/************************************************************************** * * Copyright 2008-2009 VMware, Inc. * Copyright 2007 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. * **************************************************************************/ /* Vertices are just an array of floats, with all the attributes * packed. We currently assume a layout like: * * attr[0][0..3] - window position * attr[1..n][0..3] - remaining attributes. * * Attributes are assumed to be 4 floats wide but are packed so that * all the enabled attributes run contiguously. */ #include "util/u_math.h" #include "util/u_memory.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" #include "lp_context.h" #include "lp_state.h" #include "lp_quad.h" #include "lp_quad_pipe.h" #include "lp_texture.h" #include "lp_tex_sample.h" struct quad_shade_stage { struct quad_stage stage; /**< base class */ union tgsi_exec_channel ALIGN16_ATTRIB pos[NUM_CHANNELS]; struct tgsi_exec_vector ALIGN16_ATTRIB outputs[PIPE_MAX_ATTRIBS]; uint32_t ALIGN16_ATTRIB mask[NUM_CHANNELS]; }; /** cast wrapper */ static INLINE struct quad_shade_stage * quad_shade_stage(struct quad_stage *qs) { return (struct quad_shade_stage *) qs; } static void setup_pos_vector(struct quad_shade_stage *qss, const struct tgsi_interp_coef *coef, float x, float y) { uint chan; /* do X */ qss->pos[0].f[0] = x; qss->pos[0].f[1] = x + 1; qss->pos[0].f[2] = x; qss->pos[0].f[3] = x + 1; /* do Y */ qss->pos[1].f[0] = y; qss->pos[1].f[1] = y; qss->pos[1].f[2] = y + 1; qss->pos[1].f[3] = y + 1; /* do Z and W for all fragments in the quad */ for (chan = 2; chan < 4; chan++) { const float dadx = coef->dadx[chan]; const float dady = coef->dady[chan]; const float a0 = coef->a0[chan] + dadx * x + dady * y; qss->pos[chan].f[0] = a0; qss->pos[chan].f[1] = a0 + dadx; qss->pos[chan].f[2] = a0 + dady; qss->pos[chan].f[3] = a0 + dadx + dady; } } /** * Execute fragment shader for the four fragments in the quad. */ static boolean shade_quad(struct quad_stage *qs, struct quad_header *quad) { struct quad_shade_stage *qss = quad_shade_stage( qs ); struct llvmpipe_context *llvmpipe = qs->llvmpipe; void *constants; struct tgsi_sampler **samplers; unsigned chan_index; boolean z_written; /* Compute X, Y, Z, W vals for this quad */ setup_pos_vector(qss, quad->posCoef, (float)quad->input.x0, (float)quad->input.y0); constants = llvmpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; samplers = (struct tgsi_sampler **)llvmpipe->tgsi.frag_samplers_list; for (chan_index = 0; chan_index < NUM_CHANNELS; ++chan_index) qss->mask[chan_index] = ~0; /* run shader */ llvmpipe->fs->jit_function( qss->pos, quad->coef->a0, quad->coef->dadx, quad->coef->dady, constants, qss->outputs, qss->mask, samplers); for (chan_index = 0; chan_index < NUM_CHANNELS; ++chan_index) if(!qss->mask[chan_index]) quad->inout.mask &= ~(1 << chan_index); if (quad->inout.mask == 0) return FALSE; /* store outputs */ z_written = FALSE; { const ubyte *sem_name = llvmpipe->fs->info.output_semantic_name; const ubyte *sem_index = llvmpipe->fs->info.output_semantic_index; const uint n = qss->stage.llvmpipe->fs->info.num_outputs; uint i; for (i = 0; i < n; i++) { switch (sem_name[i]) { case TGSI_SEMANTIC_COLOR: { uint cbuf = sem_index[i]; memcpy(quad->output.color[cbuf], &qss->outputs[i].xyzw[0].f[0], sizeof(quad->output.color[0]) ); } break; case TGSI_SEMANTIC_POSITION: { uint j; for (j = 0; j < 4; j++) { quad->output.depth[j] = qss->outputs[0].xyzw[2].f[j]; } z_written = TRUE; } break; } } } return TRUE; } static void coverage_quad(struct quad_stage *qs, struct quad_header *quad) { struct llvmpipe_context *llvmpipe = qs->llvmpipe; uint cbuf; /* loop over colorbuffer outputs */ for (cbuf = 0; cbuf < llvmpipe->framebuffer.nr_cbufs; cbuf++) { float (*quadColor)[4] = quad->output.color[cbuf]; unsigned j; for (j = 0; j < QUAD_SIZE; j++) { assert(quad->input.coverage[j] >= 0.0); assert(quad->input.coverage[j] <= 1.0); quadColor[3][j] *= quad->input.coverage[j]; } } } static void shade_quads(struct quad_stage *qs, struct quad_header *quads[], unsigned nr) { unsigned i, pass = 0; for (i = 0; i < nr; i++) { if(!quads[i]->inout.mask) continue; if (!shade_quad(qs, quads[i])) continue; if (/*do_coverage*/ 0) coverage_quad( qs, quads[i] ); ++pass; } if (pass) qs->next->run(qs->next, quads, nr); } /** * Per-primitive (or per-begin?) setup */ static void shade_begin(struct quad_stage *qs) { qs->next->begin(qs->next); } static void shade_destroy(struct quad_stage *qs) { align_free( qs ); } struct quad_stage * lp_quad_shade_stage( struct llvmpipe_context *llvmpipe ) { struct quad_shade_stage *qss; qss = align_malloc(sizeof(struct quad_shade_stage), 16); if (!qss) return NULL; memset(qss, 0, sizeof *qss); qss->stage.llvmpipe = llvmpipe; qss->stage.begin = shade_begin; qss->stage.run = shade_quads; qss->stage.destroy = shade_destroy; return &qss->stage; }