_cbuf(0, _cbuf) + ctx.set_framebuffer(fb) + _cbuf.clear_value = 0x00000000 + ctx.surface_clear(_cbuf, _cbuf.clear_value) + del _cbuf + + # vertex shader + vs = Shader(''' + VERT1.1 + DCL IN[0], POSITION, CONSTANT + DCL IN[1], COLOR, CONSTANT + DCL OUT[0], POSITION, CONSTANT + DCL OUT[1], COLOR, CONSTANT + 0:MOV OUT[0], IN[0] + 1:MOV OUT[1], IN[1] + 2:END + ''') + ctx.set_vertex_shader(vs) + + # fragment shader + fs = Shader(''' + FRAG1.1 + DCL IN[0], COLOR, LINEAR + DCL OUT[0], COLOR, CONSTANT + 0:MOV OUT[0], IN[0] + 1:END + ''') + ctx.set_fragment_shader(fs) + + nverts = 3 + nattrs = 2 + verts = FloatArray(nverts * nattrs * 4) + + verts[ 0] = 128.0 # x1 + verts[ 1] = 32.0 # y1 + verts[ 2] = 0.0 # z1 + verts[ 3] = 1.0 # w1 + verts[ 4] = 1.0 # r1 + verts[ 5] = 0.0 # g1 + verts[ 6] = 0.0 # b1 + verts[ 7] = 1.0 # a1 + verts[ 8] = 32.0 # x2 + verts[ 9] = 224.0 # y2 + verts[10] = 0.0 # z2 + verts[11] = 1.0 # w2 + verts[12] = 0.0 # r2 + verts[13] = 1.0 # g2 + verts[14] = 0.0 # b2 + verts[15] = 1.0 # a2 + verts[16] = 224.0 # x3 + verts[17] = 224.0 # y3 + verts[18] = 0.0 # z3 + verts[19] = 1.0 # w3 + verts[20] = 0.0 # r3 + verts[21] = 0.0 # g3 + verts[22] = 1.0 # b3 + verts[23] = 1.0 # a3 + + ctx.draw_vertices(PIPE_PRIM_TRIANGLES, + nverts, + nattrs, + verts) + + ctx.flush() + + show_image(cbuf.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE)) + #save_image('tri.png', cbuf.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE)) + + + +def main(): + dev = Device() + test(dev) + + +if __name__ == '__main__': + main() diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c new file mode 100644 index 0000000000..95c1378a03 --- /dev/null +++ b/src/gallium/state_trackers/python/st_device.c @@ -0,0 +1,323 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_winsys.h" +#include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/p_inlines.h" +#include "cso_cache/cso_context.h" +#include "util/u_math.h" +#include "util/u_memory.h" +#include "util/u_simple_shaders.h" +#include "trace/tr_screen.h" +#include "trace/tr_context.h" + +#include "st_device.h" +#include "st_winsys.h" + + +static void +st_device_really_destroy(struct st_device *st_dev) +{ + if(st_dev->screen) + st_dev->screen->destroy(st_dev->screen); + + FREE(st_dev); +} + + +void +st_device_destroy(struct st_device *st_dev) +{ + if(!--st_dev->refcount) + st_device_really_destroy(st_dev); +} + + +static struct st_device * +st_device_create_from_st_winsys(const struct st_winsys *st_ws) +{ + struct st_device *st_dev; + + if(!st_ws->screen_create || + !st_ws->context_create) + return NULL; + + st_dev = CALLOC_STRUCT(st_device); + if(!st_dev) + return NULL; + + st_dev->refcount = 1; + st_dev->st_ws = st_ws; + + st_dev->real_screen = st_ws->screen_create(); + if(!st_dev->real_screen) { + st_device_destroy(st_dev); + return NULL; + } + + st_dev->screen = trace_screen_create(st_dev->real_screen); + if(!st_dev->screen) { + st_device_destroy(st_dev); + return NULL; + } + + return st_dev; +} + + +struct st_device * +st_device_create(boolean hardware) { + if(hardware) + return st_device_create_from_st_winsys(&st_hardpipe_winsys); + else + return st_device_create_from_st_winsys(&st_softpipe_winsys); +} + + +void +st_context_destroy(struct st_context *st_ctx) +{ + unsigned i; + + if(st_ctx) { + struct st_device *st_dev = st_ctx->st_dev; + + if(st_ctx->cso) { + cso_delete_vertex_shader(st_ctx->cso, st_ctx->vs); + cso_delete_fragment_shader(st_ctx->cso, st_ctx->fs); + + cso_destroy_context(st_ctx->cso); + } + + if(st_ctx->pipe) + st_ctx->pipe->destroy(st_ctx->pipe); + + for(i = 0; i < PIPE_MAX_SAMPLERS; ++i) + pipe_texture_reference(&st_ctx->sampler_textures[i], NULL); + pipe_texture_reference(&st_ctx->default_texture, NULL); + + FREE(st_ctx); + + if(!--st_dev->refcount) + st_device_really_destroy(st_dev); + } +} + + +struct st_context * +st_context_create(struct st_device *st_dev) +{ + struct st_context *st_ctx; + + st_ctx = CALLOC_STRUCT(st_context); + if(!st_ctx) + return NULL; + + st_ctx->st_dev = st_dev; + ++st_dev->refcount; + + st_ctx->real_pipe = st_dev->st_ws->context_create(st_dev->real_screen); + if(!st_ctx->real_pipe) { + st_context_destroy(st_ctx); + return NULL; + } + + st_ctx->pipe = trace_context_create(st_dev->screen, st_ctx->real_pipe); + if(!st_ctx->pipe) { + st_context_destroy(st_ctx); + return NULL; + } + + st_ctx->cso = cso_create_context(st_ctx->pipe); + if(!st_ctx->cso) { + st_context_destroy(st_ctx); + return NULL; + } + + /* disabled blending/masking */ + { + struct pipe_blend_state blend; + memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.colormask = PIPE_MASK_RGBA; + cso_set_blend(st_ctx->cso, &blend); + } + + /* no-op depth/stencil/alpha */ + { + struct pipe_depth_stencil_alpha_state depthstencil; + memset(&depthstencil, 0, sizeof(depthstencil)); + cso_set_depth_stencil_alpha(st_ctx->cso, &depthstencil); + } + + /* rasterizer */ + { + struct pipe_rasterizer_state rasterizer; + memset(&rasterizer, 0, sizeof(rasterizer)); + rasterizer.front_winding = PIPE_WINDING_CW; + rasterizer.cull_mode = PIPE_WINDING_NONE; + rasterizer.bypass_clipping = 1; + /*rasterizer.bypass_vs = 1;*/ + cso_set_rasterizer(st_ctx->cso, &rasterizer); + } + + /* identity viewport */ + { + struct pipe_viewport_state viewport; + viewport.scale[0] = 1.0; + viewport.scale[1] = 1.0; + viewport.scale[2] = 1.0; + viewport.scale[3] = 1.0; + viewport.translate[0] = 0.0; + viewport.translate[1] = 0.0; + viewport.translate[2] = 0.0; + viewport.translate[3] = 0.0; + cso_set_viewport(st_ctx->cso, &viewport); + } + + /* samplers */ + { + struct pipe_sampler_state sampler; + unsigned i; + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + sampler.normalized_coords = 1; + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + cso_single_sampler(st_ctx->cso, i, &sampler); + cso_single_sampler_done(st_ctx->cso); + } + + /* default textures */ + { + struct pipe_screen *screen = st_dev->screen; + struct pipe_texture templat; + struct pipe_surface *surface; + unsigned i; + + memset( &templat, 0, sizeof( templat ) ); + templat.target = PIPE_TEXTURE_2D; + templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; + templat.block.size = 4; + templat.block.width = 1; + templat.block.height = 1; + templat.width[0] = 1; + templat.height[0] = 1; + templat.depth[0] = 1; + templat.last_level = 0; + + st_ctx->default_texture = screen->texture_create( screen, &templat ); + if(st_ctx->default_texture) { + surface = screen->get_tex_surface( screen, + st_ctx->default_texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE ); + if(surface) { + uint32_t *map; + map = (uint32_t *) pipe_surface_map(surface, PIPE_BUFFER_USAGE_CPU_WRITE ); + if(map) { + *map = 0x00000000; + pipe_surface_unmap( surface ); + } + pipe_surface_reference(&surface, NULL); + } + } + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + pipe_texture_reference(&st_ctx->sampler_textures[i], st_ctx->default_texture); + + cso_set_sampler_textures(st_ctx->cso, PIPE_MAX_SAMPLERS, st_ctx->sampler_textures); + } + + /* vertex shader */ + { + struct pipe_shader_state vert_shader; + + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, + TGSI_SEMANTIC_GENERIC }; + const uint semantic_indexes[] = { 0, 0 }; + st_ctx->vs = util_make_vertex_passthrough_shader(st_ctx->pipe, + 2, + semantic_names, + semantic_indexes, + &vert_shader); + cso_set_vertex_shader_handle(st_ctx->cso, st_ctx->vs); + } + + /* fragment shader */ + { + struct pipe_shader_state frag_shader; + st_ctx->fs = util_make_fragment_passthrough_shader(st_ctx->pipe, + &frag_shader); + cso_set_fragment_shader_handle(st_ctx->cso, st_ctx->fs); + } + + return st_ctx; +} + + +void +st_buffer_destroy(struct st_buffer *st_buf) +{ + if(st_buf) { + struct pipe_screen *screen = st_buf->st_dev->screen; + pipe_buffer_reference(screen, &st_buf->buffer, NULL); + FREE(st_buf); + } +} + + +struct st_buffer * +st_buffer_create(struct st_device *st_dev, + unsigned alignment, unsigned usage, unsigned size) +{ + struct pipe_screen *screen = st_dev->screen; + struct st_buffer *st_buf; + + st_buf = CALLOC_STRUCT(st_buffer); + if(!st_buf) + return NULL; + + st_buf->st_dev = st_dev; + + st_buf->buffer = pipe_buffer_create(screen, alignment, usage, size); + if(!st_buf->buffer) { + st_buffer_destroy(st_buf); + return NULL; + } + + return st_buf; +} + diff --git a/src/gallium/state_trackers/python/st_device.h b/src/gallium/state_trackers/python/st_device.h new file mode 100644 index 0000000000..7cfe6de9f6 --- /dev/null +++ b/src/gallium/state_trackers/python/st_device.h @@ -0,0 +1,101 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_DEVICE_H_ +#define ST_DEVICE_H_ + + +#include "pipe/p_state.h" + +struct cso_context; +struct pipe_screen; +struct pipe_context; +struct st_winsys; + + +struct st_buffer { + struct st_device *st_dev; + + struct pipe_buffer *buffer; +}; + + +struct st_context { + struct st_device *st_dev; + + struct pipe_context *real_pipe; + struct pipe_context *pipe; + + struct cso_context *cso; + + void *vs; + void *fs; + + struct pipe_texture *default_texture; + struct pipe_texture *sampler_textures[PIPE_MAX_SAMPLERS]; + + unsigned num_vertex_buffers; + struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; + + unsigned num_vertex_elements; + struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS]; +}; + + +struct st_device { + const struct st_winsys *st_ws; + + struct pipe_screen *real_screen; + struct pipe_screen *screen; + + /* FIXME: we also need to refcount for textures and surfaces... */ + unsigned refcount; +}; + + +struct st_buffer * +st_buffer_create(struct st_device *st_dev, + unsigned alignment, unsigned usage, unsigned size); + +void +st_buffer_destroy(struct st_buffer *st_buf); + +struct st_context * +st_context_create(struct st_device *st_dev); + +void +st_context_destroy(struct st_context *st_ctx); + +struct st_device * +st_device_create(boolean hardware); + +void +st_device_destroy(struct st_device *st_dev); + + +#endif /* ST_DEVICE_H_ */ diff --git a/src/gallium/state_trackers/python/st_hardpipe_winsys.c b/src/gallium/state_trackers/python/st_hardpipe_winsys.c new file mode 100644 index 0000000000..8b33c70fd7 --- /dev/null +++ b/src/gallium/state_trackers/python/st_hardpipe_winsys.c @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * @file + * Stub for hardware pipe driver support. + */ + + +#include "pipe/p_compiler.h" + +#include "st_winsys.h" + + +/* XXX: Force init_gallium symbol to be linked */ +extern void init_gallium(void); +void (*force_init_gallium_linkage)(void) = &init_gallium; + + +static struct pipe_screen * +st_hardpipe_screen_create(void) +{ + return st_softpipe_winsys.screen_create(); +} + + +static struct pipe_context * +st_hardpipe_context_create(struct pipe_screen *screen) +{ + return st_softpipe_winsys.context_create(screen); +} + + +const struct st_winsys st_hardpipe_winsys = { + &st_hardpipe_screen_create, + &st_hardpipe_context_create +}; diff --git a/src/gallium/state_trackers/python/st_sample.c b/src/gallium/state_trackers/python/st_sample.c new file mode 100644 index 0000000000..7765df3c4a --- /dev/null +++ b/src/gallium/state_trackers/python/st_sample.c @@ -0,0 +1,549 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_compiler.h" +#include "pipe/p_format.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" +#include "util/u_tile.h" +#include "util/u_math.h" +#include "util/u_memory.h" + +#include "st_sample.h" + + +/** + * Use our own pseudo random generator to ensure consistent runs among + * multiple runs and platforms. + * + * @sa http://en.wikipedia.org/wiki/Linear_congruential_generator + */ +static uint32_t st_random(void) { + static uint64_t seed = UINT64_C(0xbb9a063afb0a739d); + + seed = UINT64_C(134775813) * seed + UINT64_C(1); + + return (uint16_t)(seed >> 32); +} + + +/** + * We don't want to include the patent-encumbered DXT code here, so instead + * we store several uncompressed/compressed data pairs for hardware testing + * purposes. + */ +struct dxt_data +{ + uint8_t rgba[16*4]; + uint8_t raw[16]; +}; + + +static const struct dxt_data +dxt1_rgb_data[] = { + { + { + 0x99, 0xb0, 0x8e, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x99, 0xb0, 0x8e, 0xff, + 0x99, 0xb0, 0x8e, 0xff, + 0xd6, 0xff, 0x94, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x99, 0xb0, 0x8e, 0xff, + 0xd6, 0xff, 0x94, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x99, 0xb0, 0x8e, 0xff, + 0x21, 0x14, 0x84, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x21, 0x14, 0x84, 0xff, + 0x21, 0x14, 0x84, 0xff, + 0x99, 0xb0, 0x8e, 0xff + }, + {0xf2, 0xd7, 0xb0, 0x20, 0xae, 0x2c, 0x6f, 0x97} + }, + { + { + 0xb5, 0xcf, 0x9c, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x21, 0x08, 0x6b, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x52, 0x4a, 0x7b, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0xb5, 0xcf, 0x9c, 0xff, + 0x21, 0x08, 0x6b, 0xff, + 0xb5, 0xcf, 0x9c, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x52, 0x4a, 0x7b, 0xff, + 0xb5, 0xcf, 0x9c, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x52, 0x4a, 0x7b, 0xff, + 0x83, 0x8c, 0x8b, 0xff + }, + {0x73, 0xb6, 0x4d, 0x20, 0x98, 0x2b, 0xe1, 0xb8} + }, + { + { + 0x00, 0x2c, 0xff, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0x4a, 0x5c, 0xbd, 0xff, + 0x4a, 0x5c, 0xbd, 0xff, + 0x4a, 0x5c, 0xbd, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0x94, 0x8d, 0x7b, 0xff + }, + {0xe7, 0xdd, 0x7f, 0x01, 0xf9, 0xab, 0x08, 0x80} + }, + { + { + 0x6b, 0x24, 0x21, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x8b, 0x7a, 0x99, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x6b, 0x24, 0x21, 0xff, + 0x8b, 0x7a, 0x99, 0xff, + 0x9c, 0xa6, 0xd6, 0xff, + 0x6b, 0x24, 0x21, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x8b, 0x7a, 0x99, 0xff, + 0x6b, 0x24, 0x21, 0xff, + 0x8b, 0x7a, 0x99, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x9c, 0xa6, 0xd6, 0xff + }, + {0x3a, 0x9d, 0x24, 0x69, 0xbd, 0x9f, 0xb4, 0x39} + } +}; + + +static const struct dxt_data +dxt1_rgba_data[] = { + { + { + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff, + 0x4e, 0xaa, 0x90, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff, + 0x29, 0xff, 0xff, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff, + 0x73, 0x55, 0x21, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff, + 0x4e, 0xaa, 0x90, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff + }, + {0xff, 0x2f, 0xa4, 0x72, 0xeb, 0xb2, 0xbd, 0xbe} + }, + { + { + 0xb5, 0xe3, 0x63, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x6b, 0x24, 0x84, 0xff, + 0xb5, 0xe3, 0x63, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0xb5, 0xe3, 0x63, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x6b, 0x24, 0x84, 0xff, + 0x6b, 0x24, 0x84, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0xb5, 0xe3, 0x63, 0xff, + 0x90, 0x83, 0x73, 0xff, + 0xb5, 0xe3, 0x63, 0xff + }, + {0x30, 0x69, 0x0c, 0xb7, 0x4d, 0xf7, 0x0f, 0x67} + }, + { + { + 0x00, 0x00, 0x00, 0x00, + 0xc6, 0x86, 0x8c, 0xff, + 0xc6, 0x86, 0x8c, 0xff, + 0x21, 0x65, 0x42, 0xff, + 0x21, 0x65, 0x42, 0xff, + 0x21, 0x65, 0x42, 0xff, + 0x21, 0x65, 0x42, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x21, 0x65, 0x42, 0xff, + 0xc6, 0x86, 0x8c, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0xc6, 0x86, 0x8c, 0xff + }, + {0x28, 0x23, 0x31, 0xc4, 0x17, 0xc0, 0xd3, 0x7f} + }, + { + { + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0xc6, 0xe3, 0x9c, 0xff, + 0x7b, 0x1c, 0x52, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x7b, 0x1c, 0x52, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x7b, 0x1c, 0x52, 0xff, + 0xa0, 0x7f, 0x77, 0xff, + 0xc6, 0xe3, 0x9c, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0xa0, 0x7f, 0x77, 0xff + }, + {0xea, 0x78, 0x13, 0xc7, 0x7f, 0xfc, 0x33, 0xb6} + }, +}; + + +static const struct dxt_data +dxt3_rgba_data[] = { + { + { + 0x6d, 0xc6, 0x96, 0x77, + 0x6d, 0xc6, 0x96, 0xee, + 0x6d, 0xc6, 0x96, 0xaa, + 0x8c, 0xff, 0xb5, 0x44, + 0x6d, 0xc6, 0x96, 0xff, + 0x6d, 0xc6, 0x96, 0x88, + 0x31, 0x55, 0x5a, 0x66, + 0x6d, 0xc6, 0x96, 0x99, + 0x31, 0x55, 0x5a, 0xbb, + 0x31, 0x55, 0x5a, 0x55, + 0x31, 0x55, 0x5a, 0x11, + 0x6d, 0xc6, 0x96, 0xcc, + 0x6d, 0xc6, 0x96, 0xcc, + 0x6d, 0xc6, 0x96, 0x11, + 0x31, 0x55, 0x5a, 0x44, + 0x31, 0x55, 0x5a, 0x88 + }, + {0xe7, 0x4a, 0x8f, 0x96, 0x5b, 0xc1, 0x1c, 0x84, 0xf6, 0x8f, 0xab, 0x32, 0x2a, 0x9a, 0x95, 0x5a} + }, + { + { + 0xad, 0xeb, 0x73, 0x99, + 0x97, 0xaa, 0x86, 0x66, + 0x6b, 0x28, 0xad, 0x99, + 0xad, 0xeb, 0x73, 0x99, + 0x6b, 0x28, 0xad, 0x22, + 0xad, 0xeb, 0x73, 0xff, + 0x97, 0xaa, 0x86, 0x55, + 0x6b, 0x28, 0xad, 0x55, + 0x6b, 0x28, 0xad, 0x44, + 0xad, 0xeb, 0x73, 0x33, + 0x6b, 0x28, 0xad, 0xee, + 0x6b, 0x28, 0xad, 0x99, + 0x97, 0xaa, 0x86, 0x66, + 0xad, 0xeb, 0x73, 0xbb, + 0x97, 0xaa, 0x86, 0x99, + 0xad, 0xeb, 0x73, 0xbb + }, + {0x69, 0x99, 0xf2, 0x55, 0x34, 0x9e, 0xb6, 0xb9, 0x4e, 0xaf, 0x55, 0x69, 0x18, 0x61, 0x51, 0x22} + }, + { + { + 0x63, 0xd7, 0xd6, 0x00, + 0x57, 0x62, 0x5d, 0xdd, + 0x57, 0x62, 0x5d, 0xcc, + 0x57, 0x62, 0x5d, 0xbb, + 0x52, 0x28, 0x21, 0xaa, + 0x57, 0x62, 0x5d, 0xcc, + 0x57, 0x62, 0x5d, 0xcc, + 0x57, 0x62, 0x5d, 0x66, + 0x57, 0x62, 0x5d, 0x22, + 0x57, 0x62, 0x5d, 0xdd, + 0x63, 0xd7, 0xd6, 0xee, + 0x57, 0x62, 0x5d, 0x33, + 0x63, 0xd7, 0xd6, 0x55, + 0x52, 0x28, 0x21, 0x55, + 0x57, 0x62, 0x5d, 0x11, + 0x5d, 0x9c, 0x99, 0xee + }, + {0xd0, 0xbc, 0xca, 0x6c, 0xd2, 0x3e, 0x55, 0xe1, 0xba, 0x66, 0x44, 0x51, 0xfc, 0xfd, 0xcf, 0xb4} + }, + { + { + 0x94, 0x6f, 0x60, 0x22, + 0x94, 0x6f, 0x60, 0x22, + 0xc5, 0xab, 0x76, 0x11, + 0xc5, 0xab, 0x76, 0xee, + 0x63, 0x34, 0x4a, 0xdd, + 0x63, 0x34, 0x4a, 0x33, + 0x94, 0x6f, 0x60, 0x77, + 0xf7, 0xe7, 0x8c, 0x00, + 0x94, 0x6f, 0x60, 0x33, + 0x63, 0x34, 0x4a, 0xaa, + 0x94, 0x6f, 0x60, 0x77, + 0x63, 0x34, 0x4a, 0xcc, + 0x94, 0x6f, 0x60, 0xaa, + 0xf7, 0xe7, 0x8c, 0x99, + 0x63, 0x34, 0x4a, 0x44, + 0xc5, 0xab, 0x76, 0xaa + }, + {0x22, 0xe1, 0x3d, 0x07, 0xa3, 0xc7, 0x9a, 0xa4, 0x31, 0xf7, 0xa9, 0x61, 0xaf, 0x35, 0x77, 0x93} + }, +}; + + +static const struct dxt_data +dxt5_rgba_data[] = { + { + { + 0x6d, 0xc6, 0x96, 0x74, + 0x6d, 0xc6, 0x96, 0xf8, + 0x6d, 0xc6, 0x96, 0xb6, + 0x8c, 0xff, 0xb5, 0x53, + 0x6d, 0xc6, 0x96, 0xf8, + 0x6d, 0xc6, 0x96, 0x95, + 0x31, 0x55, 0x5a, 0x53, + 0x6d, 0xc6, 0x96, 0x95, + 0x31, 0x55, 0x5a, 0xb6, + 0x31, 0x55, 0x5a, 0x53, + 0x31, 0x55, 0x5a, 0x11, + 0x6d, 0xc6, 0x96, 0xd7, + 0x6d, 0xc6, 0x96, 0xb6, + 0x6d, 0xc6, 0x96, 0x11, + 0x31, 0x55, 0x5a, 0x32, + 0x31, 0x55, 0x5a, 0x95 + }, + {0xf8, 0x11, 0xc5, 0x0c, 0x9a, 0x73, 0xb4, 0x9c, 0xf6, 0x8f, 0xab, 0x32, 0x2a, 0x9a, 0x95, 0x5a} + }, + { + { + 0xad, 0xeb, 0x73, 0xa1, + 0x97, 0xaa, 0x86, 0x65, + 0x6b, 0x28, 0xad, 0xa1, + 0xad, 0xeb, 0x73, 0xa1, + 0x6b, 0x28, 0xad, 0x2a, + 0xad, 0xeb, 0x73, 0xfb, + 0x97, 0xaa, 0x86, 0x47, + 0x6b, 0x28, 0xad, 0x65, + 0x6b, 0x28, 0xad, 0x47, + 0xad, 0xeb, 0x73, 0x47, + 0x6b, 0x28, 0xad, 0xdd, + 0x6b, 0x28, 0xad, 0xa1, + 0x97, 0xaa, 0x86, 0x65, + 0xad, 0xeb, 0x73, 0xbf, + 0x97, 0xaa, 0x86, 0xa1, + 0xad, 0xeb, 0x73, 0xbf + }, + {0xfb, 0x2a, 0x34, 0x19, 0xdc, 0xbf, 0xe8, 0x71, 0x4e, 0xaf, 0x55, 0x69, 0x18, 0x61, 0x51, 0x22} + }, + { + { + 0x63, 0xd7, 0xd6, 0x00, + 0x57, 0x62, 0x5d, 0xf5, + 0x57, 0x62, 0x5d, 0xd2, + 0x57, 0x62, 0x5d, 0xaf, + 0x52, 0x28, 0x21, 0xaf, + 0x57, 0x62, 0x5d, 0xd2, + 0x57, 0x62, 0x5d, 0xd2, + 0x57, 0x62, 0x5d, 0x69, + 0x57, 0x62, 0x5d, 0x23, + 0x57, 0x62, 0x5d, 0xd2, + 0x63, 0xd7, 0xd6, 0xf5, + 0x57, 0x62, 0x5d, 0x46, + 0x63, 0xd7, 0xd6, 0x46, + 0x52, 0x28, 0x21, 0x69, + 0x57, 0x62, 0x5d, 0x23, + 0x5d, 0x9c, 0x99, 0xf5 + }, + {0xf5, 0x00, 0x81, 0x36, 0xa9, 0x17, 0xec, 0x1e, 0xba, 0x66, 0x44, 0x51, 0xfc, 0xfd, 0xcf, 0xb4} + }, + { + { + 0x94, 0x6f, 0x60, 0x25, + 0x94, 0x6f, 0x60, 0x25, + 0xc5, 0xab, 0x76, 0x05, + 0xc5, 0xab, 0x76, 0xe8, + 0x63, 0x34, 0x4a, 0xe8, + 0x63, 0x34, 0x4a, 0x25, + 0x94, 0x6f, 0x60, 0x86, + 0xf7, 0xe7, 0x8c, 0x05, + 0x94, 0x6f, 0x60, 0x25, + 0x63, 0x34, 0x4a, 0xa7, + 0x94, 0x6f, 0x60, 0x66, + 0x63, 0x34, 0x4a, 0xc7, + 0x94, 0x6f, 0x60, 0xa7, + 0xf7, 0xe7, 0x8c, 0xa7, + 0x63, 0x34, 0x4a, 0x45, + 0xc5, 0xab, 0x76, 0xa7 + }, + {0xe8, 0x05, 0x7f, 0x80, 0x33, 0x5f, 0xb5, 0x79, 0x31, 0xf7, 0xa9, 0x61, 0xaf, 0x35, 0x77, 0x93} + }, +}; + + +static INLINE void +st_sample_dxt_pixel_block(enum pipe_format format, + const struct pipe_format_block *block, + uint8_t *raw, + float *rgba, unsigned rgba_stride, + unsigned w, unsigned h) +{ + const struct dxt_data *data; + unsigned n; + unsigned i; + unsigned x, y, ch; + + switch(format) { + case PIPE_FORMAT_DXT1_RGB: + data = dxt1_rgb_data; + n = sizeof(dxt1_rgb_data)/sizeof(dxt1_rgb_data[0]); + break; + case PIPE_FORMAT_DXT1_RGBA: + data = dxt1_rgba_data; + n = sizeof(dxt1_rgba_data)/sizeof(dxt1_rgba_data[0]); + break; + case PIPE_FORMAT_DXT3_RGBA: + data = dxt3_rgba_data; + n = sizeof(dxt3_rgba_data)/sizeof(dxt3_rgba_data[0]); + break; + case PIPE_FORMAT_DXT5_RGBA: + data = dxt5_rgba_data; + n = sizeof(dxt5_rgba_data)/sizeof(dxt5_rgba_data[0]); + break; + default: + assert(0); + } + + i = st_random() % n; + + for(y = 0; y < h; ++y) + for(x = 0; x < w; ++x) + for(ch = 0; ch < 4; ++ch) + rgba[y*rgba_stride + x*4 + ch] = (float)(data[i].rgba[y*4*4 + x*4 + ch])/255.0f; + + memcpy(raw, data[i].raw, block->size); +} + + +static INLINE void +st_sample_generic_pixel_block(enum pipe_format format, + const struct pipe_format_block *block, + uint8_t *raw, + float *rgba, unsigned rgba_stride, + unsigned w, unsigned h) +{ + unsigned i; + unsigned x, y, ch; + + for(i = 0; i < block->size; ++i) + raw[i] = (uint8_t)st_random(); + + + pipe_tile_raw_to_rgba(format, + raw, + w, h, + rgba, rgba_stride); + + if(format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV) { + for(y = 0; y < h; ++y) { + for(x = 0; x < w; ++x) { + for(ch = 0; ch < 4; ++ch) { + unsigned offset = y*rgba_stride + x*4 + ch; + rgba[offset] = CLAMP(rgba[offset], 0.0f, 1.0f); + } + } + } + } +} + + +/** + * Randomly sample pixels. + */ +void +st_sample_pixel_block(enum pipe_format format, + const struct pipe_format_block *block, + void *raw, + float *rgba, unsigned rgba_stride, + unsigned w, unsigned h) +{ + switch(format) { + case PIPE_FORMAT_DXT1_RGB: + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: + st_sample_dxt_pixel_block(format, block, raw, rgba, rgba_stride, w, h); + break; + + default: + st_sample_generic_pixel_block(format, block, raw, rgba, rgba_stride, w, h); + break; + } +} + + +void +st_sample_surface(struct pipe_surface *surface, float *rgba) +{ + const struct pipe_format_block *block = &surface->block; + unsigned rgba_stride = surface->width*4; + void *raw; + unsigned x, y; + + raw = pipe_surface_map(surface, PIPE_BUFFER_USAGE_CPU_READ); + if(!raw) + return; + + for (y = 0; y < surface->nblocksy; ++y) { + for(x = 0; x < surface->nblocksx; ++x) { + st_sample_pixel_block(surface->format, + block, + (uint8_t*)raw + y*surface->stride + x*block->size, + rgba + y*block->height*rgba_stride + x*block->width*4, + rgba_stride, + MIN2(block->width, surface->width - x*block->width), + MIN2(block->height, surface->height - y*block->height)); + } + } + + pipe_surface_unmap(surface); +} diff --git a/src/gallium/state_trackers/python/st_sample.h b/src/gallium/state_trackers/python/st_sample.h new file mode 100644 index 0000000000..ff04a12613 --- /dev/null +++ b/src/gallium/state_trackers/python/st_sample.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_SAMPLE_H_ +#define ST_SAMPLE_H_ + + +#include "pipe/p_format.h" + + +void +st_sample_pixel_block(enum pipe_format format, + const struct pipe_format_block *block, + void *raw, + float *rgba, unsigned rgba_stride, + unsigned w, unsigned h); + +void +st_sample_surface(struct pipe_surface *surface, float *rgba); + + +#endif /* ST_SAMPLE_H_ */ diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c new file mode 100644 index 0000000000..f62113a469 --- /dev/null +++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c @@ -0,0 +1,311 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * @file + * Softpipe support. + * + * @author Keith Whitwell + * @author Brian Paul + * @author Jose Fonseca + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_format.h" +#include "pipe/p_context.h" +#include "pipe/p_inlines.h" +#include "util/u_math.h" +#include "util/u_memory.h" +#include "softpipe/sp_winsys.h" +#include "st_winsys.h" + + +struct st_softpipe_buffer +{ + struct pipe_buffer base; + boolean userBuffer; /** Is this a user-space buffer? */ + void *data; + void *mapped; +}; + + +/** Cast wrapper */ +static INLINE struct st_softpipe_buffer * +st_softpipe_buffer( struct pipe_buffer *buf ) +{ + return (struct st_softpipe_buffer *)buf; +} + + +static void * +st_softpipe_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags) +{ + struct st_softpipe_buffer *st_softpipe_buf = st_softpipe_buffer(buf); + st_softpipe_buf->mapped = st_softpipe_buf->data; + return st_softpipe_buf->mapped; +} + + +static void +st_softpipe_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + struct st_softpipe_buffer *st_softpipe_buf = st_softpipe_buffer(buf); + st_softpipe_buf->mapped = NULL; +} + + +static void +st_softpipe_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + struct st_softpipe_buffer *oldBuf = st_softpipe_buffer(buf); + + if (oldBuf->data) { + if (!oldBuf->userBuffer) + align_free(oldBuf->data); + + oldBuf->data = NULL; + } + + FREE(oldBuf); +} + + +static void +st_softpipe_flush_frontbuffer(struct pipe_winsys *winsys, + struct pipe_surface *surf, + void *context_private) +{ +} + + + +static const char * +st_softpipe_get_name(struct pipe_winsys *winsys) +{ + return "softpipe"; +} + + +static struct pipe_buffer * +st_softpipe_buffer_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct st_softpipe_buffer *buffer = CALLOC_STRUCT(st_softpipe_buffer); + + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + + buffer->data = align_malloc(size, alignment); + + return &buffer->base; +} + + +/** + * Create buffer which wraps user-space data. + */ +static struct pipe_buffer * +st_softpipe_user_buffer_create(struct pipe_winsys *winsys, + void *ptr, + unsigned bytes) +{ + struct st_softpipe_buffer *buffer; + + buffer = CALLOC_STRUCT(st_softpipe_buffer); + if(!buffer) + return NULL; + + buffer->base.refcount = 1; + buffer->base.size = bytes; + buffer->userBuffer = TRUE; + buffer->data = ptr; + + return &buffer->base; +} + + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + + +static int +st_softpipe_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags, + unsigned tex_usage) +{ + const unsigned alignment = 64; + + surf->width = width; + surf->height = height; + surf->format = format; + pf_get_block(format, &surf->block); + surf->nblocksx = pf_get_nblocksx(&surf->block, width); + surf->nblocksy = pf_get_nblocksy(&surf->block, height); + surf->stride = round_up(surf->nblocksx * surf->block.size, alignment); + surf->usage = flags; + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->stride * surf->nblocksy); + if(!surf->buffer) + return -1; + + return 0; +} + + +static struct pipe_surface * +st_softpipe_surface_alloc(struct pipe_winsys *winsys) +{ + struct pipe_surface *surface = CALLOC_STRUCT(pipe_surface); + + assert(winsys); + + surface->refcount = 1; + surface->winsys = winsys; + + return surface; +} + + +static void +st_softpipe_surface_release(struct pipe_winsys *winsys, + struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + assert(!surf->texture); + surf->refcount--; + if (surf->refcount == 0) { + if (surf->buffer) + winsys_buffer_reference(winsys, &surf->buffer, NULL); + free(surf); + } + *s = NULL; +} + + +static void +st_softpipe_fence_reference(struct pipe_winsys *winsys, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) +{ +} + + +static int +st_softpipe_fence_signalled(struct pipe_winsys *winsys, + struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + + +static int +st_softpipe_fence_finish(struct pipe_winsys *winsys, + struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + + +static void +st_softpipe_destroy(struct pipe_winsys *winsys) +{ + FREE(winsys); +} + + +static struct pipe_screen * +st_softpipe_screen_create(void) +{ + static struct pipe_winsys *winsys; + struct pipe_screen *screen; + + winsys = CALLOC_STRUCT(pipe_winsys); + if(!winsys) + return NULL; + + winsys->destroy = st_softpipe_destroy; + + winsys->buffer_create = st_softpipe_buffer_create; + winsys->user_buffer_create = st_softpipe_user_buffer_create; + winsys->buffer_map = st_softpipe_buffer_map; + winsys->buffer_unmap = st_softpipe_buffer_unmap; + winsys->buffer_destroy = st_softpipe_buffer_destroy; + + winsys->surface_alloc = st_softpipe_surface_alloc; + winsys->surface_alloc_storage = st_softpipe_surface_alloc_storage; + winsys->surface_release = st_softpipe_surface_release; + + winsys->fence_reference = st_softpipe_fence_reference; + winsys->fence_signalled = st_softpipe_fence_signalled; + winsys->fence_finish = st_softpipe_fence_finish; + + winsys->flush_frontbuffer = st_softpipe_flush_frontbuffer; + winsys->get_name = st_softpipe_get_name; + + screen = softpipe_create_screen(winsys); + if(!screen) + st_softpipe_destroy(winsys); + + return screen; +} + + +static struct pipe_context * +st_softpipe_context_create(struct pipe_screen *screen) +{ + return softpipe_create(screen, screen->winsys, NULL); +} + + +const struct st_winsys st_softpipe_winsys = { + &st_softpipe_screen_create, + &st_softpipe_context_create, +}; diff --git a/src/gallium/state_trackers/python/st_winsys.h b/src/gallium/state_trackers/python/st_winsys.h new file mode 100644 index 0000000000..b8cb612d86 --- /dev/null +++ b/src/gallium/state_trackers/python/st_winsys.h @@ -0,0 +1,52 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_WINSYS_H_ +#define ST_WINSYS_H_ + + +struct pipe_screen; +struct pipe_context; + + +struct st_winsys +{ + struct pipe_screen * + (*screen_create)(void); + + struct pipe_context * + (*context_create)(struct pipe_screen *screen); +}; + + +extern const struct st_winsys st_softpipe_winsys; + +extern const struct st_winsys st_hardpipe_winsys; + + +#endif /* ST_WINSYS_H_ */ diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py new file mode 100644 index 0000000000..8477aa5fc9 --- /dev/null +++ b/src/gallium/state_trackers/python/tests/base.py @@ -0,0 +1,193 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +"""Base classes for tests. + +Loosely inspired on Python's unittest module. +""" + + +from gallium import * + + +# Enumerate all pixel formats +formats = {} +for name, value in globals().items(): + if name.startswith("PIPE_FORMAT_") and isinstance(value, int): + formats[value] = name + + +def make_image(width, height, rgba): + import Image + outimage = Image.new( + mode='RGB', + size=(width, height), + color=(0,0,0)) + outpixels = outimage.load() + for y in range(0, height): + for x in range(0, width): + offset = (y*width + x)*4 + r, g, b, a = [int(min(max(rgba[offset + ch], 0.0), 1.0)*255) for ch in range(4)] + outpixels[x, y] = r, g, b + return outimage + +def save_image(width, height, rgba, filename): + outimage = make_image(width, height, rgba) + outimage.save(filename, "PNG") + +def show_image(width, height, **rgbas): + import Tkinter as tk + from PIL import Image, ImageTk + + root = tk.Tk() + + x = 64 + y = 64 + + labels = rgbas.keys() + labels.sort() + for i in range(len(labels)): + label = labels[i] + outimage = make_image(width, height, rgbas[label]) + + if i: + window = tk.Toplevel(root) + else: + window = root + window.title(label) + image1 = ImageTk.PhotoImage(outimage) + w = image1.width() + h = image1.height() + window.geometry("%dx%d+%d+%d" % (w, h, x, y)) + panel1 = tk.Label(window, image=image1) + panel1.pack(side='top', fill='both', expand='yes') + panel1.image = image1 + x += w + 2 + + root.mainloop() + + +class TestFailure(Exception): + + pass + +class TestSkip(Exception): + + pass + + +class Test: + + def __init__(self): + pass + + def _run(self, result): + raise NotImplementedError + + def run(self): + result = TestResult() + self._run(result) + result.summary() + + +class TestCase(Test): + + def __init__(self, dev, **kargs): + Test.__init__(self) + self.dev = dev + self.__dict__.update(kargs) + + def description(self): + raise NotImplementedError + + def test(self): + raise NotImplementedError + + def _run(self, result): + result.test_start(self) + try: + self.test() + except KeyboardInterrupt: + raise + except TestSkip: + result.test_skipped(self) + except TestFailure: + result.test_failed(self) + else: + result.test_passed(self) + + +class TestSuite(Test): + + def __init__(self, tests = None): + Test.__init__(self) + if tests is None: + self.tests = [] + else: + self.tests = tests + + def add_test(self, test): + self.tests.append(test) + + def _run(self, result): + for test in self.tests: + test._run(result) + + +class TestResult: + + def __init__(self): + self.tests = 0 + self.passed = 0 + self.skipped = 0 + self.failed = 0 + self.failed_descriptions = [] + + def test_start(self, test): + self.tests += 1 + print "Running %s..." % test.description() + + def test_passed(self, test): + self.passed += 1 + print "PASS" + + def test_skipped(self, test): + self.skipped += 1 + print "SKIP" + + def test_failed(self, test): + self.failed += 1 + self.failed_descriptions.append(test.description()) + print "FAIL" + + def summary(self): + print "%u tests, %u passed, %u skipped, %u failed" % (self.tests, self.passed, self.skipped, self.failed) + for description in self.failed_descriptions: + print " %s" % description +
\ No newline at end of file diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py new file mode 100644 index 0000000000..880a61306c --- /dev/null +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -0,0 +1,397 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +import sys +from gallium import * +from base import * + + +def lods(*dims): + size = max(dims) + lods = 0 + while size: + lods += 1 + size >>= 1 + return lods + + +def minify(dims, level = 1): + return [max(dim>>level, 1) for dim in dims] + + +def tex_coords(texture, face, level, zslice): + st = [ + [0.0, 0.0], + [1.0, 0.0], + [1.0, 1.0], + [0.0, 1.0], + ] + + if texture.target == PIPE_TEXTURE_2D: + return [[s, t, 0.0] for s, t in st] + elif texture.target == PIPE_TEXTURE_3D: + depth = texture.get_depth(level) + if depth > 1: + r = float(zslice)/float(depth - 1) + else: + r = 0.0 + return [[s, t, r] for s, t in st] + elif texture.target == PIPE_TEXTURE_CUBE: + result = [] + for s, t in st: + # See http://developer.nvidia.com/object/cube_map_ogl_tutorial.html + sc = 2.0*s - 1.0 + tc = 2.0*t - 1.0 + if face == PIPE_TEX_FACE_POS_X: + rx = 1.0 + ry = -tc + rz = -sc + if face == PIPE_TEX_FACE_NEG_X: + rx = -1.0 + ry = -tc + rz = sc + if face == PIPE_TEX_FACE_POS_Y: + rx = sc + ry = 1.0 + rz = tc + if face == PIPE_TEX_FACE_NEG_Y: + rx = sc + ry = -1.0 + rz = -tc + if face == PIPE_TEX_FACE_POS_Z: + rx = sc + ry = -tc + rz = 1.0 + if face == PIPE_TEX_FACE_NEG_Z: + rx = -sc + ry = -tc + rz = -1.0 + result.append([rx, ry, rz]) + return result + +def is_pot(n): + return n & (n - 1) == 0 + + +class TextureTest(TestCase): + + def description(self): + target = { + PIPE_TEXTURE_1D: "1d", + PIPE_TEXTURE_2D: "2d", + PIPE_TEXTURE_3D: "3d", + PIPE_TEXTURE_CUBE: "cube", + }[self.target] + format = formats[self.format] + if self.target == PIPE_TEXTURE_CUBE: + face = { + PIPE_TEX_FACE_POS_X: "+x", + PIPE_TEX_FACE_NEG_X: "-x", + PIPE_TEX_FACE_POS_Y: "+y", + PIPE_TEX_FACE_NEG_Y: "-y", + PIPE_TEX_FACE_POS_Z: "+z", + PIPE_TEX_FACE_NEG_Z: "-z", + }[self.face] + else: + face = "" + return "%s %s %ux%ux%u last_level=%u face=%s level=%u zslice=%u" % ( + target, format, + self.width, self.height, self.depth, self.last_level, + face, self.level, self.zslice, + ) + + def test(self): + dev = self.dev + + target = self.target + format = self.format + width = self.width + height = self.height + depth = self.depth + last_level = self.last_level + face = self.face + level = self.level + zslice = self.zslice + + tex_usage = PIPE_TEXTURE_USAGE_SAMPLER + geom_flags = 0 + if width != height: + geom_flags |= PIPE_TEXTURE_GEOM_NON_SQUARE + if not is_pot(width) or not is_pot(height) or not is_pot(depth): + geom_flags |= PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO + + if not dev.is_format_supported(format, target, tex_usage, geom_flags): + raise TestSkip + + ctx = self.dev.context_create() + + # disabled blending/masking + blend = Blend() + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.colormask = PIPE_MASK_RGBA + ctx.set_blend(blend) + + # no-op depth/stencil/alpha + depth_stencil_alpha = DepthStencilAlpha() + ctx.set_depth_stencil_alpha(depth_stencil_alpha) + + # rasterizer + rasterizer = Rasterizer() + rasterizer.front_winding = PIPE_WINDING_CW + rasterizer.cull_mode = PIPE_WINDING_NONE + rasterizer.bypass_clipping = 1 + #rasterizer.bypass_vs = 1 + ctx.set_rasterizer(rasterizer) + + # viewport (identity, we setup vertices in wincoords) + viewport = Viewport() + scale = FloatArray(4) + scale[0] = 1.0 + scale[1] = 1.0 + scale[2] = 1.0 + scale[3] = 1.0 + viewport.scale = scale + translate = FloatArray(4) + translate[0] = 0.0 + translate[1] = 0.0 + translate[2] = 0.0 + translate[3] = 0.0 + viewport.translate = translate + ctx.set_viewport(viewport) + + # samplers + sampler = Sampler() + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.normalized_coords = 1 + sampler.min_lod = 0 + sampler.max_lod = PIPE_MAX_TEXTURE_LEVELS - 1 + ctx.set_sampler(0, sampler) + + # texture + texture = dev.texture_create( + target = target, + format = format, + width = width, + height = height, + depth = depth, + last_level = last_level, + tex_usage = tex_usage, + ) + + expected_rgba = FloatArray(height*width*4) + texture.get_surface( + usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE, + face = face, + level = level, + zslice = zslice, + ).sample_rgba(expected_rgba) + + ctx.set_sampler_texture(0, texture) + + # framebuffer + cbuf_tex = dev.texture_create( + PIPE_FORMAT_A8R8G8B8_UNORM, + width, + height, + tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET, + ) + + cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE|PIPE_BUFFER_USAGE_GPU_READ) + fb = Framebuffer() + fb.width = width + fb.height = height + fb.num_cbufs = 1 + fb.set_cbuf(0, cbuf) + ctx.set_framebuffer(fb) + ctx.surface_clear(cbuf, 0x00000000) + del fb + + # vertex shader + vs = Shader(''' + VERT1.1 + DCL IN[0], POSITION, CONSTANT + DCL IN[1], GENERIC, CONSTANT + DCL OUT[0], POSITION, CONSTANT + DCL OUT[1], GENERIC, CONSTANT + 0:MOV OUT[0], IN[0] + 1:MOV OUT[1], IN[1] + 2:END + ''') + #vs.dump() + ctx.set_vertex_shader(vs) + + # fragment shader + op = { + PIPE_TEXTURE_1D: "1D", + PIPE_TEXTURE_2D: "2D", + PIPE_TEXTURE_3D: "3D", + PIPE_TEXTURE_CUBE: "CUBE", + }[target] + fs = Shader(''' + FRAG1.1 + DCL IN[0], GENERIC[0], LINEAR + DCL OUT[0], COLOR, CONSTANT + DCL SAMP[0], CONSTANT + 0:TEX OUT[0], IN[0], SAMP[0], %s + 1:END + ''' % op) + #fs.dump() + ctx.set_fragment_shader(fs) + + nverts = 4 + nattrs = 2 + verts = FloatArray(nverts * nattrs * 4) + + x = 0 + y = 0 + w, h = minify((width, height), level) + + pos = [ + [x, y], + [x+w, y], + [x+w, y+h], + [x, y+h], + ] + + tex = tex_coords(texture, face, level, zslice) + + for i in range(0, 4): + j = 8*i + verts[j + 0] = pos[i][0] # x + verts[j + 1] = pos[i][1] # y + verts[j + 2] = 0.0 # z + verts[j + 3] = 1.0 # w + verts[j + 4] = tex[i][0] # s + verts[j + 5] = tex[i][1] # r + verts[j + 6] = tex[i][2] # q + verts[j + 7] = 1.0 + + ctx.draw_vertices(PIPE_PRIM_TRIANGLE_FAN, + nverts, + nattrs, + verts) + + ctx.flush() + + cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ) + + total = h*w + different = cbuf.compare_tile_rgba(x, y, w, h, expected_rgba, tol=4.0/256) + if different: + sys.stderr.write("%u out of %u pixels differ\n" % (different, total)) + + if float(total - different)/float(total) < 0.85: + + if 0: + rgba = FloatArray(h*w*4) + cbuf.get_tile_rgba(x, y, w, h, rgba) + show_image(w, h, Result=rgba, Expected=expected_rgba) + save_image(w, h, rgba, "result.png") + save_image(w, h, expected_rgba, "expected.png") + #sys.exit(0) + + raise TestFailure + + del ctx + + + +def main(): + dev = Device() + suite = TestSuite() + + targets = [] + targets += [PIPE_TEXTURE_2D] + targets += [PIPE_TEXTURE_CUBE] + targets += [PIPE_TEXTURE_3D] + + formats = [] + formats += [PIPE_FORMAT_A8R8G8B8_UNORM] + formats += [PIPE_FORMAT_R5G6B5_UNORM] + formats += [PIPE_FORMAT_L8_UNORM] + formats += [PIPE_FORMAT_YCBCR] + formats += [PIPE_FORMAT_DXT1_RGB] + + sizes = [64, 32, 16, 8, 4, 2, 1] + #sizes = [1020, 508, 252, 62, 30, 14, 6, 3] + #sizes = [64] + #sizes = [63] + + for target in targets: + for format in formats: + for size in sizes: + if target == PIPE_TEXTURE_CUBE: + faces = [ + PIPE_TEX_FACE_POS_X, + PIPE_TEX_FACE_NEG_X, + PIPE_TEX_FACE_POS_Y, + PIPE_TEX_FACE_NEG_Y, + PIPE_TEX_FACE_POS_Z, + PIPE_TEX_FACE_NEG_Z, + ] + #faces = [PIPE_TEX_FACE_NEG_X] + else: + faces = [0] + if target == PIPE_TEXTURE_3D: + depth = size + else: + depth = 1 + for face in faces: + levels = lods(size) + for last_level in range(levels): + for level in range(0, last_level + 1): + zslice = 0 + while zslice < depth >> level: + test = TextureTest( + dev = dev, + target = target, + format = format, + width = size, + height = size, + depth = depth, + last_level = last_level, + face = face, + level = level, + zslice = zslice, + ) + suite.add_test(test) + zslice = (zslice + 1)*2 - 1 + suite.run() + + +if __name__ == '__main__': + main() |