summaryrefslogtreecommitdiff
path: root/src/amd
AgeCommit message (Expand)AuthorFilesLines
23 hoursac/surface: Fix GFX9 sparse mip info.Bas Nieuwenhuizen1-2/+6
2 daysradv: do not invalidate the L2 metadata cache on compute queuesSamuel Pitoiset1-0/+1
3 daysac: add ifdef __cplusplus guard to headerPierre-Eric Pelloux-Prayer1-0/+9
4 daysnir/loop_unroll: unroll more aggressively if it can improve load schedulingRhys Perry1-0/+1
4 daysaco/ra: Use PhysRegInterval for count_zeroTony Wasserka1-5/+5
4 daysaco/ra: Use PhysRegInterval for collect_vars parametersTony Wasserka1-7/+9
4 daysaco/ra: Use PhysReg when indexing into RegisterFile's containersTony Wasserka1-24/+24
4 daysaco/ra: Use PhysReg for member functions of PhysRegIntervalTony Wasserka1-49/+49
4 daysaco/ra: Remove unused function parameterTony Wasserka1-8/+7
4 daysaco/ra: Use std::all_of to simplify a loopTony Wasserka1-7/+6
4 daysaco/ra: Add helpers to test for intersection/containment of reg intervalsTony Wasserka1-7/+18
4 daysaco/ra: Move commonly repeated code to a helper functionTony Wasserka1-18/+21
4 daysaco/ra: Conservatively refactor get_reg_specified to use PhysRegIntervalTony Wasserka1-13/+9
4 daysaco/ra: Use std::all_of to simplify a loopTony Wasserka1-14/+6
4 daysaco/ra: Use std::find_if(_not) to clean up get_reg_simpleTony Wasserka1-21/+23
4 daysaco/ra: Add iterator interface for PhysRegIntervalTony Wasserka1-4/+49
4 daysaco/ra: Remove always-false conditionsTony Wasserka1-3/+3
4 daysaco/ra: Conservatively refactor existing code to use PhysRegIntervalTony Wasserka1-112/+93
4 daysaco/ra: Introduce PhysRegInterval helper classTony Wasserka1-0/+30
4 daysaco/ra: Update register use bounds before recursing in get_regs_for_copiesTony Wasserka1-2/+1
4 daysaco: remove divergent branches which only jump over very few instructionsDaniel Schürmann1-13/+43
4 daysaco: propagate swizzles when optimizing packed clamp & fmaDaniel Schürmann1-11/+34
4 daysaco: optimize v_pk_fma_f16 -> v_pk_fmac_f16 on GFX10Daniel Schürmann1-1/+6
4 daysaco: optimize packed fnegDaniel Schürmann1-0/+38
4 daysaco: optimize packed clampDaniel Schürmann1-2/+21
4 daysaco: optimize packed mul+add to v_pk_fma_f16Daniel Schürmann1-2/+97
4 daysaco: simplify multiply-add combiningDaniel Schürmann1-48/+36
4 daysradv: vectorize 16bit instructionsDaniel Schürmann2-0/+36
4 daysaco: emit packed 16bit instructionsDaniel Schürmann2-4/+66
4 daysaco: create helpers to emit vop3p instructionsDaniel Schürmann3-11/+61
4 daysaco: change usesModifiers() considering opsel_hi on packed instructionsDaniel Schürmann1-1/+5
4 daysaco: allow SGPRs on every src position for VOP3PDaniel Schürmann1-2/+4
4 daysaco: allow constants/literals on every src position for VOP3PDaniel Schürmann1-4/+12
4 daysaco/RA: fix subdword operands on VOP3P instructionsDaniel Schürmann1-2/+14
4 daysaco: fix VOP3P assembly, VN and validationDaniel Schürmann4-29/+43
4 daysradv: enable DCC for MSAA on GFX10+21.0-branchpointSamuel Pitoiset1-6/+4
4 daysradv: Use L2 coherency on GFX9+.Bas Nieuwenhuizen3-64/+94
4 daysradv: Use L2 for CP DMA on GFX9+.Bas Nieuwenhuizen1-4/+13
4 daysradv: Use access helpers for flushing with meta operations.Bas Nieuwenhuizen10-72/+96
4 daysradv: Do dst invalidations for write accesses.Bas Nieuwenhuizen1-1/+7
4 daysradv: Invalidate CB on SHADER_WRITE for meta operations.Bas Nieuwenhuizen1-0/+12
4 daysradv: Remove redundant WB_L2 flush.Bas Nieuwenhuizen1-1/+0
4 daysradv: disable VK_EXT_sample_locations again on GFX10+Samuel Pitoiset1-1/+1
4 daysradv: enable DCC for mipmaps on GFX10+Samuel Pitoiset1-7/+2
4 daysradv: do not enable DCC for 3D images with mipmaps on GFX10+Samuel Pitoiset1-0/+6
4 daysradv: add support for fast-clearing DCC levels on GFX10+Samuel Pitoiset1-5/+4
4 daysradv: prevent fast-clearing uncompressed DCC levelsSamuel Pitoiset1-0/+4
4 daysac/surface: store DCC mip info into the surfaceSamuel Pitoiset2-0/+11
4 daysaco/tests: don't rely on argument evaluation orderRhys Perry1-21/+21
4 daysradv: skip fast-clear eliminate for CMASK based on a predicateSamuel Pitoiset1-1/+3