Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | r300g: use util_format_short_name and util_format_is_plain | Marek Olšák | 2010-05-03 | 4 | -29/+17 |
| | |||||
* | util: add util_format_short_name | Marek Olšák | 2010-05-03 | 1 | -0/+13 |
| | |||||
* | util: add util_format_is_plain | Marek Olšák | 2010-05-03 | 1 | -0/+15 |
| | |||||
* | r300g: refuse to create a texture with size 0 | Marek Olšák | 2010-05-03 | 1 | -0/+11 |
| | |||||
* | r300g: improve texture debug output, split into TEX and TEXALLOC flags | Marek Olšák | 2010-05-03 | 3 | -6/+22 |
| | |||||
* | Reorder LLVM passes, running mem2reg earlier. | Török Edwin | 2010-05-03 | 2 | -4/+4 |
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This gives a ~30% shader optimization time improvement on blender. Tested by comparing the dumped LLVM modules. Current ordering: time ~/llvm-git/obj/Release-Asserts/bin/opt l.bc -constprop -instcombine -mem2reg -gvn -simplifycfg real 0m1.126s user 0m1.108s sys 0m0.012s With this patch: time ~/llvm-git/obj/Release-Asserts/bin/opt l.bc -mem2reg -constprop -instcombine -gvn -simplifycfg real 0m0.885s user 0m0.880s sys 0m0.000s The overall improvement in blender is ~15%. Blender without the patch takes 1m13s: edwin 5934 87.6 11.5 729440 458296 pts/5 SLl+ 17:35 1:13 blender Blender with the patch takes 1m3s: edwin 5726 94.2 11.2 716424 446168 pts/5 SLl+ 17:32 1:03 blender It is still slow with the patch, but better (most of the optimization time is taken up by GVN, see LLVM PR7023). Signed-off-by: Török Edwin <[email protected]> Signed-off-by: José Fonseca <[email protected]> | ||||
* | svga: Silent warning. | José Fonseca | 2010-05-03 | 1 | -1/+1 |
| | |||||
* | gallivm: Display message instead of crashing when sampler generator was not ↵ | José Fonseca | 2010-05-03 | 1 | -0/+8 |
| | | | | supplied for tgsi translation. | ||||
* | translate: Make translate_generic truly generic. | José Fonseca | 2010-05-03 | 1 | -301/+90 |
| | | | | | PIPE_FORMAT_R10G10B10X2_USCALED, half floats, were not supported, so just rely on u_format for (almost) universal format support. | ||||
* | draw: Remove draw_pt_fetch_prepare call from llvm middle end. | José Fonseca | 2010-05-03 | 1 | -12/+0 |
| | | | | Unneeded since we code generate the vertex fecthes. | ||||
* | tgsi: Parse and dump predicates. | José Fonseca | 2010-05-03 | 2 | -1/+63 |
| | |||||
* | gallium: Remove loop register file. | José Fonseca | 2010-05-03 | 9 | -44/+6 |
| | | | | | It was only used for D3D's REP/END/BGNFOR/ENDFOR. D3D's aL register is just like another address register now. | ||||
* | gallivm: Replace predicate assertion failure with warning message. | José Fonseca | 2010-05-03 | 1 | -8/+6 |
| | |||||
* | draw: Disable rtasm compilation when using LLVM. | José Fonseca | 2010-05-03 | 1 | -6/+10 |
| | | | | Saves time and trouble. | ||||
* | softpipe: invalidate cache view when swizzles are different. | Dave Airlie | 2010-05-03 | 1 | -1/+15 |
| | | | | | | | | Current code only invalidated if the texture was different, however we store swizzled values in the cache, so we need to invalidate in that case also. Signed-off-by: Dave Airlie <[email protected]> | ||||
* | gallivm: fix nested break and continue statements | Zack Rusin | 2010-05-02 | 1 | -16/+6 |
| | | | | | | | we were resetting the mask on each new break/continue statement within the same scope. we always need to and the current execution mask with the current break/continue mask to get the correct result (the masks are always ~1 initially) | ||||
* | identity: Move sampler view create and destroy to id_objects.c | Jakob Bornecrantz | 2010-05-03 | 3 | -34/+67 |
| | |||||
* | identity: s/texture/resource/ | Jakob Bornecrantz | 2010-05-03 | 2 | -21/+21 |
| | |||||
* | identity: Whitespace & Style | Jakob Bornecrantz | 2010-05-03 | 4 | -52/+52 |
| | |||||
* | gallium/util: print \n after DXTn printf | Luca Barbieri | 2010-05-03 | 1 | -2/+2 |
| | | | | | | | Re-add commit 2d65a7caf97684aa654088c76a74b632fbd685fa Signed-off-by: Xavier Chantry <[email protected]> Signed-off-by: José Fonseca <[email protected]> | ||||
* | llvmpipe: add lp_test_* to .gitignore | Xavier Chantry | 2010-05-03 | 1 | -0/+4 |
| | | | | | Signed-off-by: Xavier Chantry <[email protected]> Signed-off-by: José Fonseca <[email protected]> | ||||
* | svga: Remove the screen private context. | José Fonseca | 2010-05-03 | 11 | -106/+63 |
| | | | | | | All affected operations have already been moved to context. More cleanup work can be done, in particular with the buffer transfers. | ||||
* | svga: Remove empty file. | José Fonseca | 2010-05-03 | 1 | -2/+0 |
| | |||||
* | svga: Update flags documentation. | José Fonseca | 2010-05-03 | 1 | -8/+5 |
| | |||||
* | softpipe: Fix alpha blending for formats without alpha channel. | José Fonseca | 2010-05-03 | 1 | -60/+108 |
| | | | | | Don't use the dst alpha channel from the tile cache when it does not exist in the true format. | ||||
* | llvmpipe: Fix alpha blending for formats without alpha channel. | José Fonseca | 2010-05-03 | 1 | -2/+51 |
| | | | | | Don't use the dst alpha channel from the swizzled tile when it does not exist in the true format. | ||||
* | svga: Include svga_surface.h in svga_screen.c. | Vinson Lee | 2010-05-02 | 1 | -0/+1 |
| | | | | Fixes svga_screen_init_surface_functions implicit declaration warning. | ||||
* | r300g: Remove unnecessary header. | Vinson Lee | 2010-05-02 | 1 | -1/+0 |
| | |||||
* | Update dri state tracker to use new API aware context create | Xavier Chantry | 2010-05-02 | 2 | -2/+4 |
| | |||||
* | r300g: do not validate buffers in check_cs | Marek Olšák | 2010-05-02 | 1 | -1/+1 |
| | | | | | It's already done in r300_emit_buffer_validate. This also fixes Total Annihilation 3D on debug builds at least. | ||||
* | r300g: fix surface_copy for compressed formats | Marek Olšák | 2010-05-02 | 1 | -1/+2 |
| | | | | No accelerated blitting for these, it's too messy. | ||||
* | retrace: Add sampler_view_destroy | José Fonseca | 2010-05-02 | 1 | -0/+7 |
| | |||||
* | trace: Fix typo in method name. | José Fonseca | 2010-05-02 | 1 | -1/+1 |
| | |||||
* | r300g: always set the pitch of the first miplevel in the tiling flags | Marek Olšák | 2010-05-02 | 1 | -2/+2 |
| | | | | This further reduces the number of DRM calls and flushes. | ||||
* | Revert "r300g: disable point sprites (again)" | Marek Olšák | 2010-05-02 | 1 | -4/+1 |
| | | | | This reverts commit 3e0f1719d411b34f77fd2a053ecf83b43bb836ca. | ||||
* | r300g: merge VS output mapping state (VAP) to rasterizer block state (RS) | Marek Olšák | 2010-05-02 | 8 | -207/+96 |
| | | | | | | | | | | | | These two should be tied together because what's set in VAP or stuffed in GA should be rasterized in RS. Not doing so causes a hardlock. The reason for the merge is that if stuffed texture coordinates (e.g. point sprite texgen) happen to occupy the texcoord slot dedicated to fog or wpos, the two must be relocated to other free slots, which needs remapping the vertex shader outputs. The rasterizer code is now literally a sequence read-rasterize-write. | ||||
* | r300g: inline update_derived_shader_state | Marek Olšák | 2010-05-02 | 1 | -12/+5 |
| | |||||
* | r300g: fix segfault when resizing a DRI2 framebuffer | Marek Olšák | 2010-05-02 | 1 | -37/+4 |
| | | | | | | Resetting tiling flags might happen after a texture is destroyed... Looking at the kernel sources, we don't actually need to reset the tiling flags. | ||||
* | r300g: cache tiling flags to reduce the number of DRM calls | Marek Olšák | 2010-05-02 | 1 | -6/+9 |
| | |||||
* | i965g: s/sprintf/util_snprintf/ | Vinson Lee | 2010-05-02 | 1 | -5/+6 |
| | |||||
* | llvmpipe: Remove unused variable. | Vinson Lee | 2010-05-01 | 1 | -2/+0 |
| | |||||
* | swrastg: Use trace | Jakob Bornecrantz | 2010-05-01 | 2 | -2/+4 |
| | |||||
* | nv50: raise constant buffers size to maximum | Christoph Bumiller | 2010-04-30 | 3 | -15/+10 |
| | | | | | Removed the param heaps, haven't been using them for a long time now. | ||||
* | nv50: relax restriction on surface_copy format equality a bit | Christoph Bumiller | 2010-04-30 | 1 | -1/+27 |
| | |||||
* | nv50: don't segfault on OPCODE_END for empty programs | Christoph Bumiller | 2010-04-30 | 1 | -5/+6 |
| | |||||
* | llvmpipe: fix out-of-bounds texture sampling | Brian Paul | 2010-04-30 | 1 | -0/+12 |
| | | | | | | | | | | | If we're using a wrap mode in which border color sampling is possible it means that texcoords may be outside of the texture image bounds. Fetching the texel may result in a segfault. Use the 'use_border' variable to catch such texcoords and replace the texel offset with zero (which will be in bounds). Fixes segfault in Lightsmark demo, fd.o bug 27877. | ||||
* | llvmpipe: added lp_build_sample_nop() for debugging | Brian Paul | 2010-04-30 | 1 | -7/+29 |
| | |||||
* | svga: Fix index offset | Jakob Bornecrantz | 2010-04-21 | 1 | -2/+3 |
| | |||||
* | draw: Fix memory leaks in llvm code. | José Fonseca | 2010-04-30 | 2 | -0/+26 |
| | |||||
* | svga: Init surface functions | Jakob Bornecrantz | 2010-04-20 | 1 | -0/+1 |
| |