diff options
author | Ian Romanick <ian.d.romanick@intel.com> | 2010-08-13 20:22:46 -0700 |
---|---|---|
committer | Ian Romanick <ian.d.romanick@intel.com> | 2010-08-13 20:22:46 -0700 |
commit | 443a7e4e9a360acbc3e662c098be436f180bf81d (patch) | |
tree | 71125fc394bdcbc17cd7ed74c592b9a7761541f8 /src/mesa/drivers/dri | |
parent | 2f8ee757ab324d599fcb8287789eb5f1a7890d74 (diff) | |
parent | 27041d7cb3faeaed483538a228573466363ec1c7 (diff) |
Merge branch 'master' into glsl2
Diffstat (limited to 'src/mesa/drivers/dri')
48 files changed, 1034 insertions, 743 deletions
diff --git a/src/mesa/drivers/dri/common/dri_metaops.c b/src/mesa/drivers/dri/common/dri_metaops.c index 86e59a8e51..a2f404b616 100644 --- a/src/mesa/drivers/dri/common/dri_metaops.c +++ b/src/mesa/drivers/dri/common/dri_metaops.c @@ -29,6 +29,7 @@ #include "main/arbprogram.h" #include "main/arrayobj.h" #include "main/bufferobj.h" +#include "main/context.h" #include "main/enable.h" #include "main/matrix.h" #include "main/texstate.h" diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile index 71ee753748..65fd658c04 100644 --- a/src/mesa/drivers/dri/i915/Makefile +++ b/src/mesa/drivers/dri/i915/Makefile @@ -56,7 +56,7 @@ C_SOURCES = \ ASM_SOURCES = -DRIVER_DEFINES = -I../intel -I../intel/server -DI915 \ +DRIVER_DEFINES = -I../intel -DI915 \ $(shell pkg-config libdrm --atleast-version=2.3.1 \ && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile index 831981558d..e381a5c714 100644 --- a/src/mesa/drivers/dri/i965/Makefile +++ b/src/mesa/drivers/dri/i965/Makefile @@ -106,7 +106,7 @@ C_SOURCES = \ ASM_SOURCES = -DRIVER_DEFINES = -I../intel -I../intel/server +DRIVER_DEFINES = -I../intel INCLUDES += $(INTEL_CFLAGS) DRI_LIB_DEPS += $(INTEL_LIBS) diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h index 40eece276b..af08446f2d 100644 --- a/src/mesa/drivers/dri/i965/brw_state.h +++ b/src/mesa/drivers/dri/i965/brw_state.h @@ -46,68 +46,68 @@ brw_add_validated_bo(struct brw_context *brw, drm_intel_bo *bo) } }; -const struct brw_tracked_state brw_blend_constant_color; -const struct brw_tracked_state brw_cc_unit; -const struct brw_tracked_state brw_check_fallback; -const struct brw_tracked_state brw_clip_prog; -const struct brw_tracked_state brw_clip_unit; -const struct brw_tracked_state brw_vs_constants; -const struct brw_tracked_state brw_wm_constants; -const struct brw_tracked_state brw_constant_buffer; -const struct brw_tracked_state brw_curbe_offsets; -const struct brw_tracked_state brw_invarient_state; -const struct brw_tracked_state brw_gs_prog; -const struct brw_tracked_state brw_gs_unit; -const struct brw_tracked_state brw_line_stipple; -const struct brw_tracked_state brw_aa_line_parameters; -const struct brw_tracked_state brw_pipelined_state_pointers; -const struct brw_tracked_state brw_binding_table_pointers; -const struct brw_tracked_state brw_depthbuffer; -const struct brw_tracked_state brw_polygon_stipple_offset; -const struct brw_tracked_state brw_polygon_stipple; -const struct brw_tracked_state brw_program_parameters; -const struct brw_tracked_state brw_recalculate_urb_fence; -const struct brw_tracked_state brw_sf_prog; -const struct brw_tracked_state brw_sf_unit; -const struct brw_tracked_state brw_sf_vp; -const struct brw_tracked_state brw_state_base_address; -const struct brw_tracked_state brw_urb_fence; -const struct brw_tracked_state brw_vertex_state; -const struct brw_tracked_state brw_vs_surfaces; -const struct brw_tracked_state brw_vs_prog; -const struct brw_tracked_state brw_vs_unit; -const struct brw_tracked_state brw_wm_input_sizes; -const struct brw_tracked_state brw_wm_prog; -const struct brw_tracked_state brw_wm_samplers; -const struct brw_tracked_state brw_wm_constant_surface; -const struct brw_tracked_state brw_wm_surfaces; -const struct brw_tracked_state brw_wm_binding_table; -const struct brw_tracked_state brw_wm_unit; - -const struct brw_tracked_state brw_psp_urb_cbs; - -const struct brw_tracked_state brw_pipe_control; - -const struct brw_tracked_state brw_drawing_rect; -const struct brw_tracked_state brw_indices; -const struct brw_tracked_state brw_vertices; -const struct brw_tracked_state brw_index_buffer; -const struct brw_tracked_state gen6_binding_table_pointers; -const struct brw_tracked_state gen6_blend_state; -const struct brw_tracked_state gen6_cc_state_pointers; -const struct brw_tracked_state gen6_clip_state; -const struct brw_tracked_state gen6_clip_vp; -const struct brw_tracked_state gen6_color_calc_state; -const struct brw_tracked_state gen6_depth_stencil_state; -const struct brw_tracked_state gen6_gs_state; -const struct brw_tracked_state gen6_sampler_state; -const struct brw_tracked_state gen6_scissor_state; -const struct brw_tracked_state gen6_sf_state; -const struct brw_tracked_state gen6_sf_vp; -const struct brw_tracked_state gen6_urb; -const struct brw_tracked_state gen6_viewport_state; -const struct brw_tracked_state gen6_vs_state; -const struct brw_tracked_state gen6_wm_state; +extern const struct brw_tracked_state brw_blend_constant_color; +extern const struct brw_tracked_state brw_cc_unit; +extern const struct brw_tracked_state brw_check_fallback; +extern const struct brw_tracked_state brw_clip_prog; +extern const struct brw_tracked_state brw_clip_unit; +extern const struct brw_tracked_state brw_vs_constants; +extern const struct brw_tracked_state brw_wm_constants; +extern const struct brw_tracked_state brw_constant_buffer; +extern const struct brw_tracked_state brw_curbe_offsets; +extern const struct brw_tracked_state brw_invarient_state; +extern const struct brw_tracked_state brw_gs_prog; +extern const struct brw_tracked_state brw_gs_unit; +extern const struct brw_tracked_state brw_line_stipple; +extern const struct brw_tracked_state brw_aa_line_parameters; +extern const struct brw_tracked_state brw_pipelined_state_pointers; +extern const struct brw_tracked_state brw_binding_table_pointers; +extern const struct brw_tracked_state brw_depthbuffer; +extern const struct brw_tracked_state brw_polygon_stipple_offset; +extern const struct brw_tracked_state brw_polygon_stipple; +extern const struct brw_tracked_state brw_program_parameters; +extern const struct brw_tracked_state brw_recalculate_urb_fence; +extern const struct brw_tracked_state brw_sf_prog; +extern const struct brw_tracked_state brw_sf_unit; +extern const struct brw_tracked_state brw_sf_vp; +extern const struct brw_tracked_state brw_state_base_address; +extern const struct brw_tracked_state brw_urb_fence; +extern const struct brw_tracked_state brw_vertex_state; +extern const struct brw_tracked_state brw_vs_surfaces; +extern const struct brw_tracked_state brw_vs_prog; +extern const struct brw_tracked_state brw_vs_unit; +extern const struct brw_tracked_state brw_wm_input_sizes; +extern const struct brw_tracked_state brw_wm_prog; +extern const struct brw_tracked_state brw_wm_samplers; +extern const struct brw_tracked_state brw_wm_constant_surface; +extern const struct brw_tracked_state brw_wm_surfaces; +extern const struct brw_tracked_state brw_wm_binding_table; +extern const struct brw_tracked_state brw_wm_unit; + +extern const struct brw_tracked_state brw_psp_urb_cbs; + +extern const struct brw_tracked_state brw_pipe_control; + +extern const struct brw_tracked_state brw_drawing_rect; +extern const struct brw_tracked_state brw_indices; +extern const struct brw_tracked_state brw_vertices; +extern const struct brw_tracked_state brw_index_buffer; +extern const struct brw_tracked_state gen6_binding_table_pointers; +extern const struct brw_tracked_state gen6_blend_state; +extern const struct brw_tracked_state gen6_cc_state_pointers; +extern const struct brw_tracked_state gen6_clip_state; +extern const struct brw_tracked_state gen6_clip_vp; +extern const struct brw_tracked_state gen6_color_calc_state; +extern const struct brw_tracked_state gen6_depth_stencil_state; +extern const struct brw_tracked_state gen6_gs_state; +extern const struct brw_tracked_state gen6_sampler_state; +extern const struct brw_tracked_state gen6_scissor_state; +extern const struct brw_tracked_state gen6_sf_state; +extern const struct brw_tracked_state gen6_sf_vp; +extern const struct brw_tracked_state gen6_urb; +extern const struct brw_tracked_state gen6_viewport_state; +extern const struct brw_tracked_state gen6_vs_state; +extern const struct brw_tracked_state gen6_wm_state; /*********************************************************************** * brw_state.c diff --git a/src/mesa/drivers/dri/i965/brw_util.c b/src/mesa/drivers/dri/i965/brw_util.c index 1db2a210d4..e878da3850 100644 --- a/src/mesa/drivers/dri/i965/brw_util.c +++ b/src/mesa/drivers/dri/i965/brw_util.c @@ -30,6 +30,8 @@ */ +#include <assert.h> + #include "main/mtypes.h" #include "program/prog_parameter.h" #include "brw_util.h" diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index 5f2035d79c..e19f44035f 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -29,6 +29,7 @@ #include "main/glheader.h" #include "main/context.h" #include "main/extensions.h" +#include "main/fbobject.h" #include "main/framebuffer.h" #include "main/imports.h" #include "main/points.h" @@ -39,8 +40,6 @@ #include "drivers/common/driverfuncs.h" #include "drivers/common/meta.h" -#include "i830_dri.h" - #include "intel_chipset.h" #include "intel_buffers.h" #include "intel_tex.h" @@ -420,7 +419,7 @@ intel_prepare_render(struct intel_context *intel) __DRIdrawable *drawable; drawable = driContext->driDrawablePriv; - if (drawable->dri2.stamp != driContext->dri2.draw_stamp) { + if (drawable && drawable->dri2.stamp != driContext->dri2.draw_stamp) { if (drawable->lastStamp != drawable->dri2.stamp) intel_update_renderbuffers(driContext, drawable); intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); @@ -428,7 +427,7 @@ intel_prepare_render(struct intel_context *intel) } drawable = driContext->driReadablePriv; - if (drawable->dri2.stamp != driContext->dri2.read_stamp) { + if (drawable && drawable->dri2.stamp != driContext->dri2.read_stamp) { if (drawable->lastStamp != drawable->dri2.stamp) intel_update_renderbuffers(driContext, drawable); driContext->dri2.read_stamp = drawable->dri2.stamp; @@ -613,6 +612,7 @@ intelInitContext(struct intel_context *intel, __DRIscreen *sPriv = driContextPriv->driScreenPriv; struct intel_screen *intelScreen = sPriv->private; int bo_reuse_mode; + __GLcontextModes visual; /* we can't do anything without a connection to the device */ if (intelScreen->bufmgr == NULL) @@ -624,6 +624,11 @@ intelInitContext(struct intel_context *intel, functions->Viewport = intel_viewport; } + if (mesaVis == NULL) { + memset(&visual, 0, sizeof visual); + mesaVis = &visual; + } + if (!_mesa_initialize_context_for_api(&intel->ctx, api, mesaVis, shareCtx, functions, (void *) intel)) { printf("%s: failed to init mesa context\n", __FUNCTION__); @@ -890,14 +895,21 @@ intelMakeCurrent(__DRIcontext * driContextPriv, } if (driContextPriv) { - struct gl_framebuffer *fb = driDrawPriv->driverPrivate; - struct gl_framebuffer *readFb = driReadPriv->driverPrivate; + struct gl_framebuffer *fb, *readFb; + + if (driDrawPriv == NULL && driReadPriv == NULL) { + fb = _mesa_get_incomplete_framebuffer(); + readFb = _mesa_get_incomplete_framebuffer(); + } else { + fb = driDrawPriv->driverPrivate; + readFb = driReadPriv->driverPrivate; + driContextPriv->dri2.draw_stamp = driDrawPriv->dri2.stamp - 1; + driContextPriv->dri2.read_stamp = driReadPriv->dri2.stamp - 1; + } - driContextPriv->dri2.draw_stamp = driDrawPriv->dri2.stamp - 1; - driContextPriv->dri2.read_stamp = driReadPriv->dri2.stamp - 1; intel_prepare_render(intel); _mesa_make_current(&intel->ctx, fb, readFb); - + /* We do this in intel_prepare_render() too, but intel->ctx.DrawBuffer * is NULL at that point. We can't call _mesa_makecurrent() * first, since we need the buffer size for the initial diff --git a/src/mesa/drivers/dri/intel/intel_extensions_es2.c b/src/mesa/drivers/dri/intel/intel_extensions_es2.c index baf8e13001..de34bbb2ae 100644 --- a/src/mesa/drivers/dri/intel/intel_extensions_es2.c +++ b/src/mesa/drivers/dri/intel/intel_extensions_es2.c @@ -28,7 +28,6 @@ #include "main/extensions.h" #include "intel_extensions.h" -#include "utils.h" static const char *es2_extensions[] = { /* Used by mesa internally (cf all_mesa_extensions in ../common/utils.c) */ diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c index 0e2fe893fe..02c0ffce31 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c @@ -45,6 +45,7 @@ #include "main/attrib.h" #include "main/enable.h" #include "main/viewport.h" +#include "main/context.h" #include "swrast/swrast.h" #include "intel_screen.h" diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c index fe4de18960..680d18ba29 100644 --- a/src/mesa/drivers/dri/intel/intel_regions.c +++ b/src/mesa/drivers/dri/intel/intel_regions.c @@ -155,6 +155,9 @@ intel_region_alloc_internal(struct intel_context *intel, } region = calloc(sizeof(*region), 1); + if (region == NULL) + return region; + region->cpp = cpp; region->width = width; region->height = height; @@ -189,6 +192,9 @@ intel_region_alloc(struct intel_context *intel, region = intel_region_alloc_internal(intel, cpp, width, height, aligned_pitch / cpp, buffer); + if (region == NULL) + return region; + region->tiling = tiling; return region; diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c index 224b506c05..6efb2ddc55 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_copy.c +++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c @@ -102,7 +102,7 @@ do_copy_texsubimage(struct intel_context *intel, GLcontext *ctx = &intel->ctx; const struct intel_region *src = get_teximage_source(intel, internalFormat); - if (!intelImage->mt || !src) { + if (!intelImage->mt || !src || !src->buffer) { if (INTEL_DEBUG & DEBUG_FALLBACKS) fprintf(stderr, "%s fail %p %p (0x%08x)\n", __FUNCTION__, intelImage->mt, src, internalFormat); diff --git a/src/mesa/drivers/dri/intel/intel_tex_format.c b/src/mesa/drivers/dri/intel/intel_tex_format.c index 5f813c0efa..e03b203fb4 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_format.c +++ b/src/mesa/drivers/dri/intel/intel_tex_format.c @@ -19,7 +19,6 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, GLenum format, GLenum type) { struct intel_context *intel = intel_context(ctx); - const GLboolean do32bpt = (intel->ctx.Visual.rgbBits >= 24); #if 0 printf("%s intFmt=0x%x format=0x%x type=0x%x\n", @@ -30,39 +29,28 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case 4: case GL_RGBA: case GL_COMPRESSED_RGBA: - if (format == GL_BGRA) { - if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - return MESA_FORMAT_ARGB8888; - } - else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - return MESA_FORMAT_ARGB4444; - } - else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - return MESA_FORMAT_ARGB1555; - } - } - return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; + if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) + return MESA_FORMAT_ARGB4444; + else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) + return MESA_FORMAT_ARGB1555; + else + return MESA_FORMAT_ARGB8888; case 3: case GL_RGB: case GL_COMPRESSED_RGB: - if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return MESA_FORMAT_RGB565; - } - if (do32bpt) { - if (intel->has_xrgb_textures) - return MESA_FORMAT_XRGB8888; - else - return MESA_FORMAT_ARGB8888; - } else { + if (type == GL_UNSIGNED_SHORT_5_6_5) return MESA_FORMAT_RGB565; - } + else if (intel->has_xrgb_textures) + return MESA_FORMAT_XRGB8888; + else + return MESA_FORMAT_ARGB8888; case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; + return MESA_FORMAT_ARGB8888; case GL_RGBA4: case GL_RGBA2: diff --git a/src/mesa/drivers/dri/intel/server/i830_dri.h b/src/mesa/drivers/dri/intel/server/i830_dri.h deleted file mode 100644 index def049e7a6..0000000000 --- a/src/mesa/drivers/dri/intel/server/i830_dri.h +++ /dev/null @@ -1,62 +0,0 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.6 2003/09/28 20:15:59 alanh Exp $ */ - -#ifndef _I830_DRI_H -#define _I830_DRI_H - -#include "xf86drm.h" - -#define I830_MAX_DRAWABLES 256 - -#define I830_MAJOR_VERSION 1 -#define I830_MINOR_VERSION 9 -#define I830_PATCHLEVEL 0 - -#define I830_REG_SIZE 0x80000 - -typedef struct _I830DRIRec { - drm_handle_t regs; - drmSize regsSize; - - drmSize unused1; /* backbufferSize */ - drm_handle_t unused2; /* backbuffer */ - - drmSize unused3; /* depthbufferSize */ - drm_handle_t unused4; /* depthbuffer */ - - drmSize unused5; /* rotatedSize */ - drm_handle_t unused6; /* rotatedbuffer */ - - drm_handle_t unused7; /* textures */ - int unused8; /* textureSize */ - - drm_handle_t unused9; /* agp_buffers */ - drmSize unused10; /* agp_buf_size */ - - int deviceID; - int width; - int height; - int mem; - int cpp; - int bitsPerPixel; - - int unused11[8]; /* was front/back/depth/rotated offset/pitch */ - - int unused12; /* logTextureGranularity */ - int unused13; /* textureOffset */ - - int irq; - int sarea_priv_offset; -} I830DRIRec, *I830DRIPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830ConfigPrivRec, *I830ConfigPrivPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830DRIContextRec, *I830DRIContextPtr; - - -#endif diff --git a/src/mesa/drivers/dri/intel/server/intel.h b/src/mesa/drivers/dri/intel/server/intel.h deleted file mode 100644 index 6ea72499c1..0000000000 --- a/src/mesa/drivers/dri/intel/server/intel.h +++ /dev/null @@ -1,331 +0,0 @@ -#ifndef _INTEL_H_ -#define _INTEL_H_ - -#include "xf86drm.h" /* drm_handle_t, etc */ - -/* Intel */ -#ifndef PCI_CHIP_I810 -#define PCI_CHIP_I810 0x7121 -#define PCI_CHIP_I810_DC100 0x7123 -#define PCI_CHIP_I810_E 0x7125 -#define PCI_CHIP_I815 0x1132 -#define PCI_CHIP_I810_BRIDGE 0x7120 -#define PCI_CHIP_I810_DC100_BRIDGE 0x7122 -#define PCI_CHIP_I810_E_BRIDGE 0x7124 -#define PCI_CHIP_I815_BRIDGE 0x1130 -#endif - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 - -#ifndef PCI_CHIP_I855_GM -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I855_GM_BRIDGE 0x3580 -#endif - -#ifndef PCI_CHIP_I865_G -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I865_G_BRIDGE 0x2570 -#endif - -#ifndef PCI_CHIP_I915_G -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I915_GM -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I915_GM_BRIDGE 0x2590 -#endif - -#ifndef PCI_CHIP_E7221_G -#define PCI_CHIP_E7221_G 0x258A -/* Same as I915_G_BRIDGE */ -#define PCI_CHIP_E7221_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I945_G -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_G_BRIDGE 0x2770 -#endif - -#ifndef PCI_CHIP_I945_GM -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GM_BRIDGE 0x27A0 -#endif - -#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 || \ - pI810->Chipset == PCI_CHIP_I810_DC100 || \ - pI810->Chipset == PCI_CHIP_I810_E) -#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815) -#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M) -#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G) -#define IS_I85X(pI810) (pI810->Chipset == PCI_CHIP_I855_GM) -#define IS_I852(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME)) -#define IS_I855(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME)) -#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G) - -#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G) -#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM) -#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G) -#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM) -#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810)) - -#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810)) - -#define I830_GMCH_CTRL 0x52 - -#define I830_GMCH_MEM_MASK 0x1 -#define I830_GMCH_MEM_64M 0x1 -#define I830_GMCH_MEM_128M 0 - -#define I830_GMCH_GMS_MASK 0x70 -#define I830_GMCH_GMS_DISABLED 0x00 -#define I830_GMCH_GMS_LOCAL 0x10 -#define I830_GMCH_GMS_STOLEN_512 0x20 -#define I830_GMCH_GMS_STOLEN_1024 0x30 -#define I830_GMCH_GMS_STOLEN_8192 0x40 - -#define I855_GMCH_GMS_MASK (0x7 << 4) -#define I855_GMCH_GMS_DISABLED 0x00 -#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) -#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) -#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) -#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) -#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) -#define I915G_GMCH_GMS_STOLEN_48M (0x6 << 4) -#define I915G_GMCH_GMS_STOLEN_64M (0x7 << 4) - -typedef unsigned char Bool; -#define TRUE 1 -#define FALSE 0 - -#define PIPE_NONE 0<<0 -#define PIPE_CRT 1<<0 -#define PIPE_TV 1<<1 -#define PIPE_DFP 1<<2 -#define PIPE_LFP 1<<3 -#define PIPE_CRT2 1<<4 -#define PIPE_TV2 1<<5 -#define PIPE_DFP2 1<<6 -#define PIPE_LFP2 1<<7 - -typedef struct _I830MemPool *I830MemPoolPtr; -typedef struct _I830MemRange *I830MemRangePtr; -typedef struct _I830MemRange { - long Start; - long End; - long Size; - unsigned long Physical; - unsigned long Offset; /* Offset of AGP-allocated portion */ - unsigned long Alignment; - drm_handle_t Key; - unsigned long Pitch; // add pitch - I830MemPoolPtr Pool; -} I830MemRange; - -typedef struct _I830MemPool { - I830MemRange Total; - I830MemRange Free; - I830MemRange Fixed; - I830MemRange Allocated; -} I830MemPool; - -typedef struct { - int tail_mask; - I830MemRange mem; - unsigned char *virtual_start; - int head; - int tail; - int space; -} I830RingBuffer; - -typedef struct _I830Rec { - unsigned char *MMIOBase; - unsigned char *FbBase; - int cpp; - uint32_t aper_size; - unsigned int bios_version; - - /* These are set in PreInit and never changed. */ - long FbMapSize; - long TotalVideoRam; - I830MemRange StolenMemory; /* pre-allocated memory */ - long BIOSMemorySize; /* min stolen pool size */ - int BIOSMemSizeLoc; - - /* These change according to what has been allocated. */ - long FreeMemory; - I830MemRange MemoryAperture; - I830MemPool StolenPool; - long allocatedMemory; - - /* Regions allocated either from the above pools, or from agpgart. */ - /* for single and dual head configurations */ - I830MemRange FrontBuffer; - I830MemRange FrontBuffer2; - I830MemRange Scratch; - I830MemRange Scratch2; - - I830RingBuffer *LpRing; - - I830MemRange BackBuffer; - I830MemRange DepthBuffer; - I830MemRange TexMem; - int TexGranularity; - I830MemRange ContextMem; - int drmMinor; - Bool have3DWindows; - - Bool NeedRingBufferLow; - Bool allowPageFlip; - Bool disableTiling; - - int Chipset; - unsigned long LinearAddr; - unsigned long MMIOAddr; - - drmSize registerSize; /**< \brief MMIO register map size */ - drm_handle_t registerHandle; /**< \brief MMIO register map handle */ - // IOADDRESS ioBase; - int irq; /**< \brief IRQ number */ - int GttBound; - - drm_handle_t ring_map; - unsigned int Fence[8]; - -} I830Rec; - -/* - * 12288 is set as the maximum, chosen because it is enough for - * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare. - */ -#define I830_MAXIMUM_VBIOS_MEM 12288 -#define I830_DEFAULT_VIDEOMEM_2D (MB(32) / 1024) -#define I830_DEFAULT_VIDEOMEM_3D (MB(64) / 1024) - -/* Flags for memory allocation function */ -#define FROM_ANYWHERE 0x00000000 -#define FROM_POOL_ONLY 0x00000001 -#define FROM_NEW_ONLY 0x00000002 -#define FROM_MASK 0x0000000f - -#define ALLOCATE_AT_TOP 0x00000010 -#define ALLOCATE_AT_BOTTOM 0x00000020 -#define FORCE_GAPS 0x00000040 - -#define NEED_PHYSICAL_ADDR 0x00000100 -#define ALIGN_BOTH_ENDS 0x00000200 -#define FORCE_LOW 0x00000400 - -#define ALLOC_NO_TILING 0x00001000 -#define ALLOC_INITIAL 0x00002000 - -#define ALLOCATE_DRY_RUN 0x80000000 - -/* Chipset registers for VIDEO BIOS memory RW access */ -#define _855_DRAM_RW_CONTROL 0x58 -#define _845_DRAM_RW_CONTROL 0x90 -#define DRAM_WRITE 0x33330000 - -#define KB(x) ((x) * 1024) -#define MB(x) ((x) * KB(1024)) - -#define GTT_PAGE_SIZE KB(4) -#define ROUND_TO(x, y) (((x) + (y) - 1) / (y) * (y)) -#define ROUND_DOWN_TO(x, y) ((x) / (y) * (y)) -#define ROUND_TO_PAGE(x) ROUND_TO((x), GTT_PAGE_SIZE) -#define ROUND_TO_MB(x) ROUND_TO((x), MB(1)) -#define PRIMARY_RINGBUFFER_SIZE KB(128) - - -/* Ring buffer registers, p277, overview p19 - */ -#define LP_RING 0x2030 -#define HP_RING 0x2040 - -#define RING_TAIL 0x00 -#define TAIL_ADDR 0x000FFFF8 -#define I830_TAIL_MASK 0x001FFFF8 - -#define RING_HEAD 0x04 -#define HEAD_WRAP_COUNT 0xFFE00000 -#define HEAD_WRAP_ONE 0x00200000 -#define HEAD_ADDR 0x001FFFFC -#define I830_HEAD_MASK 0x001FFFFC - -#define RING_START 0x08 -#define START_ADDR 0x03FFFFF8 -#define I830_RING_START_MASK 0xFFFFF000 - -#define RING_LEN 0x0C -#define RING_NR_PAGES 0x001FF000 -#define I830_RING_NR_PAGES 0x001FF000 -#define RING_REPORT_MASK 0x00000006 -#define RING_REPORT_64K 0x00000002 -#define RING_REPORT_128K 0x00000004 -#define RING_NO_REPORT 0x00000000 -#define RING_VALID_MASK 0x00000001 -#define RING_VALID 0x00000001 -#define RING_INVALID 0x00000000 - - -/* Fence/Tiling ranges [0..7] - */ -#define FENCE 0x2000 -#define FENCE_NR 8 - -#define I915G_FENCE_START_MASK 0x0ff00000 - -#define I830_FENCE_START_MASK 0x07f80000 - -#define FENCE_START_MASK 0x03F80000 -#define FENCE_X_MAJOR 0x00000000 -#define FENCE_Y_MAJOR 0x00001000 -#define FENCE_SIZE_MASK 0x00000700 -#define FENCE_SIZE_512K 0x00000000 -#define FENCE_SIZE_1M 0x00000100 -#define FENCE_SIZE_2M 0x00000200 -#define FENCE_SIZE_4M 0x00000300 -#define FENCE_SIZE_8M 0x00000400 -#define FENCE_SIZE_16M 0x00000500 -#define FENCE_SIZE_32M 0x00000600 -#define FENCE_SIZE_64M 0x00000700 -#define I915G_FENCE_SIZE_1M 0x00000000 -#define I915G_FENCE_SIZE_2M 0x00000100 -#define I915G_FENCE_SIZE_4M 0x00000200 -#define I915G_FENCE_SIZE_8M 0x00000300 -#define I915G_FENCE_SIZE_16M 0x00000400 -#define I915G_FENCE_SIZE_32M 0x00000500 -#define I915G_FENCE_SIZE_64M 0x00000600 -#define I915G_FENCE_SIZE_128M 0x00000700 -#define FENCE_PITCH_1 0x00000000 -#define FENCE_PITCH_2 0x00000010 -#define FENCE_PITCH_4 0x00000020 -#define FENCE_PITCH_8 0x00000030 -#define FENCE_PITCH_16 0x00000040 -#define FENCE_PITCH_32 0x00000050 -#define FENCE_PITCH_64 0x00000060 -#define FENCE_VALID 0x00000001 - -#include <mmio.h> - -# define MMIO_IN8(base, offset) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) -# define MMIO_IN32(base, offset) \ - read_MMIO_LE32(base, offset) -# define MMIO_OUT8(base, offset, val) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val) -# define MMIO_OUT32(base, offset, val) \ - *(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val) - - - /* Memory mapped register access macros */ -#define INREG8(addr) MMIO_IN8(MMIO, addr) -#define INREG(addr) MMIO_IN32(MMIO, addr) -#define OUTREG8(addr, val) MMIO_OUT8(MMIO, addr, val) -#define OUTREG(addr, val) MMIO_OUT32(MMIO, addr, val) - -#define DSPABASE 0x70184 - -#endif diff --git a/src/mesa/drivers/dri/mach64/mach64_ioctl.h b/src/mesa/drivers/dri/mach64/mach64_ioctl.h index 1ffda1932f..9145ee6e6c 100644 --- a/src/mesa/drivers/dri/mach64/mach64_ioctl.h +++ b/src/mesa/drivers/dri/mach64/mach64_ioctl.h @@ -32,6 +32,9 @@ #ifndef __MACH64_IOCTL_H__ #define __MACH64_IOCTL_H__ +#include <stdio.h> +#include <stdlib.h> + #include "mach64_dri.h" #include "mach64_reg.h" #include "mach64_lock.h" diff --git a/src/mesa/drivers/dri/nouveau/nouveau_fbo.c b/src/mesa/drivers/dri/nouveau/nouveau_fbo.c index 8be7edb150..bd1273beea 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_fbo.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_fbo.c @@ -220,7 +220,7 @@ get_tex_format(struct gl_texture_image *ti) case MESA_FORMAT_RGB565: return GL_RGB5; default: - assert(0); + return GL_NONE; } } @@ -231,7 +231,6 @@ nouveau_render_texture(GLcontext *ctx, struct gl_framebuffer *fb, struct gl_renderbuffer *rb = att->Renderbuffer; struct gl_texture_image *ti = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; - int ret; /* Allocate a renderbuffer object for the texture if we * haven't already done so. */ @@ -244,9 +243,7 @@ nouveau_render_texture(GLcontext *ctx, struct gl_framebuffer *fb, } /* Update the renderbuffer fields from the texture. */ - ret = set_renderbuffer_format(rb, get_tex_format(ti)); - assert(ret); - + set_renderbuffer_format(rb, get_tex_format(ti)); rb->Width = ti->Width; rb->Height = ti->Height; nouveau_surface_ref(&to_nouveau_teximage(ti)->surface, diff --git a/src/mesa/drivers/dri/nouveau/nouveau_texture.c b/src/mesa/drivers/dri/nouveau/nouveau_texture.c index dbf9a5cc61..442f4e899e 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_texture.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_texture.c @@ -38,6 +38,7 @@ #include "main/mipmap.h" #include "main/texfetch.h" #include "main/teximage.h" +#include "drivers/common/meta.h" static struct gl_texture_object * nouveau_texture_new(GLcontext *ctx, GLuint name, GLenum target) @@ -182,10 +183,10 @@ teximage_fits(struct gl_texture_object *t, int level) struct nouveau_surface *s = &to_nouveau_texture(t)->surfaces[level]; struct gl_texture_image *ti = t->Image[0][level]; - return ti && (t->Target == GL_TEXTURE_RECTANGLE || - (s->bo && s->width == ti->Width && - s->height == ti->Height && - s->format == ti->TexFormat)); + return ti && to_nouveau_teximage(ti)->surface.bo && + (t->Target == GL_TEXTURE_RECTANGLE || + (s->bo && s->format == ti->TexFormat && + s->width == ti->Width && s->height == ti->Height)); } static GLboolean @@ -589,6 +590,53 @@ nouveau_texture_unmap(GLcontext *ctx, struct gl_texture_object *t) } } +static void +store_mipmap(GLcontext *ctx, GLenum target, int first, int last, + struct gl_texture_object *t) +{ + struct gl_pixelstore_attrib packing = { + .BufferObj = ctx->Shared->NullBufferObj, + .Alignment = 1 + }; + GLenum format = t->Image[0][first]->TexFormat; + unsigned base_format, type, comps; + int i; + + base_format = _mesa_get_format_base_format(format); + _mesa_format_to_type_and_comps(format, &type, &comps); + + for (i = first; i <= last; i++) { + struct gl_texture_image *ti = t->Image[0][i]; + void *data = ti->Data; + + nouveau_teximage(ctx, 3, target, i, ti->InternalFormat, + ti->Width, ti->Height, ti->Depth, + ti->Border, base_format, type, data, + &packing, t, ti); + + _mesa_free_texmemory(data); + } +} + +static void +nouveau_generate_mipmap(GLcontext *ctx, GLenum target, + struct gl_texture_object *t) +{ + if (_mesa_meta_check_generate_mipmap_fallback(ctx, target, t)) { + struct gl_texture_image *base = t->Image[0][t->BaseLevel]; + + nouveau_teximage_map(ctx, base); + _mesa_generate_mipmap(ctx, target, t); + nouveau_teximage_unmap(ctx, base); + + store_mipmap(ctx, target, t->BaseLevel + 1, + get_last_level(t), t); + + } else { + _mesa_meta_GenerateMipmap(ctx, target, t); + } +} + void nouveau_texture_functions_init(struct dd_function_table *functions) { @@ -607,4 +655,5 @@ nouveau_texture_functions_init(struct dd_function_table *functions) functions->BindTexture = nouveau_bind_texture; functions->MapTexture = nouveau_texture_map; functions->UnmapTexture = nouveau_texture_unmap; + functions->GenerateMipmap = nouveau_generate_mipmap; } diff --git a/src/mesa/drivers/dri/nouveau/nv20_state_fb.c b/src/mesa/drivers/dri/nouveau/nv20_state_fb.c index 21da4f7af1..95691cad04 100644 --- a/src/mesa/drivers/dri/nouveau/nv20_state_fb.c +++ b/src/mesa/drivers/dri/nouveau/nv20_state_fb.c @@ -72,7 +72,7 @@ nv20_emit_framebuffer(GLcontext *ctx, int emit) fb->_ColorDrawBuffers[0])->surface; rt_format |= get_rt_format(s->format); - zeta_pitch = rt_pitch = s->pitch; + rt_pitch = s->pitch; nouveau_bo_markl(bctx, kelvin, NV20TCL_COLOR_OFFSET, s->bo, 0, bo_flags); @@ -88,6 +88,9 @@ nv20_emit_framebuffer(GLcontext *ctx, int emit) nouveau_bo_markl(bctx, kelvin, NV20TCL_ZETA_OFFSET, s->bo, 0, bo_flags); + } else { + rt_format |= get_rt_format(MESA_FORMAT_Z24_S8); + zeta_pitch = rt_pitch; } BEGIN_RING(chan, kelvin, NV20TCL_RT_FORMAT, 2); diff --git a/src/mesa/drivers/dri/nouveau/nv20_state_tex.c b/src/mesa/drivers/dri/nouveau/nv20_state_tex.c index e46118e4fc..2d45513bb4 100644 --- a/src/mesa/drivers/dri/nouveau/nv20_state_tex.c +++ b/src/mesa/drivers/dri/nouveau/nv20_state_tex.c @@ -194,7 +194,8 @@ nv20_emit_tex_obj(GLcontext *ctx, int emit) | nvgl_wrap_mode(t->WrapS) << 0; tx_filter = nvgl_filter_mode(t->MagFilter) << 24 - | nvgl_filter_mode(t->MinFilter) << 16; + | nvgl_filter_mode(t->MinFilter) << 16 + | 2 << 12; tx_enable = NV20TCL_TX_ENABLE_ENABLE | log2i(t->MaxAnisotropy) << 4; diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 262fe3cdde..dbf4ad477d 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -612,6 +612,8 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); + radeon_prepare_render(&rmesa->radeon); + if (rmesa->radeon.swtcl.hw_primitive != hwprim) { /* need to disable perspective-correct texturing for point sprites */ if ((hwprim & 0xf) == R200_VF_PRIM_POINT_SPRITES && ctx->Point.PointSprite) { diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c index d43e14581e..4ae0f30491 100644 --- a/src/mesa/drivers/dri/r200/r200_tcl.c +++ b/src/mesa/drivers/dri/r200/r200_tcl.c @@ -264,6 +264,8 @@ void r200TclPrimitive( GLcontext *ctx, r200ContextPtr rmesa = R200_CONTEXT(ctx); GLuint newprim = hw_prim | R200_VF_TCL_OUTPUT_VTX_ENABLE; + radeon_prepare_render(&rmesa->radeon); + if (newprim != rmesa->tcl.hw_primitive || !discrete_prim[hw_prim&0xf]) { /* need to disable perspective-correct texturing for point sprites */ diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c index a326ee4c4f..c6246a81a2 100644 --- a/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c @@ -109,13 +109,13 @@ void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c) debug_program_log(c, "before compilation"); if (c->Base.is_r500){ - r500_transform_unroll_loops(&c->Base, &loop_state); - debug_program_log(c, "after r500 transform loops"); + rc_unroll_loops(&c->Base, R500_PFS_MAX_INST); + debug_program_log(c, "after unroll loops"); } else{ - rc_transform_unroll_loops(&c->Base, &loop_state); + rc_transform_loops(&c->Base, &loop_state); debug_program_log(c, "after transform loops"); - + rc_emulate_branches(&c->Base); debug_program_log(c, "after emulate branches"); } diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c index d347b4df9c..e940fedec2 100644 --- a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c @@ -633,7 +633,7 @@ static struct rc_swizzle_caps r300_vertprog_swizzle_caps = { void r3xx_compile_vertex_program(struct r300_vertex_program_compiler* compiler) { struct emulate_loop_state loop_state; - + compiler->Base.SwizzleCaps = &r300_vertprog_swizzle_caps; addArtificialOutputs(compiler); @@ -643,13 +643,12 @@ void r3xx_compile_vertex_program(struct r300_vertex_program_compiler* compiler) /* XXX Ideally this should be done only for r3xx, but since * we don't have branching support for r5xx, we use the emulation * on all chipsets. */ - rc_transform_unroll_loops(&compiler->Base, &loop_state); - - debug_program_log(compiler, "after transform loops"); - + if (compiler->Base.is_r500){ + rc_transform_loops(&compiler->Base, &loop_state); rc_emulate_loops(&loop_state, R500_VS_MAX_ALU); } else { + rc_transform_loops(&compiler->Base, &loop_state); rc_emulate_loops(&loop_state, R300_VS_MAX_ALU); } debug_program_log(compiler, "after emulate loops"); diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c index e6b5522c5b..80a120497e 100644 --- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c @@ -30,7 +30,6 @@ #include <stdio.h> #include "../r300_reg.h" -#include "radeon_emulate_loops.h" /** * Rewrite IF instructions to use the ALU result special register. @@ -60,31 +59,6 @@ int r500_transform_IF( return 1; } -/** - * Rewrite loops to make them easier to emit. This is not a local - * transformation, because it modifies and reorders an entire block of code. - */ -void r500_transform_unroll_loops(struct radeon_compiler * c, - struct emulate_loop_state *s) -{ - int i; - - rc_transform_unroll_loops(c, s); - - for( i = s->LoopCount - 1; i >= 0; i-- ){ - struct rc_instruction * inst_continue; - if(!s->Loops[i].EndLoop){ - continue; - } - /* Insert a continue instruction at the end of the loop. This - * is required in order to emit loops correctly. */ - inst_continue = rc_insert_new_instruction(c, - s->Loops[i].EndIf->Prev); - inst_continue->U.I.Opcode = RC_OPCODE_CONTINUE; - } - -} - static int r500_swizzle_is_native(rc_opcode opcode, struct rc_src_register reg) { unsigned int relevant; diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h index 0d005a794f..34173351f8 100644 --- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h +++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h @@ -49,6 +49,4 @@ extern int r500_transform_IF( struct rc_instruction * inst, void* data); -void r500_transform_unroll_loops(struct radeon_compiler * c, - struct emulate_loop_state * s); #endif diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c index 0bd8f0a239..c3f817ad4e 100644 --- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c +++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c @@ -64,7 +64,12 @@ struct branch_info { }; struct loop_info { - int LoopStart; + int BgnLoop; + + int BranchDepth; + int * Brks; + int BrkCount; + int BrkReserved; }; struct emit_state { @@ -368,6 +373,12 @@ static void emit_flowcontrol(struct emit_state * s, struct rc_instruction * inst unsigned int newip = ++s->Code->inst_end; + /* Currently all loops use the same integer constant to intialize + * the loop variables. */ + if(!s->Code->int_constants[0]) { + s->Code->int_constants[0] = R500_FC_INT_CONST_KR(0xff); + s->Code->int_constant_count = 1; + } s->Code->inst[newip].inst0 = R500_INST_TYPE_FC | R500_INST_ALU_WAIT; switch(inst->U.I.Opcode){ @@ -378,32 +389,69 @@ static void emit_flowcontrol(struct emit_state * s, struct rc_instruction * inst s->Loops, s->CurrentLoopDepth, s->LoopsReserved, 1); loop = &s->Loops[s->CurrentLoopDepth++]; - - /* We don't emit an instruction for BGNLOOP, so we need to - * decrement the instruction counter, but first we need to - * set LoopStart to the current value of inst_end, which - * will end up being the first real instruction in the loop.*/ - loop->LoopStart = s->Code->inst_end--; + memset(loop, 0, sizeof(struct loop_info)); + loop->BranchDepth = s->CurrentBranchDepth; + loop->BgnLoop = newip; + + s->Code->inst[newip].inst2 = R500_FC_OP_LOOP + | R500_FC_JUMP_FUNC(0x00) + | R500_FC_IGNORE_UNCOVERED + ; break; - case RC_OPCODE_BRK: - /* Don't emit an instruction for BRK */ - s->Code->inst_end--; + loop = &s->Loops[s->CurrentLoopDepth - 1]; + memory_pool_array_reserve(&s->C->Pool, int, loop->Brks, + loop->BrkCount, loop->BrkReserved, 1); + + loop->Brks[loop->BrkCount++] = newip; + s->Code->inst[newip].inst2 = R500_FC_OP_BREAKLOOP + | R500_FC_JUMP_FUNC(0xff) + | R500_FC_B_OP1_DECR + | R500_FC_B_POP_CNT( + s->CurrentBranchDepth - loop->BranchDepth) + | R500_FC_IGNORE_UNCOVERED + ; break; case RC_OPCODE_CONTINUE: loop = &s->Loops[s->CurrentLoopDepth - 1]; - s->Code->inst[newip].inst2 = R500_FC_OP_JUMP | - R500_FC_JUMP_FUNC(0xff); - s->Code->inst[newip].inst3 = R500_FC_JUMP_ADDR(loop->LoopStart); + s->Code->inst[newip].inst2 = R500_FC_OP_JUMP + | R500_FC_JUMP_FUNC(0xff) + | R500_FC_B_OP1_DECR + | R500_FC_B_POP_CNT( + s->CurrentBranchDepth - loop->BranchDepth) + ; + s->Code->inst[newip].inst3 = R500_FC_JUMP_ADDR(loop->BgnLoop); break; case RC_OPCODE_ENDLOOP: - /* Don't emit an instruction for ENDLOOP */ - s->Code->inst_end--; + { + unsigned int i; + loop = &s->Loops[s->CurrentLoopDepth - 1]; + /* Emit ENDLOOP */ + s->Code->inst[newip].inst2 = R500_FC_OP_ENDLOOP + | R500_FC_JUMP_FUNC(0xff) + | R500_FC_JUMP_ANY + | R500_FC_IGNORE_UNCOVERED + ; + /* The constant integer at index 0 is used by all loops. */ + s->Code->inst[newip].inst3 = R500_FC_INT_ADDR(0) + | R500_FC_JUMP_ADDR(loop->BgnLoop + 1) + ; + + /* Set jump address and int constant for BGNLOOP */ + s->Code->inst[loop->BgnLoop].inst3 = R500_FC_INT_ADDR(0) + | R500_FC_JUMP_ADDR(newip) + ; + + /* Set jump address for the BRK instructions. */ + while(loop->BrkCount--) { + s->Code->inst[loop->Brks[loop->BrkCount]].inst3 = + R500_FC_JUMP_ADDR(newip + 1); + } s->CurrentLoopDepth--; break; - + } case RC_OPCODE_IF: if ( s->CurrentBranchDepth >= MAX_BRANCH_DEPTH_FULL) { rc_error(s->C, "Branch depth exceeds hardware limit"); @@ -442,24 +490,16 @@ static void emit_flowcontrol(struct emit_state * s, struct rc_instruction * inst } branch = &s->Branches[s->CurrentBranchDepth - 1]; - - if(inst->Prev->U.I.Opcode == RC_OPCODE_BRK){ - branch->Endif = --s->Code->inst_end; - s->Code->inst[branch->Endif].inst2 |= - R500_FC_B_OP0_DECR; - } - else{ - branch->Endif = newip; - - s->Code->inst[branch->Endif].inst2 = R500_FC_OP_JUMP - | R500_FC_A_OP_NONE /* no address stack */ - | R500_FC_JUMP_ANY /* docs says set this, but I don't understand why */ - | R500_FC_B_OP0_DECR /* decrement branch counter if stay */ - | R500_FC_B_OP1_NONE /* no branch counter if stay */ - | R500_FC_B_POP_CNT(1) + branch->Endif = newip; + + s->Code->inst[branch->Endif].inst2 = R500_FC_OP_JUMP + | R500_FC_A_OP_NONE /* no address stack */ + | R500_FC_JUMP_ANY /* docs says set this, but I don't understand why */ + | R500_FC_B_OP0_DECR /* decrement branch counter if stay */ + | R500_FC_B_OP1_NONE /* no branch counter if stay */ + | R500_FC_B_POP_CNT(1) ; - s->Code->inst[branch->Endif].inst3 = R500_FC_JUMP_ADDR(branch->Endif + 1); - } + s->Code->inst[branch->Endif].inst3 = R500_FC_JUMP_ADDR(branch->Endif + 1); s->Code->inst[branch->If].inst2 = R500_FC_OP_JUMP | R500_FC_A_OP_NONE /* no address stack */ | R500_FC_JUMP_FUNC(0x0f) /* jump if ALU result is false */ @@ -544,11 +584,9 @@ void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compi code->inst[ip].inst0 = R500_INST_TYPE_OUT | R500_INST_TEX_SEM_WAIT; } - /* Use FULL flow control mode if branches are nested deep enough. - * We don not need to enable FULL flow control mode for loops, becasue - * we aren't using the hardware loop instructions. - */ - if (s.MaxBranchDepth >= 4) { + /* Enable full flow control mode if we are using loops or have if + * statements nested at least four deep. */ + if (s.MaxBranchDepth >= 4 || s.LoopsReserved > 0) { if (code->max_temp_idx < 1) code->max_temp_idx = 1; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_code.h b/src/mesa/drivers/dri/r300/compiler/radeon_code.h index d03689763b..e14a3520dd 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_code.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_code.h @@ -221,6 +221,9 @@ struct r500_fragment_program_code { int max_temp_idx; uint32_t us_fc_ctrl; + + uint32_t int_constants[32]; + uint32_t int_constant_count; }; struct rX00_fragment_program_code { diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c index fbb4235c22..31566a937f 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c @@ -43,6 +43,12 @@ struct instruction_state { unsigned char SrcReg[3]; }; +struct loopinfo { + struct updatemask_state * Breaks; + unsigned int BreakCount; + unsigned int BreaksReserved; +}; + struct branchinfo { unsigned int HaveElse:1; @@ -59,6 +65,10 @@ struct deadcode_state { struct branchinfo * BranchStack; unsigned int BranchStackSize; unsigned int BranchStackReserved; + + struct loopinfo * LoopStack; + unsigned int LoopStackSize; + unsigned int LoopStackReserved; }; @@ -78,6 +88,22 @@ static void or_updatemasks( dst->Address = a->Address | b->Address; } +static void push_break(struct deadcode_state *s) +{ + struct loopinfo * loop = &s->LoopStack[s->LoopStackSize - 1]; + memory_pool_array_reserve(&s->C->Pool, struct updatemask_state, + loop->Breaks, loop->BreakCount, loop->BreaksReserved, 1); + + memcpy(&loop->Breaks[loop->BreakCount++], &s->R, sizeof(s->R)); +} + +static void push_loop(struct deadcode_state * s) +{ + memory_pool_array_reserve(&s->C->Pool, struct loopinfo, s->LoopStack, + s->LoopStackSize, s->LoopStackReserved, 1); + memset(&s->LoopStack[s->LoopStackSize++], 0, sizeof(struct loopinfo)); +} + static void push_branch(struct deadcode_state * s) { memory_pool_array_reserve(&s->C->Pool, struct branchinfo, s->BranchStack, @@ -233,11 +259,22 @@ void rc_dataflow_deadcode(struct radeon_compiler * c, rc_dataflow_mark_outputs_f } } } + push_loop(&s); break; } - case RC_OPCODE_CONTINUE: case RC_OPCODE_BRK: + push_break(&s); + break; case RC_OPCODE_BGNLOOP: + { + unsigned int i; + struct loopinfo * loop = &s.LoopStack[s.LoopStackSize-1]; + for(i = 0; i < loop->BreakCount; i++) { + or_updatemasks(&s.R, &s.R, &loop->Breaks[i]); + } + break; + } + case RC_OPCODE_CONTINUE: break; case RC_OPCODE_ENDIF: push_branch(&s); diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c index 131e9e7436..24c3ae57b6 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c @@ -39,7 +39,6 @@ #define DBG(...) do { if (VERBOSE) fprintf(stderr, __VA_ARGS__); } while(0) struct const_value { - struct radeon_compiler * C; struct rc_src_register * Src; float Value; @@ -78,17 +77,17 @@ static int src_reg_is_immediate(struct rc_src_register * src, c->Program.Constants.Constants[src->Index].Type==RC_CONSTANT_IMMEDIATE; } -static unsigned int loop_calc_iterations(struct emulate_loop_state *s, - struct loop_info * loop, unsigned int max_instructions) +static unsigned int loop_max_possible_iterations(struct radeon_compiler *c, + struct loop_info * loop, unsigned int prog_inst_limit) { - unsigned int total_i = rc_recompute_ips(s->C); + unsigned int total_i = rc_recompute_ips(c); unsigned int loop_i = (loop->EndLoop->IP - loop->BeginLoop->IP) - 1; /* +1 because the program already has one iteration of the loop. */ - return 1 + ((max_instructions - total_i) / (s->LoopCount * loop_i)); + return 1 + ((prog_inst_limit - total_i) / loop_i); } -static void loop_unroll(struct emulate_loop_state * s, - struct loop_info *loop, unsigned int iterations) +static void unroll_loop(struct radeon_compiler * c, struct loop_info * loop, + unsigned int iterations) { unsigned int i; struct rc_instruction * ptr; @@ -99,7 +98,7 @@ static void loop_unroll(struct emulate_loop_state * s, rc_remove_instruction(loop->EndLoop); for( i = 1; i < iterations; i++){ for(ptr = first; ptr != last->Next; ptr = ptr->Next){ - struct rc_instruction *new = rc_alloc_instruction(s->C); + struct rc_instruction *new = rc_alloc_instruction(c); memcpy(new, ptr, sizeof(struct rc_instruction)); rc_insert_instruction(append_to, new); append_to = new; @@ -115,7 +114,7 @@ static void update_const_value(void * data, struct rc_instruction * inst, if(value->Src->File != file || value->Src->Index != index || !(1 << GET_SWZ(value->Src->Swizzle, 0) & mask)){ - return; + return; } switch(inst->U.I.Opcode){ case RC_OPCODE_MOV: @@ -140,7 +139,7 @@ static void get_incr_amount(void * data, struct rc_instruction * inst, if(file != RC_FILE_TEMPORARY || count_inst->Index != index || (1 << GET_SWZ(count_inst->Swz,0) != mask)){ - return; + return; } /* Find the index of the counter register. */ opcode = rc_get_opcode_info(inst->U.I.Opcode); @@ -185,13 +184,16 @@ static void get_incr_amount(void * data, struct rc_instruction * inst, count_inst->Unknown = 1; return; } - } -static int transform_const_loop(struct emulate_loop_state * s, - struct loop_info * loop) +/** + * If prog_inst_limit is -1, then all eligible loops will be unrolled regardless + * of how many iterations they have. + */ +static int try_unroll_loop(struct radeon_compiler * c, struct loop_info * loop, + unsigned int prog_inst_limit) { - int end_loops = 1; + int end_loops; int iterations; struct count_inst count_inst; float limit_value; @@ -201,12 +203,12 @@ static int transform_const_loop(struct emulate_loop_state * s, struct rc_instruction * inst; /* Find the counter and the upper limit */ - - if(src_reg_is_immediate(&loop->Cond->U.I.SrcReg[0], s->C)){ + + if(src_reg_is_immediate(&loop->Cond->U.I.SrcReg[0], c)){ limit = &loop->Cond->U.I.SrcReg[0]; counter = &loop->Cond->U.I.SrcReg[1]; } - else if(src_reg_is_immediate(&loop->Cond->U.I.SrcReg[1], s->C)){ + else if(src_reg_is_immediate(&loop->Cond->U.I.SrcReg[1], c)){ limit = &loop->Cond->U.I.SrcReg[1]; counter = &loop->Cond->U.I.SrcReg[0]; } @@ -214,13 +216,13 @@ static int transform_const_loop(struct emulate_loop_state * s, DBG("No constant limit.\n"); return 0; } - + /* Find the initial value of the counter */ counter_value.Src = counter; counter_value.Value = 0.0f; counter_value.HasValue = 0; - counter_value.C = s->C; - for(inst = s->C->Program.Instructions.Next; inst != loop->BeginLoop; + counter_value.C = c; + for(inst = c->Program.Instructions.Next; inst != loop->BeginLoop; inst = inst->Next){ rc_for_all_writes_mask(inst, update_const_value, &counter_value); } @@ -230,11 +232,12 @@ static int transform_const_loop(struct emulate_loop_state * s, } DBG("Initial counter value is %f\n", counter_value.Value); /* Determine how the counter is modified each loop */ - count_inst.C = s->C; + count_inst.C = c; count_inst.Index = counter->Index; count_inst.Swz = counter->Swizzle; count_inst.Amount = 0.0f; count_inst.Unknown = 0; + end_loops = 1; for(inst = loop->BeginLoop->Next; end_loops > 0; inst = inst->Next){ switch(inst->U.I.Opcode){ /* XXX In the future we might want to try to unroll nested @@ -246,6 +249,16 @@ static int transform_const_loop(struct emulate_loop_state * s, loop->EndLoop = inst; end_loops--; break; + case RC_OPCODE_BRK: + /* Don't unroll loops if it has a BRK instruction + * other one used when testing the main conditional + * of the loop. */ + + /* Make sure we haven't entered a nested loops. */ + if(inst != loop->Brk && end_loops == 1) { + return 0; + } + break; /* XXX Check if the counter is modified within an if statement. */ case RC_OPCODE_IF: @@ -266,17 +279,20 @@ static int transform_const_loop(struct emulate_loop_state * s, /* Calculate the number of iterations of this loop. Keeping this * simple, since we only support increment and decrement loops. */ - limit_value = get_constant_value(s->C, limit, 0); + limit_value = get_constant_value(c, limit, 0); DBG("Limit is %f.\n", limit_value); + /* The iteration calculations are opposite of what you would expect. + * In a normal loop, if the condition is met, then loop continues, but + * with our loops, if the condition is met, the is exited. */ switch(loop->Cond->U.I.Opcode){ - case RC_OPCODE_SGT: - case RC_OPCODE_SLT: + case RC_OPCODE_SGE: + case RC_OPCODE_SLE: iterations = (int) ceilf((limit_value - counter_value.Value) / count_inst.Amount); break; - case RC_OPCODE_SLE: - case RC_OPCODE_SGE: + case RC_OPCODE_SGT: + case RC_OPCODE_SLT: iterations = (int) floorf((limit_value - counter_value.Value) / count_inst.Amount) + 1; break; @@ -284,77 +300,85 @@ static int transform_const_loop(struct emulate_loop_state * s, return 0; } + if (prog_inst_limit > 0 + && iterations > loop_max_possible_iterations(c, loop, + prog_inst_limit)) { + return 0; + } + DBG("Loop will have %d iterations.\n", iterations); - + /* Prepare loop for unrolling */ rc_remove_instruction(loop->Cond); rc_remove_instruction(loop->If); rc_remove_instruction(loop->Brk); rc_remove_instruction(loop->EndIf); - - loop_unroll(s, loop, iterations); + + unroll_loop(c, loop, iterations); loop->EndLoop = NULL; return 1; } -/** - * This function prepares a loop to be unrolled by converting it into an if - * statement. Here is an outline of the conversion process: - * BGNLOOP; -> BGNLOOP; - * <Additional conditional code> -> <Additional conditional code> - * SGE/SLT temp[0], temp[1], temp[2]; -> SLT/SGE temp[0], temp[1], temp[2]; - * IF temp[0]; -> IF temp[0]; - * BRK; -> - * ENDIF; -> <Loop Body> - * <Loop Body> -> ENDIF; - * ENDLOOP; -> ENDLOOP - * +/** + * @param c + * @param loop * @param inst A pointer to a BGNLOOP instruction. - * @return If the loop can be unrolled, a pointer to the first instruction of - * the unrolled loop. - * Otherwise, A pointer to the ENDLOOP instruction. - * Null if there is an error. + * @return 1 if all of the members of loop where set. + * @return 0 if there was an error and some members of loop are still NULL. */ -static struct rc_instruction * transform_loop(struct emulate_loop_state * s, +static int build_loop_info(struct radeon_compiler * c, struct loop_info * loop, struct rc_instruction * inst) { - struct loop_info *loop; struct rc_instruction * ptr; - memory_pool_array_reserve(&s->C->Pool, struct loop_info, - s->Loops, s->LoopCount, s->LoopReserved, 1); - - loop = &s->Loops[s->LoopCount++]; - memset(loop, 0, sizeof(struct loop_info)); if(inst->U.I.Opcode != RC_OPCODE_BGNLOOP){ - rc_error(s->C, "expected BGNLOOP\n", __FUNCTION__); - return NULL; + rc_error(c, "%s: expected BGNLOOP", __FUNCTION__); + return 0; } + + memset(loop, 0, sizeof(struct loop_info)); + loop->BeginLoop = inst; - for(ptr = loop->BeginLoop->Next; !loop->EndLoop; ptr = ptr->Next){ + for(ptr = loop->BeginLoop->Next; !loop->EndLoop; ptr = ptr->Next) { + + if (ptr == &c->Program.Instructions) { + rc_error(c, "%s: BGNLOOP without an ENDLOOOP.\n", + __FUNCTION__); + return 0; + } + switch(ptr->U.I.Opcode){ case RC_OPCODE_BGNLOOP: - /* Nested loop */ - ptr = transform_loop(s, ptr); - if(!ptr){ - return NULL; + { + /* Nested loop, skip ahead to the end. */ + unsigned int loop_depth = 1; + for(ptr = ptr->Next; ptr != &c->Program.Instructions; + ptr = ptr->Next){ + if (ptr->U.I.Opcode == RC_OPCODE_BGNLOOP) { + loop_depth++; + } else if (ptr->U.I.Opcode == RC_OPCODE_ENDLOOP) { + if (!--loop_depth) { + break; + } + } + } + if (ptr == &c->Program.Instructions) { + rc_error(c, "%s: BGNLOOP without an ENDLOOOP\n", + __FUNCTION__); + return 0; } break; + } case RC_OPCODE_BRK: - loop->Brk = ptr; - if(ptr->Next->U.I.Opcode != RC_OPCODE_ENDIF){ - rc_error(s->C, - "%s: expected ENDIF\n",__FUNCTION__); - return NULL; - } - loop->EndIf = ptr->Next; - if(ptr->Prev->U.I.Opcode != RC_OPCODE_IF){ - rc_error(s->C, - "%s: expected IF\n", __FUNCTION__); - return NULL; + if(ptr->Next->U.I.Opcode != RC_OPCODE_ENDIF + || ptr->Prev->U.I.Opcode != RC_OPCODE_IF + || loop->Brk){ + continue; } + loop->Brk = ptr; loop->If = ptr->Prev; + loop->EndIf = ptr->Next; switch(loop->If->Prev->U.I.Opcode){ case RC_OPCODE_SLT: case RC_OPCODE_SGE: @@ -364,18 +388,61 @@ static struct rc_instruction * transform_loop(struct emulate_loop_state * s, case RC_OPCODE_SNE: break; default: - rc_error(s->C, "%s expected conditional\n", + rc_error(c, "%s: expected conditional", __FUNCTION__); - return NULL; + return 0; } loop->Cond = loop->If->Prev; - ptr = loop->EndIf; break; + case RC_OPCODE_ENDLOOP: loop->EndLoop = ptr; break; } } + + if (loop->BeginLoop && loop->Brk && loop->If && loop->EndIf + && loop->Cond && loop->EndLoop) { + return 1; + } + return 0; +} + +/** + * This function prepares a loop to be unrolled by converting it into an if + * statement. Here is an outline of the conversion process: + * BGNLOOP; -> BGNLOOP; + * <Additional conditional code> -> <Additional conditional code> + * SGE/SLT temp[0], temp[1], temp[2]; -> SLT/SGE temp[0], temp[1], temp[2]; + * IF temp[0]; -> IF temp[0]; + * BRK; -> + * ENDIF; -> <Loop Body> + * <Loop Body> -> ENDIF; + * ENDLOOP; -> ENDLOOP + * + * @param inst A pointer to a BGNLOOP instruction. + * @return If the loop can be unrolled, a pointer to the first instruction of + * the unrolled loop. + * Otherwise, A pointer to the ENDLOOP instruction. + * Null if there is an error. + */ +static struct rc_instruction * transform_loop(struct emulate_loop_state * s, + struct rc_instruction * inst) +{ + struct loop_info * loop; + + memory_pool_array_reserve(&s->C->Pool, struct loop_info, + s->Loops, s->LoopCount, s->LoopReserved, 1); + + loop = &s->Loops[s->LoopCount++]; + + if (!build_loop_info(s->C, loop, inst)) + return NULL; + + if(try_unroll_loop(s->C, loop, -1)){ + return loop->BeginLoop->Next; + } + /* Reverse the conditional instruction */ switch(loop->Cond->U.I.Opcode){ case RC_OPCODE_SGE: @@ -400,24 +467,19 @@ static struct rc_instruction * transform_loop(struct emulate_loop_state * s, rc_error(s->C, "loop->Cond is not a conditional.\n"); return NULL; } - - /* Check if the number of loops is known at compile time. */ - if(transform_const_loop(s, loop)){ - return loop->BeginLoop->Next; - } - /* Prepare the loop to be unrolled */ + /* Prepare the loop to be emulated */ rc_remove_instruction(loop->Brk); rc_remove_instruction(loop->EndIf); rc_insert_instruction(loop->EndLoop->Prev, loop->EndIf); return loop->EndLoop; } -void rc_transform_unroll_loops(struct radeon_compiler *c, - struct emulate_loop_state * s) +void rc_transform_loops(struct radeon_compiler *c, + struct emulate_loop_state * s) { struct rc_instruction * ptr; - + memset(s, 0, sizeof(struct emulate_loop_state)); s->C = c; ptr = s->C->Program.Instructions.Next; @@ -433,8 +495,23 @@ void rc_transform_unroll_loops(struct radeon_compiler *c, } } -void rc_emulate_loops(struct emulate_loop_state *s, - unsigned int max_instructions) +void rc_unroll_loops(struct radeon_compiler *c, int prog_inst_limit) +{ + struct rc_instruction * inst; + struct loop_info loop; + + for(inst = c->Program.Instructions.Next; + inst != &c->Program.Instructions; inst = inst->Next) { + + if (inst->U.I.Opcode == RC_OPCODE_BGNLOOP) { + if (build_loop_info(c, &loop, inst)) { + try_unroll_loop(c, &loop, prog_inst_limit); + } + } + } +} + +void rc_emulate_loops(struct emulate_loop_state *s, int prog_inst_limit) { int i; /* Iterate backwards of the list of loops so that loops that nested @@ -444,8 +521,8 @@ void rc_emulate_loops(struct emulate_loop_state *s, if(!s->Loops[i].EndLoop){ continue; } - unsigned int iterations = loop_calc_iterations(s, &s->Loops[i], - max_instructions); - loop_unroll(s, &s->Loops[i], iterations); + unsigned int iterations = loop_max_possible_iterations( + s->C, &s->Loops[i], prog_inst_limit); + unroll_loop(s->C, &s->Loops[i], iterations); } } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.h b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.h index 7748813c4e..86d91ef14b 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.h @@ -23,10 +23,11 @@ struct emulate_loop_state { unsigned int LoopReserved; }; -void rc_transform_unroll_loops(struct radeon_compiler *c, - struct emulate_loop_state * s); +void rc_transform_loops(struct radeon_compiler *c, + struct emulate_loop_state * s); -void rc_emulate_loops(struct emulate_loop_state *s, - unsigned int max_instructions); +void rc_unroll_loops(struct radeon_compiler * c, int prog_inst_limit); + +void rc_emulate_loops(struct emulate_loop_state * s, int prog_inst_limit); #endif /* RADEON_EMULATE_LOOPS_H */ diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c index eca0651536..7a3f35950a 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c @@ -164,7 +164,8 @@ static void peephole(struct radeon_compiler * c, struct rc_instruction * inst_mo inst = inst->Next) { /* XXX In the future we might be able to make the optimizer * smart enough to handle loops. */ - if(inst->U.I.Opcode == RC_OPCODE_BGNLOOP){ + if(inst->U.I.Opcode == RC_OPCODE_BGNLOOP + || inst->U.I.Opcode == RC_OPCODE_ENDLOOP){ return; } rc_for_all_reads_mask(inst, peephole_scan_read, &s); diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c index 3cc2897293..857aae5514 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c @@ -988,17 +988,22 @@ void radeonTransformKILP(struct radeon_compiler * c) for (inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) { - if (inst->U.I.Opcode != RC_OPCODE_KILP - || inst->Prev->U.I.Opcode != RC_OPCODE_IF - || inst->Next->U.I.Opcode != RC_OPCODE_ENDIF) { + if (inst->U.I.Opcode != RC_OPCODE_KILP) continue; - } + inst->U.I.Opcode = RC_OPCODE_KIL; - inst->U.I.SrcReg[0] = negate(absolute(inst->Prev->U.I.SrcReg[0])); - /* Remove IF */ - rc_remove_instruction(inst->Prev); - /* Remove ENDIF */ - rc_remove_instruction(inst->Next); + if (inst->Prev->U.I.Opcode != RC_OPCODE_IF + || inst->Next->U.I.Opcode != RC_OPCODE_ENDIF) { + inst->U.I.SrcReg[0] = negate(builtin_one); + } else { + + inst->U.I.SrcReg[0] = + negate(absolute(inst->Prev->U.I.SrcReg[0])); + /* Remove IF */ + rc_remove_instruction(inst->Prev); + /* Remove ENDIF */ + rc_remove_instruction(inst->Next); + } } } diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index bb8f91491f..cf89ab7ec3 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -327,6 +327,8 @@ void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim) BATCH_LOCALS(&rmesa->radeon); int type, num_verts; + radeon_prepare_render(&rmesa->radeon); + type = r300PrimitiveType(rmesa, prim); num_verts = r300NumVerts(rmesa, end - start, prim); diff --git a/src/mesa/drivers/dri/r600/r600_blit.c b/src/mesa/drivers/dri/r600/r600_blit.c index 172f85eb26..619678214f 100644 --- a/src/mesa/drivers/dri/r600/r600_blit.c +++ b/src/mesa/drivers/dri/r600/r600_blit.c @@ -390,13 +390,20 @@ set_render_target(context_t *context, struct radeon_bo *bo, gl_format mesa_forma 0, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT, 0); END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(12); + BEGIN_BATCH_NO_AUTOSTATE(9); R600_OUT_BATCH_REGVAL(CB_COLOR0_SIZE + (4 * id), cb_color0_size); R600_OUT_BATCH_REGVAL(CB_COLOR0_VIEW + (4 * id), cb_color0_view); - R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), cb_color0_info); R600_OUT_BATCH_REGVAL(CB_COLOR0_MASK + (4 * id), 0); END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), cb_color0_info); + R600_OUT_BATCH_RELOC(0, + bo, + 0, + 0, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT, 0); + END_BATCH(); + COMMIT_BATCH(); } @@ -1447,7 +1454,7 @@ set_default_state(context_t *context) SETbit(sq_dyn_gpr_cntl_ps_flush_req, VS_PC_LIMIT_ENABLE_bit); } - BEGIN_BATCH_NO_AUTOSTATE(117); + BEGIN_BATCH_NO_AUTOSTATE(114); R600_OUT_BATCH_REGSEQ(SQ_CONFIG, 6); R600_OUT_BATCH(sq_config); R600_OUT_BATCH(sq_gpr_resource_mgmt_1); @@ -1477,7 +1484,6 @@ set_default_state(context_t *context) (CLRCMP_SEL_SRC << CLRCMP_FCN_SEL_shift)); R600_OUT_BATCH_REGVAL(SQ_VTX_BASE_VTX_LOC, 0); R600_OUT_BATCH_REGVAL(SQ_VTX_START_INST_LOC, 0); - R600_OUT_BATCH_REGVAL(DB_DEPTH_INFO, 0); R600_OUT_BATCH_REGVAL(DB_DEPTH_CONTROL, 0); R600_OUT_BATCH_REGVAL(CB_SHADER_MASK, (OUTPUT0_ENABLE_mask)); R600_OUT_BATCH_REGVAL(CB_TARGET_MASK, (TARGET0_ENABLE_mask)); @@ -1607,7 +1613,7 @@ unsigned r600_blit(GLcontext *ctx, /* Flush is needed to make sure that source buffer has correct data */ radeonFlush(ctx); - rcommonEnsureCmdBufSpace(&context->radeon, 304, __FUNCTION__); + rcommonEnsureCmdBufSpace(&context->radeon, 305, __FUNCTION__); /* load shaders */ load_shaders(context->radeon.glCtx); @@ -1616,7 +1622,7 @@ unsigned r600_blit(GLcontext *ctx, return GL_FALSE; /* set clear state */ - /* 117 */ + /* 114 */ set_default_state(context); /* shaders */ @@ -1632,7 +1638,7 @@ unsigned r600_blit(GLcontext *ctx, set_tex_sampler(context); /* dst */ - /* 27 */ + /* 31 */ set_render_target(context, dst_bo, dst_mesaformat, dst_pitch, dst_width, dst_height, dst_offset); /* scissors */ diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index 99a33df4fc..9c954cbf70 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -275,7 +275,10 @@ GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size) case 2: format = FMT_8_8; break; case 3: - format = FMT_8_8_8; break; + /* for some (small/unaligned) strides using 4 comps works + * better, probably same as GL_SHORT below + * test piglit/draw-vertices */ + format = FMT_8_8_8_8; break; case 4: format = FMT_8_8_8_8; break; default: @@ -2872,25 +2875,92 @@ GLboolean assemble_CMP(r700_AssemblerBase *pAsm) GLboolean assemble_TRIG(r700_AssemblerBase *pAsm, BITS opcode) { + /* + * r600 - trunc to -PI..PI range + * r700 - normalize by dividing by 2PI + * see fdo bug 27901 + */ + int tmp; checkop1(pAsm); tmp = gethelpr(pAsm); - pAsm->D.dst.opcode = SQ_OP2_INST_MUL; + pAsm->D.dst.opcode = SQ_OP3_INST_MULADD; + pAsm->D.dst.op3 = 1; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); pAsm->D.dst.rtype = DST_REG_TEMPORARY; pAsm->D.dst.reg = tmp; - pAsm->D.dst.writex = 1; assemble_src(pAsm, 0, -1); pAsm->S[1].src.rtype = SRC_REC_LITERAL; setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X); + + pAsm->S[2].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_Y); + pAsm->D2.dst2.literal_slots = 1; pAsm->C[0].f = 1/(3.1415926535 * 2); - pAsm->C[1].f = 0.0F; - next_ins(pAsm); + pAsm->C[1].f = 0.5f; + + if ( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = SQ_OP2_INST_FRACT; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + + if(( GL_FALSE == next_ins(pAsm) )) + { + return GL_FALSE; + } + pAsm->D.dst.opcode = SQ_OP3_INST_MULADD; + pAsm->D.dst.op3 = 1; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X); + + pAsm->S[2].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_Y); + + pAsm->D2.dst2.literal_slots = 1; + + if (pAsm->bR6xx) + { + pAsm->C[0].f = 3.1415926535897f * 2.0f; + pAsm->C[1].f = -3.1415926535897f; + } + else + { + pAsm->C[0].f = 1.0f; + pAsm->C[1].f = -0.5f; + } + + if(( GL_FALSE == next_ins(pAsm) )) + { + return GL_FALSE; + } pAsm->D.dst.opcode = opcode; pAsm->D.dst.math = 1; @@ -4030,22 +4100,79 @@ GLboolean assemble_SCS(r700_AssemblerBase *pAsm) checkop1(pAsm); tmp = gethelpr(pAsm); - /* tmp.x = src /2*PI */ - pAsm->D.dst.opcode = SQ_OP2_INST_MUL; + + pAsm->D.dst.opcode = SQ_OP3_INST_MULADD; + pAsm->D.dst.op3 = 1; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); pAsm->D.dst.rtype = DST_REG_TEMPORARY; pAsm->D.dst.reg = tmp; - pAsm->D.dst.writex = 1; assemble_src(pAsm, 0, -1); pAsm->S[1].src.rtype = SRC_REC_LITERAL; setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X); + + pAsm->S[2].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_Y); + pAsm->D2.dst2.literal_slots = 1; pAsm->C[0].f = 1/(3.1415926535 * 2); - pAsm->C[1].f = 0.0F; + pAsm->C[1].f = 0.5F; - next_ins(pAsm); + if ( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = SQ_OP2_INST_FRACT; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + + if(( GL_FALSE == next_ins(pAsm) )) + { + return GL_FALSE; + } + pAsm->D.dst.opcode = SQ_OP3_INST_MULADD; + pAsm->D.dst.op3 = 1; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X); + + pAsm->S[2].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_Y); + + pAsm->D2.dst2.literal_slots = 1; + + if(pAsm->bR6xx) { + pAsm->C[0].f = 3.1415926535897f * 2.0f; + pAsm->C[1].f = -3.1415926535897f; + } else { + pAsm->C[0].f = 1.0f; + pAsm->C[1].f = -0.5f; + } + + if(( GL_FALSE == next_ins(pAsm) )) + { + return GL_FALSE; + } // COS dst.x, a.x pAsm->D.dst.opcode = SQ_OP2_INST_COS; @@ -6473,7 +6600,7 @@ GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, * results are undefined anyway */ if(export_count == 0) { - Process_Export(pR700AsmCode, SQ_EXPORT_PIXEL, 0, 1, 0, GL_FALSE); + Process_Export(pR700AsmCode, SQ_EXPORT_PIXEL, 0, 1, pR700AsmCode->starting_export_register_number, GL_FALSE); } if(pR700AsmCode->cf_last_export_ptr != NULL) diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index cefda3ac4b..1e955b93b2 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -617,18 +617,25 @@ static void r700SendDepthTargetState(GLcontext *ctx, struct radeon_state_atom *a r700SetDepthTarget(context); - BEGIN_BATCH_NO_AUTOSTATE(8 + 2); + BEGIN_BATCH_NO_AUTOSTATE(7 + 2); R600_OUT_BATCH_REGSEQ(DB_DEPTH_SIZE, 2); R600_OUT_BATCH(r700->DB_DEPTH_SIZE.u32All); R600_OUT_BATCH(r700->DB_DEPTH_VIEW.u32All); - R600_OUT_BATCH_REGSEQ(DB_DEPTH_BASE, 2); + R600_OUT_BATCH_REGSEQ(DB_DEPTH_BASE, 1); R600_OUT_BATCH(r700->DB_DEPTH_BASE.u32All); - R600_OUT_BATCH(r700->DB_DEPTH_INFO.u32All); R600_OUT_BATCH_RELOC(r700->DB_DEPTH_BASE.u32All, rrb->bo, r700->DB_DEPTH_BASE.u32All, 0, RADEON_GEM_DOMAIN_VRAM, 0); END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGSEQ(DB_DEPTH_INFO, 1); + R600_OUT_BATCH(r700->DB_DEPTH_INFO.u32All); + R600_OUT_BATCH_RELOC(r700->DB_DEPTH_INFO.u32All, + rrb->bo, + r700->DB_DEPTH_INFO.u32All, + 0, RADEON_GEM_DOMAIN_VRAM, 0); + END_BATCH(); if ((context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) && (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)) { @@ -687,27 +694,35 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom * BEGIN_BATCH_NO_AUTOSTATE(3 + 2); R600_OUT_BATCH_REGSEQ(CB_COLOR0_TILE + (4 * id), 1); R600_OUT_BATCH(r700->render_target[id].CB_COLOR0_TILE.u32All); - R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_BASE.u32All, + R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_TILE.u32All, rrb->bo, - r700->render_target[id].CB_COLOR0_BASE.u32All, + r700->render_target[id].CB_COLOR0_TILE.u32All, 0, RADEON_GEM_DOMAIN_VRAM, 0); END_BATCH(); BEGIN_BATCH_NO_AUTOSTATE(3 + 2); R600_OUT_BATCH_REGSEQ(CB_COLOR0_FRAG + (4 * id), 1); R600_OUT_BATCH(r700->render_target[id].CB_COLOR0_FRAG.u32All); - R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_BASE.u32All, + R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_FRAG.u32All, rrb->bo, - r700->render_target[id].CB_COLOR0_BASE.u32All, + r700->render_target[id].CB_COLOR0_FRAG.u32All, 0, RADEON_GEM_DOMAIN_VRAM, 0); END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(12); + BEGIN_BATCH_NO_AUTOSTATE(9); R600_OUT_BATCH_REGVAL(CB_COLOR0_SIZE + (4 * id), r700->render_target[id].CB_COLOR0_SIZE.u32All); R600_OUT_BATCH_REGVAL(CB_COLOR0_VIEW + (4 * id), r700->render_target[id].CB_COLOR0_VIEW.u32All); - R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), r700->render_target[id].CB_COLOR0_INFO.u32All); R600_OUT_BATCH_REGVAL(CB_COLOR0_MASK + (4 * id), r700->render_target[id].CB_COLOR0_MASK.u32All); END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), r700->render_target[id].CB_COLOR0_INFO.u32All); + R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_INFO.u32All, + rrb->bo, + r700->render_target[id].CB_COLOR0_INFO.u32All, + 0, RADEON_GEM_DOMAIN_VRAM, 0); + + END_BATCH(); + COMMIT_BATCH(); } @@ -1567,7 +1582,7 @@ void r600InitAtoms(context_t *context) ALLOC_STATE(sq, always, 34, r700SendSQConfig); ALLOC_STATE(db, always, 17, r700SendDBState); ALLOC_STATE(stencil, always, 4, r700SendStencilState); - ALLOC_STATE(db_target, always, 12, r700SendDepthTargetState); + ALLOC_STATE(db_target, always, 16, r700SendDepthTargetState); ALLOC_STATE(sc, always, 15, r700SendSCState); ALLOC_STATE(scissor, always, 22, r700SendScissorState); ALLOC_STATE(aa, always, 12, r700SendAAState); @@ -1578,7 +1593,7 @@ void r600InitAtoms(context_t *context) ALLOC_STATE(poly, always, 10, r700SendPolyState); ALLOC_STATE(cb, cb, 18, r700SendCBState); ALLOC_STATE(clrcmp, always, 6, r700SendCBCLRCMPState); - ALLOC_STATE(cb_target, always, 29, r700SendRenderTargetState); + ALLOC_STATE(cb_target, always, 31, r700SendRenderTargetState); ALLOC_STATE(blnd, blnd, (6 + (R700_MAX_RENDER_TARGETS * 3)), r700SendCBBlendState); ALLOC_STATE(blnd_clr, always, 6, r700SendCBBlendColorState); ALLOC_STATE(sx, always, 9, r700SendSXState); diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index 09c48565b6..d1008f28b9 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -48,6 +48,7 @@ static GLboolean r700ClearFast(context_t *context, GLbitfield mask) void r700Clear(GLcontext * ctx, GLbitfield mask) { context_t *context = R700_CONTEXT(ctx); + radeonContextPtr radeon = &context->radeon; __DRIdrawable *dPriv = radeon_get_drawable(&context->radeon); const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask[0]); GLbitfield swrast_mask = 0, tri_mask = 0; @@ -60,6 +61,8 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) context->radeon.front_buffer_dirty = GL_TRUE; } + radeon_prepare_render(radeon); + if( GL_TRUE == r700ClearFast(context, mask) ) { return; diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index 1929b7cc12..ba55f38e05 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -625,11 +625,11 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input stride = (input[i]->StrideB == 0) ? getTypeSize(input[i]->Type) * input[i]->Size : input[i]->StrideB; - if (input[i]->Type == GL_DOUBLE || input[i]->Type == GL_UNSIGNED_INT || input[i]->Type == GL_INT || + if (input[i]->Type == GL_DOUBLE || input[i]->Type == GL_UNSIGNED_INT || input[i]->Type == GL_INT #if MESA_BIG_ENDIAN - getTypeSize(input[i]->Type) != 4 || + || getTypeSize(input[i]->Type) != 4 #endif - stride < 4) + ) { r700ConvertAttrib(ctx, count, input[i], &context->stream_desc[index]); } @@ -637,19 +637,10 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input { if (input[i]->BufferObj->Name) { - if (stride % 4 != 0) - { - assert(((intptr_t) input[i]->Ptr) % input[i]->StrideB == 0); - r700AlignDataToDword(ctx, input[i], count, &context->stream_desc[index]); - context->stream_desc[index].is_named_bo = GL_FALSE; - } - else - { - context->stream_desc[index].stride = input[i]->StrideB; - context->stream_desc[index].bo_offset = (intptr_t) input[i]->Ptr; - context->stream_desc[index].bo = get_radeon_buffer_object(input[i]->BufferObj)->bo; - context->stream_desc[index].is_named_bo = GL_TRUE; - } + context->stream_desc[index].stride = input[i]->StrideB; + context->stream_desc[index].bo_offset = (intptr_t) input[i]->Ptr; + context->stream_desc[index].bo = get_radeon_buffer_object(input[i]->BufferObj)->bo; + context->stream_desc[index].is_named_bo = GL_TRUE; } else { @@ -977,6 +968,10 @@ static void r700DrawPrims(GLcontext *ctx, { GLboolean retval = GL_FALSE; + context_t *context = R700_CONTEXT(ctx); + radeonContextPtr radeon = &context->radeon; + radeon_prepare_render(radeon); + /* This check should get folded into just the places that * min/max index are really needed. */ diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index 137f3007ce..6a2a09eaf1 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -461,11 +461,11 @@ static void r700TranslateAttrib(GLcontext *ctx, GLuint unLoc, int count, const s stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size : input->StrideB; - if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT || + if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT #if MESA_BIG_ENDIAN - getTypeSize(input->Type) != 4 || + || getTypeSize(input->Type) != 4 #endif - stride < 4) + ) { pStreamDesc->type = GL_FLOAT; diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c index 13f1f0611b..c1a660af3d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.c +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -708,7 +708,6 @@ void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb) if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) { rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); radeon->front_cliprects = GL_TRUE; - radeon->front_buffer_dirty = GL_TRUE; } else { rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer); radeon->front_cliprects = GL_FALSE; @@ -1132,17 +1131,13 @@ flush_front: if (screen->dri2.loader && (screen->dri2.loader->base.version >= 2) && (screen->dri2.loader->flushFrontBuffer != NULL)) { __DRIdrawable * drawable = radeon_get_drawable(radeon); - (*screen->dri2.loader->flushFrontBuffer)(drawable, drawable->loaderPrivate); - /* Only clear the dirty bit if front-buffer rendering is no longer - * enabled. This is done so that the dirty bit can only be set in - * glDrawBuffer. Otherwise the dirty bit would have to be set at - * each of N places that do rendering. This has worse performances, - * but it is much easier to get correct. + /* We set the dirty bit in radeon_prepare_render() if we're + * front buffer rendering once we get there. */ - if (!radeon->is_front_buffer_rendering) { - radeon->front_buffer_dirty = GL_FALSE; - } + radeon->front_buffer_dirty = GL_FALSE; + + (*screen->dri2.loader->flushFrontBuffer)(drawable, drawable->loaderPrivate); } } } diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index 5a7d52c4d2..92663bf66d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -493,6 +493,50 @@ radeon_bits_per_pixel(const struct radeon_renderbuffer *rb) return _mesa_get_format_bytes(rb->base.Format) * 8; } +/* + * Check if drawable has been invalidated by dri2InvalidateDrawable(). + * Update renderbuffers if so. This prevents a client from accessing + * a backbuffer that has a swap pending but not yet completed. + * + * See intel_prepare_render for equivalent code in intel driver. + * + */ +void radeon_prepare_render(radeonContextPtr radeon) +{ + __DRIcontext *driContext = radeon->dri.context; + __DRIdrawable *drawable; + __DRIscreen *screen; + + screen = driContext->driScreenPriv; + if (!screen->dri2.loader) + return; + + drawable = driContext->driDrawablePriv; + if (drawable->dri2.stamp != driContext->dri2.draw_stamp) { + if (drawable->lastStamp != drawable->dri2.stamp) + radeon_update_renderbuffers(driContext, drawable, GL_FALSE); + + /* Intel driver does the equivalent of this, no clue if it is needed: + * radeon_draw_buffer(radeon->glCtx, &(drawable->driverPrivate)->base); + */ + driContext->dri2.draw_stamp = drawable->dri2.stamp; + } + + drawable = driContext->driReadablePriv; + if (drawable->dri2.stamp != driContext->dri2.read_stamp) { + if (drawable->lastStamp != drawable->dri2.stamp) + radeon_update_renderbuffers(driContext, drawable, GL_FALSE); + driContext->dri2.read_stamp = drawable->dri2.stamp; + } + + /* If we're currently rendering to the front buffer, the rendering + * that will happen next will probably dirty the front buffer. So + * mark it as dirty here. + */ + if (radeon->is_front_buffer_rendering) + radeon->front_buffer_dirty = GL_TRUE; +} + void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable, GLboolean front_only) @@ -514,6 +558,11 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable, screen = context->driScreenPriv; radeon = (radeonContextPtr) context->driverPrivate; + /* Set this up front, so that in case our buffers get invalidated + * while we're getting new buffers, we don't clobber the stamp and + * thus ignore the invalidate. */ + drawable->lastStamp = drawable->dri2.stamp; + if (screen->dri2.loader && (screen->dri2.loader->base.version > 2) && (screen->dri2.loader->getBuffersWithFormat != NULL)) { @@ -650,6 +699,13 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable, rb->base.Height = drawable->h; rb->has_surface = 0; + /* r6xx+ tiling */ + rb->tile_config = radeon->radeonScreen->tile_config; + rb->group_bytes = radeon->radeonScreen->group_bytes; + rb->num_channels = radeon->radeonScreen->num_channels; + rb->num_banks = radeon->radeonScreen->num_banks; + rb->r7xx_bank_op = radeon->radeonScreen->r7xx_bank_op; + if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_bo) { if (RADEON_DEBUG & RADEON_DRI) fprintf(stderr, "(reusing depth buffer as stencil)\n"); @@ -678,7 +734,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable, bo->flags |= RADEON_BO_FLAGS_MACRO_TILE; if (tiling_flags & RADEON_TILING_MICRO) bo->flags |= RADEON_BO_FLAGS_MICRO_TILE; - + } if (buffers[i].attachment == __DRI_BUFFER_DEPTH) { diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index 5156c5d0d0..f06e5fdf24 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -93,6 +93,13 @@ struct radeon_renderbuffer GLuint pf_pending; /**< sequence number of pending flip */ GLuint vbl_pending; /**< vblank sequence number of pending flip */ __DRIdrawable *dPriv; + + /* r6xx+ tiling */ + GLuint tile_config; + GLint group_bytes; + GLint num_channels; + GLint num_banks; + GLint r7xx_bank_op; }; struct radeon_framebuffer @@ -614,5 +621,6 @@ GLboolean radeonMakeCurrent(__DRIcontext * driContextPriv, __DRIdrawable * driDrawPriv, __DRIdrawable * driReadPriv); extern void radeonDestroyContext(__DRIcontext * driContextPriv); +void radeon_prepare_render(radeonContextPtr radeon); #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_pixel_read.c b/src/mesa/drivers/dri/radeon/radeon_pixel_read.c index dadb8002c7..fb741173ca 100644 --- a/src/mesa/drivers/dri/radeon/radeon_pixel_read.c +++ b/src/mesa/drivers/dri/radeon/radeon_pixel_read.c @@ -179,6 +179,9 @@ radeonReadPixels(GLcontext * ctx, GLenum format, GLenum type, const struct gl_pixelstore_attrib *pack, GLvoid * pixels) { + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + radeon_prepare_render(radeon); + if (do_blit_readpixels(ctx, x, y, width, height, format, type, pack, pixels)) return; diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index 82107cc6ae..fa97a19302 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -213,6 +213,10 @@ static const GLuint __driNConfigOptions = 17; static int getSwapInfo( __DRIdrawable *dPriv, __DRIswapInfo * sInfo ); +#ifndef RADEON_INFO_TILE_CONFIG +#define RADEON_INFO_TILE_CONFIG 0x6 +#endif + static int radeonGetParam(__DRIscreen *sPriv, int param, void *value) { @@ -232,6 +236,9 @@ radeonGetParam(__DRIscreen *sPriv, int param, void *value) case RADEON_PARAM_NUM_Z_PIPES: info.request = RADEON_INFO_NUM_Z_PIPES; break; + case RADEON_INFO_TILE_CONFIG: + info.request = RADEON_INFO_TILE_CONFIG; + break; default: return -EINVAL; } @@ -376,6 +383,21 @@ static const __DRItexBufferExtension r600TexBufferExtension = { }; #endif +static void +radeonDRI2Flush(__DRIdrawable *drawable) +{ + radeonContextPtr rmesa; + + rmesa = (radeonContextPtr) drawable->driContextPriv->driverPrivate; + radeonFlush(rmesa->glCtx); +} + +static const struct __DRI2flushExtensionRec radeonFlushExtension = { + { __DRI2_FLUSH, __DRI2_FLUSH_VERSION }, + radeonDRI2Flush, + dri2InvalidateDrawable, +}; + static int radeon_set_screen_flags(radeonScreenPtr screen, int device_id) { screen->device_id = device_id; @@ -1305,6 +1327,56 @@ radeonCreateScreen2(__DRIscreen *sPriv) else screen->chip_flags |= RADEON_CLASS_R600; + /* r6xx+ tiling */ + if (IS_R600_CLASS(screen) && (sPriv->drm_version.minor >= 6)) { + ret = radeonGetParam(sPriv, RADEON_INFO_TILE_CONFIG, &temp); + if (ret) + fprintf(stderr, "failed to get tiling info\n"); + else { + screen->tile_config = temp; + screen->r7xx_bank_op = 0; + switch((screen->tile_config & 0xe) >> 1) { + case 0: + screen->num_channels = 1; + break; + case 1: + screen->num_channels = 2; + break; + case 2: + screen->num_channels = 4; + break; + case 3: + screen->num_channels = 8; + break; + default: + fprintf(stderr, "bad channels\n"); + break; + } + switch((screen->tile_config & 0x30) >> 4) { + case 0: + screen->num_banks = 4; + break; + case 1: + screen->num_banks = 8; + break; + default: + fprintf(stderr, "bad banks\n"); + break; + } + switch((screen->tile_config & 0xc0) >> 6) { + case 0: + screen->group_bytes = 256; + break; + case 1: + screen->group_bytes = 512; + break; + default: + fprintf(stderr, "bad group_bytes\n"); + break; + } + } + } + if (IS_R300_CLASS(screen)) { ret = radeonGetParam(sPriv, RADEON_PARAM_NUM_GB_PIPES, &temp); if (ret) { @@ -1379,6 +1451,8 @@ radeonCreateScreen2(__DRIscreen *sPriv) screen->extensions[i++] = &r600TexBufferExtension.base; #endif + screen->extensions[i++] = &radeonFlushExtension.base; + screen->extensions[i++] = NULL; sPriv->extensions = screen->extensions; diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.h b/src/mesa/drivers/dri/radeon/radeon_screen.h index 0d7e335fa3..2b33201a53 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.h +++ b/src/mesa/drivers/dri/radeon/radeon_screen.h @@ -112,6 +112,13 @@ typedef struct radeon_screen { int kernel_mm; drm_radeon_sarea_t *sarea; /* Private SAREA data */ struct radeon_bo_manager *bom; + + /* r6xx+ tiling */ + GLuint tile_config; + GLint group_bytes; + GLint num_channels; + GLint num_banks; + GLint r7xx_bank_op; } radeonScreenRec, *radeonScreenPtr; #define IS_R100_CLASS(screen) \ diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c index 1adb609603..9dfe2dd243 100644 --- a/src/mesa/drivers/dri/radeon/radeon_span.c +++ b/src/mesa/drivers/dri/radeon/radeon_span.c @@ -111,7 +111,6 @@ static GLubyte *r200_depth_4byte(const struct radeon_renderbuffer * rrb, * two main types: * - 1D (akin to macro-linear/micro-tiled on older asics) * - 2D (akin to macro-tiled/micro-tiled on older asics) - * only 1D tiling is implemented below */ #if defined(RADEON_R600) static inline GLint r600_1d_tile_helper(const struct radeon_renderbuffer * rrb, @@ -208,12 +207,190 @@ static inline GLint r600_1d_tile_helper(const struct radeon_renderbuffer * rrb, return offset; } +static inline GLint r600_log2(GLint n) +{ + GLint log2 = 0; + + while (n >>= 1) + ++log2; + return log2; +} + +static inline GLint r600_2d_tile_helper(const struct radeon_renderbuffer * rrb, + GLint x, GLint y, GLint is_depth, GLint is_stencil) +{ + GLint group_bytes = rrb->group_bytes; + GLint num_channels = rrb->num_channels; + GLint num_banks = rrb->num_banks; + GLint r7xx_bank_op = rrb->r7xx_bank_op; + /* */ + GLint group_bits = r600_log2(group_bytes); + GLint channel_bits = r600_log2(num_channels); + GLint bank_bits = r600_log2(num_banks); + GLint element_bytes = rrb->cpp; + GLint num_samples = 1; + GLint tile_width = 8; + GLint tile_height = 8; + GLint tile_thickness = 1; + GLint macro_tile_width = num_banks; + GLint macro_tile_height = num_channels; + GLint pitch_elements = (rrb->pitch / element_bytes) / tile_width; + GLint height = rrb->base.Height / tile_height; + GLint z = 0; + GLint sample_number = 0; + /* */ + GLint tile_bytes; + GLint macro_tile_bytes; + GLint macro_tiles_per_row; + GLint macro_tiles_per_slice; + GLint slice_offset; + GLint macro_tile_row_index; + GLint macro_tile_column_index; + GLint macro_tile_offset; + GLint pixel_number = 0; + GLint element_offset; + GLint bank = 0; + GLint channel = 0; + GLint total_offset; + GLint group_mask = (1 << group_bits) - 1; + GLint offset_low; + GLint offset_high; + GLint offset = 0; + + switch (num_channels) { + case 2: + default: + // channel[0] = x[3] ^ y[3] + channel |= (((x >> 3) ^ (y >> 3)) & 1) << 0; + break; + case 4: + // channel[0] = x[4] ^ y[3] + channel |= (((x >> 4) ^ (y >> 3)) & 1) << 0; + // channel[1] = x[3] ^ y[4] + channel |= (((x >> 3) ^ (y >> 4)) & 1) << 1; + break; + case 8: + // channel[0] = x[5] ^ y[3] + channel |= (((x >> 5) ^ (y >> 3)) & 1) << 0; + // channel[0] = x[4] ^ x[5] ^ y[4] + channel |= (((x >> 4) ^ (x >> 5) ^ (y >> 4)) & 1) << 1; + // channel[0] = x[3] ^ y[5] + channel |= (((x >> 3) ^ (y >> 5)) & 1) << 2; + break; + } + + switch (num_banks) { + case 4: + // bank[0] = x[3] ^ y[4 + log2(num_channels)] + bank |= (((x >> 3) ^ (y >> (4 + channel_bits))) & 1) << 0; + if (r7xx_bank_op) + // bank[1] = x[3] ^ y[4 + log2(num_channels)] ^ x[5] + bank |= (((x >> 4) ^ (y >> (3 + channel_bits)) ^ (x >> 5)) & 1) << 1; + else + // bank[1] = x[4] ^ y[3 + log2(num_channels)] + bank |= (((x >> 4) ^ (y >> (3 + channel_bits))) & 1) << 1; + break; + case 8: + // bank[0] = x[3] ^ y[5 + log2(num_channels)] + bank |= (((x >> 3) ^ (y >> (5 + channel_bits))) & 1) << 0; + // bank[1] = x[4] ^ y[4 + log2(num_channels)] ^ y[5 + log2(num_channels)] + bank |= (((x >> 4) ^ (y >> (4 + channel_bits)) ^ (y >> (5 + channel_bits))) & 1) << 1; + if (r7xx_bank_op) + // bank[2] = x[5] ^ y[3 + log2(num_channels)] ^ x[6] + bank |= (((x >> 5) ^ (y >> (3 + channel_bits)) ^ (x >> 6)) & 1) << 2; + else + // bank[2] = x[5] ^ y[3 + log2(num_channels)] + bank |= (((x >> 5) ^ (y >> (3 + channel_bits))) & 1) << 2; + break; + } + + tile_bytes = tile_width * tile_height * tile_thickness * element_bytes * num_samples; + macro_tile_bytes = macro_tile_width * macro_tile_height * tile_bytes; + macro_tiles_per_row = pitch_elements / macro_tile_width; + macro_tiles_per_slice = macro_tiles_per_row * (height / macro_tile_height); + slice_offset = (z / tile_thickness) * macro_tiles_per_slice * macro_tile_bytes; + macro_tile_row_index = (y / tile_height) / macro_tile_height; + macro_tile_column_index = (x / tile_width) / macro_tile_width; + macro_tile_offset = ((macro_tile_row_index * macro_tiles_per_row) + macro_tile_column_index) * macro_tile_bytes; + + if (is_depth) { + GLint pixel_offset = 0; + + pixel_number |= ((x >> 0) & 1) << 0; // pn[0] = x[0] + pixel_number |= ((y >> 0) & 1) << 1; // pn[1] = y[0] + pixel_number |= ((x >> 1) & 1) << 2; // pn[2] = x[1] + pixel_number |= ((y >> 1) & 1) << 3; // pn[3] = y[1] + pixel_number |= ((x >> 2) & 1) << 4; // pn[4] = x[2] + pixel_number |= ((y >> 2) & 1) << 5; // pn[5] = y[2] + switch (element_bytes) { + case 2: + pixel_offset = pixel_number * element_bytes * num_samples; + break; + case 4: + /* stencil and depth data are stored separately within a tile. + * stencil is stored in a contiguous tile before the depth tile. + * stencil element is 1 byte, depth element is 3 bytes. + * stencil tile is 64 bytes. + */ + if (is_stencil) + pixel_offset = pixel_number * 1 * num_samples; + else + pixel_offset = (pixel_number * 3 * num_samples) + 64; + break; + } + element_offset = pixel_offset + (sample_number * element_bytes); + } else { + GLint sample_offset; + + switch (element_bytes) { + case 1: + pixel_number |= ((x >> 0) & 1) << 0; // pn[0] = x[0] + pixel_number |= ((x >> 1) & 1) << 1; // pn[1] = x[1] + pixel_number |= ((x >> 2) & 1) << 2; // pn[2] = x[2] + pixel_number |= ((y >> 1) & 1) << 3; // pn[3] = y[1] + pixel_number |= ((y >> 0) & 1) << 4; // pn[4] = y[0] + pixel_number |= ((y >> 2) & 1) << 5; // pn[5] = y[2] + break; + case 2: + pixel_number |= ((x >> 0) & 1) << 0; // pn[0] = x[0] + pixel_number |= ((x >> 1) & 1) << 1; // pn[1] = x[1] + pixel_number |= ((x >> 2) & 1) << 2; // pn[2] = x[2] + pixel_number |= ((y >> 0) & 1) << 3; // pn[3] = y[0] + pixel_number |= ((y >> 1) & 1) << 4; // pn[4] = y[1] + pixel_number |= ((y >> 2) & 1) << 5; // pn[5] = y[2] + break; + case 4: + pixel_number |= ((x >> 0) & 1) << 0; // pn[0] = x[0] + pixel_number |= ((x >> 1) & 1) << 1; // pn[1] = x[1] + pixel_number |= ((y >> 0) & 1) << 2; // pn[2] = y[0] + pixel_number |= ((x >> 2) & 1) << 3; // pn[3] = x[2] + pixel_number |= ((y >> 1) & 1) << 4; // pn[4] = y[1] + pixel_number |= ((y >> 2) & 1) << 5; // pn[5] = y[2] + break; + } + sample_offset = sample_number * (tile_bytes / num_samples); + element_offset = sample_offset + (pixel_number * element_bytes); + } + total_offset = (slice_offset + macro_tile_offset) >> (channel_bits + bank_bits); + total_offset += element_offset; + + offset_low = total_offset & group_mask; + offset_high = (total_offset & ~group_mask) << (channel_bits + bank_bits); + offset = (bank << (group_bits + channel_bits)) + (channel << group_bits) + offset_low + offset_high; + + return offset; +} + /* depth buffers */ static GLubyte *r600_ptr_depth(const struct radeon_renderbuffer * rrb, GLint x, GLint y) { GLubyte *ptr = rrb->bo->ptr; - GLint offset = r600_1d_tile_helper(rrb, x, y, 1, 0); + GLint offset; + if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) + offset = r600_2d_tile_helper(rrb, x, y, 1, 0); + else + offset = r600_1d_tile_helper(rrb, x, y, 1, 0); return &ptr[offset]; } @@ -221,7 +398,11 @@ static GLubyte *r600_ptr_stencil(const struct radeon_renderbuffer * rrb, GLint x, GLint y) { GLubyte *ptr = rrb->bo->ptr; - GLint offset = r600_1d_tile_helper(rrb, x, y, 1, 1); + GLint offset; + if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) + offset = r600_2d_tile_helper(rrb, x, y, 1, 1); + else + offset = r600_1d_tile_helper(rrb, x, y, 1, 1); return &ptr[offset]; } @@ -235,7 +416,10 @@ static GLubyte *r600_ptr_color(const struct radeon_renderbuffer * rrb, if (rrb->has_surface || !(rrb->bo->flags & mask)) { offset = x * rrb->cpp + y * rrb->pitch; } else { - offset = r600_1d_tile_helper(rrb, x, y, 0, 0); + if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) + offset = r600_2d_tile_helper(rrb, x, y, 0, 0); + else + offset = r600_1d_tile_helper(rrb, x, y, 0, 0); } return &ptr[offset]; } diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index f2fcb46688..67be466c3f 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -408,6 +408,8 @@ static GLboolean radeon_run_render( GLcontext *ctx, !radeon_dma_validate_render( ctx, VB )) return GL_TRUE; + radeon_prepare_render(&rmesa->radeon); + tnl->Driver.Render.Start( ctx ); for (i = 0 ; i < VB->PrimitiveCount ; i++) diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index ea796e1a45..5e1718f9df 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -252,6 +252,8 @@ void radeonTclPrimitive( GLcontext *ctx, GLuint se_cntl; GLuint newprim = hw_prim | RADEON_CP_VC_CNTL_TCL_ENABLE; + radeon_prepare_render(&rmesa->radeon); + if (newprim != rmesa->tcl.hw_primitive || !discrete_prim[hw_prim&0xf]) { RADEON_NEWPRIM( rmesa ); diff --git a/src/mesa/drivers/dri/radeon/radeon_tex_copy.c b/src/mesa/drivers/dri/radeon/radeon_tex_copy.c index 29fd31ac23..4cb0bb60c8 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex_copy.c +++ b/src/mesa/drivers/dri/radeon/radeon_tex_copy.c @@ -153,6 +153,9 @@ radeonCopyTexImage2D(GLcontext *ctx, GLenum target, GLint level, _mesa_select_tex_image(ctx, texObj, target, level); int srcx, srcy, dstx, dsty; + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + radeon_prepare_render(radeon); + if (border) goto fail; @@ -202,6 +205,9 @@ radeonCopyTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + radeon_prepare_render(radeon); + if (!do_copy_texsubimage(ctx, target, level, radeon_tex_obj(texObj), (radeon_texture_image *)texImage, xoffset, yoffset, x, y, width, height)) { |