Age | Commit message (Collapse) | Author | |
---|---|---|---|
2010-09-25 | gallivm: optimize yuv decoding | Roland Scheidegger | |
this is more a proof to show vector shifts on x86 with per-element shift count are evil. Since we can avoid the shift with a single compare/select, use that instead. Replaces more than 20 instructions (and slow ones at that) with about 3, and cuts compiled shader size with mesa's yuvsqure demo by over 10% (no performance measurements done - but selection is blazing fast). Might want to revisit that for future cpus - unfortunately AVX won't have vector shifts neither, but AMD's XOP will, but even in that case using selection here is probably not slower. | |||
2010-09-25 | gallivm: don't use URem/UDiv when calculating offsets for blocks | Roland Scheidegger | |
While it's true that llvm can and will indeed replace this with bit arithmetic (since block height/width is POT), it does so (llvm 2.7) by element and hence extracts/shifts/reinserts each element individually. This costs about 16 instructions (and extract is not really fast) vs. 1... | |||
2010-09-25 | gallivm: fix copy&paste bug | Roland Scheidegger | |
looks like pot_depth should be used, not pot_height (found by accident, not verified) | |||
2010-09-24 | llvmpipe: make texture border_color dynamic state | Brian Paul | |
2010-09-24 | llvmpipe: make min/max lod and lod bias dynamic state | Brian Paul | |
Before, changing any of these sampler values triggered generation of new JIT code. Added a new flag for the special case of min_lod == max_lod which is hit during auto mipmap generation. | |||
2010-09-22 | gallivm: Add unorm support to lp_build_lerp() | José Fonseca | |
Unfortunately this can cause segfault with LLVM 2.6, if x is a constant. | |||
2010-09-21 | gallivm: fix lp_build_sample_compare() | Brian Paul | |
The old code didn't really make sense. We only need to compare the X channel of the texture (depth) against the texcoord. For (bi)linear sampling we should move the calls to this function and compute the final result as (s1+s2+s3+s4) * 0.25. Someday. This fixes the glean glsl1 shadow2D() tests. See fd.o bug 29307. | |||
2010-09-20 | gallivm: remove debug code | Brian Paul | |
2010-09-17 | gallivm: added missing case for PIPE_TEXTURE_RECT | Brian Paul | |
Fixes fd.o bug 30245 | |||
2010-09-16 | gallivm: Remove unnecessary header. | Vinson Lee | |
2010-09-16 | gallivm: fix wrong return value in bitwise functions | Brian Paul | |
2010-09-16 | gallivm: Clamp indirect register indices to file_max. | José Fonseca | |
Prevents crashes with bogus data, or bad shader translation. | |||
2010-09-16 | gallivm: Start collecting bitwise arithmetic helpers in a new module. | José Fonseca | |
2010-09-16 | gallivm: Fix address register swizzle. | José Fonseca | |
We're actually doing a double swizzling: indirect_reg->Swizzle[indirect_reg->SwizzleX] instead of simply indirect_reg->SwizzleX | |||
2010-09-16 | gallivm: make lp_build_sample_nop public | Keith Whitwell | |
2010-09-16 | gallivm: move i32_vec_type inside the #ifdef | Brian Paul | |
2010-09-16 | gallivm: fix incorrect vector shuffle datatype | Brian Paul | |
The permutation vector must always be a vector of int32 values. | |||
2010-09-15 | gallivm: Remove unnecessary headers. | Vinson Lee | |
2010-09-15 | gallivm: expand AoS sampling to cover all filtering modes | Brian Paul | |
...and all texture targets (1D/2D/3D/CUBE). | |||
2010-09-14 | gallivm: Remove unnecessary header. | Vinson Lee | |
2010-09-11 | gallivm: nr_channels is only valid for formats with plain layout. | José Fonseca | |
This is erroneously throwing non plain formats out of the faster AoS sampling path. Doing 8bit interpolation for single channels such as L8 should be no worse than with floating point. But this may need more investigation. | |||
2010-09-11 | gallivm: Use const keyword on swizzles. | José Fonseca | |
2010-09-11 | gallivm: Allow to TGSI AoS translation to happen in BGRA ordering. | José Fonseca | |
Or any ordering. | |||
2010-09-11 | gallivm: Add a new debug flag to warn about performance issues. | José Fonseca | |
2010-09-11 | gallivm: Helper functions for pointer indirection. | José Fonseca | |
2010-09-11 | gallivm: Cleanup the TGSI <-> sampler interface. | José Fonseca | |
2010-09-11 | gallivm: Add some utility functions to set/get array elements too. | José Fonseca | |
2010-09-11 | gallivm: Basic AoS TGSI -> LLVM IR. | José Fonseca | |
Essentially a variation of the SoA version. | |||
2010-09-11 | gallivm: Move the texture modifiers to the header. | José Fonseca | |
Useful to pass these around. | |||
2010-09-11 | gallivm: s/lp_build_broadcast_aos/lp_build_swizzle_scalar_aos/ | José Fonseca | |
More accurate description of this function purpose. | |||
2010-09-05 | gallivm: Pass condition masks as an unsigned bitmask. | José Fonseca | |
Much more convenient than boolean arrays. | |||
2010-09-05 | gallivm: Cope with tgsi instruction reallocation failure. | José Fonseca | |
2010-08-31 | gallivm: fix bug in nested conditionals | Brian Paul | |
This, plus the previous commit fix fd.o bug 29806. | |||
2010-08-30 | gallivm: Compute the 4 texel offsets for linear filtering en ensemble. | José Fonseca | |
2010-08-30 | gallivm: Disable LLVM's pretty stack trace dumper. | José Fonseca | |
By default LLVM adds a signal handler to output a pretty stack trace. This signal handler is never removed, causing problems when unloading the shared object where the gallium driver resides. Thanks to Chris Li for finding this. | |||
2010-08-30 | gallivm: Correct copy'n'pasted comments. | José Fonseca | |
2010-08-30 | gallivm: Fix lp_build_sum_vector. | José Fonseca | |
The result is scalar, so when argument is zero/undef we can pass vector zero/undef. Also, support the scalar case. | |||
2010-08-29 | util: remove util_is_pot in favor of util_is_power_of_two | Marek Olšák | |
The function was duplicated. | |||
2010-08-24 | gallivm: Include missing header in lp_bld_sample.h. | Vinson Lee | |
Include p_format.h for enum pipe_format symbol. | |||
2010-08-24 | gallivm: Include missing header in lp_bld_pack.h. | Vinson Lee | |
Include p_compiler.h for boolean symbol. | |||
2010-08-21 | gallivm: Emit DIVPS instead of RCPPS. | José Fonseca | |
See comments for detailed rationale. Thanks to Michal Krol and Zack Rusin for detecting and investigating this in detail. | |||
2010-08-20 | gallium: make all checks for PIPE_TEXTURE_2D check for PIPE_TEXTURE_RECT too | Luca Barbieri | |
Searched for them with: git grep -E '[!=]=.*PIPE_TEXTURE_2D|PIPE_TEXTURE_2D.*[!=]=|case.*PIPE_TEXTURE_2D' Behavior hasn't been changed. | |||
2010-08-18 | gallivm: Use proper index to lookup predicate register array. | michal | |
Doesn't fix anything, as those indices were both always 0. | |||
2010-08-17 | gallivm: Fix and re-enable MMX-disabling code | nobled | |
Signed-off-by: José Fonseca <jfonseca@vmware.com> | |||
2010-08-14 | gallivm: Remove unnecessary header. | Vinson Lee | |
2010-08-14 | u_cpu_detect: remove arch and little_endian | Luca Barbieri | |
This logic duplicates the one in p_config.h, so remove it and adjust the only two places that were using it. | |||
2010-08-14 | gallivm: Refactor the Newton-Rapshon steps, and disable once again. | José Fonseca | |
It causes a very ugly corruption on the Earth's halo on Google Earth. | |||
2010-08-11 | gallivm: Fix and enable the extra Newton/Raphson step in lp_build_rcp(). | José Fonseca | |
Thanks to Michal for spotting this. | |||
2010-08-11 | gallivm: Use lp_build_div instead of lp_build_mul + lp_build_rcp. | José Fonseca | |
Single divide, so let lp_build_div decide how to implement this. This will save a multiplication in architectures which don't have a RCP intrinsic. | |||
2010-08-11 | gallivm: Use unsigned shift in lp_build_minify. | José Fonseca | |
Texture dimensions are unsigned. |