aboutsummaryrefslogtreecommitdiffstats
path: root/src/gallium
Commit message (Collapse)AuthorAgeFilesLines
* draw_llvm: fix segfaults on non-SSE2 CPUs where it is disabled (v2)Luca Barbieri2010-08-163-17/+32
| | | | | | | | | | | | | | | Changes in v2: - Change function name Currently draw_llvm refuses to create itself on non-SSE2 CPUs due to an alleged LLVM bug. However, this is implemented improperly, because other parts of draw still attempt to access draw->llvm, resulting in segfaults. Instead, put the check in debug_get_option_draw_use_llvm, check that before calling draw_llvm_create, and then check whether draw->llvm is non-null everywhere else.
* translate_sse: major rewrite (v5)Luca Barbieri2010-08-162-239/+936
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NOTE: Win64 is untested, and is thus currently disabled. If you have such a system, please enable it and report whether it works. To enable it, change src/gallium/auxiliary/translate/translate.c Changes in v5: - On Win64, preserve %xmm6 and %xmm7 as required by the ABI - Use _WIN64 instead of WIN64 Changes in v4: - Use x86_target() and x86_target_caps() - Enable translate_sse in x86-64, but not in Win64 Changes in v3: - Win64 support (untested) - Use u_cpu_detect.h constants instead of #ifs Changes in v2: - Minimize #ifs - Give a name to magic number CHANNELS_0001 - Add support for CPUs without SSE (only memcpy and swizzles, like non SSE2) - Fixed comments translate_sse is currently very limited to the point of being useless in essentially all cases. In particular, it only support some float32 and unorm8 formats and doesn't work on x86-64. This commit rewrites it to support: 1. Dumb memory copy for any pair of identical formats 2. All formats that are swizzles of each other 3. Converting 32/64-bit floats and all 8/16/32-bit integers to 32-bit float 4. Converting unorm8/snorm8 to snorm16 and uscaled8/sscaled8 to sscaled16 5. Support for x86-64 (doesn't take advantage of it in any way though) This new translate can even be useful to translate index buffers for cards that lack 8-bit index support. It passes the testsuite I wrote, but note that this is a major change, and more testing would be great.
* rtasm: add minimal x86-64 support and new instructions (v5)Luca Barbieri2010-08-163-40/+551
| | | | | | | | | | | | | | | | | | | | | | | | Changes in v5: - Add sse2_movdqa Changes in v4: - Use _WIN64 instead of WIN64 Changes in v3: - Add target and target caps functions, so that they could be different in principle from the current CPU and they don't need #ifs to check Changes in v2: - Win64 support (untested) - Use u_cpu_detect.h constants instead of #ifs This commit adds minimal x86-64 support: only movs between registers are supported for r8-r15, and x64_rexw() must be used to ask for 64-bit operations. It also adds several new instructions for the new translate_sse code. movdqa
* translate: add support for 8/16-bit indicesLuca Barbieri2010-08-165-19/+108
| | | | | Currently, only 32-bit indices are supported, but some use cases translate needs support for all types.
* translate_sse: remove useless generated function wrappersLuca Barbieri2010-08-161-51/+4
| | | | | | | | | | Currently translate_sse puts two trivial wrappers in the translate vtable. These slow it down and enlarge the source code for no gain, except perhaps the ability to set a breakpoint there, so remove them. Breakpoints can be set on the caller of the translate functions, with no loss of functionality.
* translate_generic: factor out common code between linear and indexedLuca Barbieri2010-08-161-115/+62
| | | | This moves the common code into a separate ALWAYS_INLINE function.
* translate_generic: use memcpy if possible (v3)Luca Barbieri2010-08-161-33/+75
| | | | | | | | | | | | | | | | | | | | | | | | Changes in v3: - If we can do a copy, don't try to get an emit func, as that can assert(0) Changes in v2: - Add comment regarding copy_size When used in GPU drivers, translate can be used to simultaneously perform a gather operation, and convert away from unsupported formats. In this use case, input and output formats will often be identical: clearly it would make sense to use a memcpy in this case. Instead, translate will insist to convert to and from 32-bit floating point numbers. This is not only extremely expensive, but it also loses precision for 32/64-bit integers and 64-bit floating point numbers. This patch changes translate_generic to just use memcpy if the formats are identical, non-blocked, and with an integral number of bytes per pixel (note that all sensible vertex formats are like this).
* drwa: Fix polygon edge flags.Chia-I Wu2010-08-161-1/+1
| | | | | Fix a copy-and-paste error introduced by f141abdc8fdbff41e16b0ce53fa3fa8fba32a7f9.
* draw: No need to make max_vertices even.Chia-I Wu2010-08-166-30/+0
| | | | | | | | | Triangle strip alternates the front/back orientation of its triangles. max_vertices was made even so that varray never splitted a triangle strip at the wrong positions. It did not work with triangle strips with adjacencies. And it is no longer relevant with vsplit.
* draw: Remove DRAW_PIPE_MAX_VERTICES and DRAW_PIPE_FLAG_MASK.Chia-I Wu2010-08-166-30/+19
| | | | | The higher bits of draw elements are no longer used for the stipple or edge flags.
* drwa: Add PRIMITIVE macro to vsplit.Chia-I Wu2010-08-162-20/+31
| | | | | PRIMITIVE is used by the indexed path to flush the entire primitive with custom vertex count checks. It replaces the existing fast path.
* draw: last_vertex_last is always true for GS and SO.Chia-I Wu2010-08-162-9/+3
| | | | | That is, OpenGL decomposition rule is assumed. There should be a pipe_context state to specify the rules.
* draw: Remove varray and vcache.Chia-I Wu2010-08-169-1284/+2
| | | | They have been deprecated by vsplit.
* draw: Replace vcache by vsplit.Chia-I Wu2010-08-163-26/+4
| | | | | | | vcache decomposes primitives while vsplit splits primitives. Splitting is generally easier to do and is faster. More importantly, vcache depends on flatshade_first to decompose. The outputs may have incorrect vertex order which is significant to GS.
* draw: Replace varray by vsplit.Chia-I Wu2010-08-162-8/+9
| | | | | vsplit is a superset of varray. It sets the split flags comparing to varray.
* draw: Add vsplit frontend.Chia-I Wu2010-08-166-1/+695
| | | | | | | | vsplit is based on varray. It sets the split flags when a primitive is splitted. It also has support for indexed primitives. For indexed primitives, unlike vcache, vsplit splits the primitives instead of decomposes them.
* draw: Add new util function draw_pt_trim_count.Chia-I Wu2010-08-163-10/+9
| | | | draw_pt_trim_count is renamed from trim in draw_pt.c.
* draw: Simplify frontend interface a little.Chia-I Wu2010-08-165-21/+11
| | | | | The run method is simplified to take the start vertex and the vertex count.
* draw: Add prim flags to middle ends.Chia-I Wu2010-08-167-26/+46
| | | | | Update the middle end interface to pass the primitive flags from the frontends to the pipeline. No frontend sets the flags yet.
* draw: Add flags to draw_prim_info.Chia-I Wu2010-08-169-11/+36
| | | | | | | | | A primitive may be splitted in frontends. The splitted primitives should convey certain flag bits so that the decomposer can correctly decide the stipple or edge flags. This commit adds flags to draw_prim_info and updates the decomposer to honor the flags. Frontends and middle ends will be updated later.
* r300g: fix an invalid pointer in freeMarek Olšák2010-08-161-1/+1
|
* r300g: Let hyperz init failnobled2010-08-163-6/+20
| | | | Signed-off-by: Marek Olšák <[email protected]>
* r300g: Fix leaks in failed context creationnobled2010-08-161-34/+48
| | | | | | | | | | | This changes r300_destroy_context() so it can be called on a partially-initialized context, and uses it when r300_create_context() hits a fatal error. This makes sure r300_create_context() doesn't leak memory or neglect to call r300_update_num_contexts() when it fails. Signed-off-by: Marek Olšák <[email protected]>
* r300g: Fix macronobled2010-08-161-1/+3
| | | | | | | This fixes a potential bug if (has_hyperz) is false (it would still init the atom as if has_hyperz were true). Signed-off-by: Marek Olšák <[email protected]>
* r300/compiler: implement DP2 opcodeMarek Olšák2010-08-161-1/+1
|
* r300/compiler: implement SSG opcodeMarek Olšák2010-08-161-1/+1
|
* llvmpipe: special case triangles which fall in a single 16x16 blockKeith Whitwell2010-08-153-0/+106
| | | | | Check for these and route them to a dedicated handler with one fewer levels of recursive rasterization.
* llvmpipe: consolidate several loops in lp_rast_triangleKeith Whitwell2010-08-151-18/+15
|
* llvmpipe: remove all traces of step arrays, pos_tablesKeith Whitwell2010-08-153-114/+0
| | | | | No need to calculate these values any longer, nor to store them in the bin data. Improves isosurf a bit more, 115->123 fps.
* llvmpipe: eliminate last usage of step array in rast_tmp.hKeith Whitwell2010-08-152-16/+57
| | | | | | | | | | | | For 16 and 64 pixel levels, calculate a mask which is linear in x and y (ie not in the swizzle layout). When iterating over full and partial masks, figure out position by manipulating the bit number set in the mask, rather than relying on postion arrays. Similarly, calculate the lower-level c values from dcdx, dcdy and the position rather than relying on the step array.
* llvmpipe: don't refer to plane->step when dcdx or dcdy would doKeith Whitwell2010-08-151-6/+6
|
* llvmpipe: also use build_mask at 16, 64 pixel levelsKeith Whitwell2010-08-151-28/+19
|
* llvmpipe: version of block4 which doesn't need the full step arrayKeith Whitwell2010-08-152-7/+27
| | | | No noticable slowdown with isosurf.
* llvmpipe: reorganize block4 loop, nice speedupKeith Whitwell2010-08-152-12/+19
| | | | | isosurf 95->115 fps just by exchanging the two inner loops in this function...
* gallium/docs: Cleanup debugging. Spelling, grammar, organization.Corbin Simpson2010-08-151-25/+30
|
* gallium/docs: Add formatting for envvar role; change debugging.Corbin Simpson2010-08-154-30/+44
| | | | | Per Jakob's request. Not super-pretty, but it's a good point for modding later.
* r300g: mark HiZ/ZMask_clear atoms as non-dirty after emission in clearMarek Olšák2010-08-151-1/+4
|
* r300g: do not use HiZ if HiZ RAM is not properly initializedMarek Olšák2010-08-155-18/+23
|
* r300g: rename dirty_zmask -> zmask_in_useMarek Olšák2010-08-155-13/+13
|
* r300g: do not clear with blitter if we clear just the ZMask RAMMarek Olšák2010-08-151-9/+37
| | | | | This skips the blitter clear path entirely if the color is not cleared and the depth+stencil is cleared with the ZMask.
* r300g: do not use fastfill if ZMask RAM is not properly initializedMarek Olšák2010-08-154-18/+31
| | | | z_fastfill -> dirty_zmask[level].
* r300g: separate num_cs_end_dwords out from prepare_for_renderingMarek Olšák2010-08-153-24/+25
|
* gallium/docs: Add rbug to distributionJakob Bornecrantz2010-08-151-0/+5
|
* gallium/docs: Add a debugging sectionJakob Bornecrantz2010-08-152-0/+97
|
* gallium: Make printing info on debug builds default offJakob Bornecrantz2010-08-154-4/+4
| | | | | | | This commit silences the printing off most of the debug information when running debug builds. The big culprits are: the tgsi sanity checker that gets run on all shaders on debug; all the options; and finaly the cpu caps printer.
* gallivm: Remove unnecessary header.Vinson Lee2010-08-141-1/+0
|
* scons: Fix r600g build.Vinson Lee2010-08-142-1/+26
|
* r600g: Remove unnecessary headers.Vinson Lee2010-08-142-2/+0
|
* r300g: Remove unnecessary header.Vinson Lee2010-08-141-1/+0
|
* u_cpu_detect: remove arch and little_endianLuca Barbieri2010-08-144-35/+9
| | | | | This logic duplicates the one in p_config.h, so remove it and adjust the only two places that were using it.