| /xsrc/external/mit/MesaLib/dist/src/panfrost/bifrost/valhall/ |
| H A D | valhall.py | 101 def __init__(self, index, size, is_float = False, swizzle = False, widen = False, lanes = False, lane = None, absneg = False, notted = False, name = ""): 109 self.lane = lane 125 if lane: 126 self.offset['lane'] = self.lane 127 self.bits['lane'] = 2 if size in (8, 32) else 1 151 self.lane = False 203 lane = el.get('lane', Non [all...] |
| H A D | asm.py | 259 elif src.lane and mod in enums[f'lane_{src.size}_bit'].bare_values: 263 encoded |= (val << src.offset['lane']) 277 die_if(not src.lane, "Instruction doesn't take a lane") 281 encoded |= (val << src.lane)
|
| /xsrc/external/mit/MesaLib/dist/src/panfrost/bifrost/ |
| H A D | bi_lower_divergent_indirects.c | 31 * if (lane == 0) 33 * else if (lane == 1) 36 * else if (lane == MAX_LANE) 89 nir_ssa_def *lane = nir_load_subgroup_invocation(b); local in function:bi_lower_divergent_indirects_impl 101 nir_push_if(b, nir_ieq_imm(b, lane, i));
|
| H A D | compiler.h | 39 * for widen = none, H00 for widen = h0, B1111 for widen = b1. For lane, also 226 bi_byte(bi_index idx, unsigned lane) argument 229 assert(lane < 4); 230 idx.swizzle = BI_SWIZZLE_B0000 + lane;
|
| /xsrc/external/mit/MesaLib/dist/src/gallium/drivers/swr/rasterizer/jitter/ |
| H A D | fetch_jit.cpp | 373 // gather SIMD full pixels per lane then shift/mask to move each component to their 397 for (uint32_t lane = 0; lane < mVWidth; ++lane) local in function:FetchJit::CreateGatherOddFormats 400 Value* index = VEXTRACT(pOffsets, C(lane)); 401 Value* mask = VEXTRACT(pMask, C(lane)); 419 Value* pDst = BITCAST(GEP(pDstMem, C(lane)), PointerType::get(mInt8Ty, 0)); 427 Value* pDst = BITCAST(GEP(pDstMem, C(lane)), PointerType::get(mInt16Ty, 0)); 437 Value* pDst = BITCAST(GEP(pDstMem, C(lane)), PointerType::get(mInt16Ty, 0)); 1121 for (int64_t lane local in function:GetSimdValidIndicesGfx 1195 for (int64_t lane = 0; lane < mVWidth; lane++) local in function:FetchJit::GetSimdValidIndicesHelper 1413 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:FetchJit::Shuffle8bpcGatherd16 1789 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:FetchJit::Shuffle16bpcGather16 2035 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:FetchJit::Shuffle16bpcGather [all...] |
| H A D | builder_mem.cpp | 234 /// @param pVecPassthru - SIMD wide vector of values to load when lane is inactive 447 // after pshufb: group components together in each 128bit lane 453 // after PERMD: move and pack xy components into each 128bit lane 479 // if x or z, extract 128bits from lane 0, else for y or w, extract from lane 1 480 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:SwrJit::Builder::Shuffle16bpcGather4 485 vGatherOutput[swizzleIndex] = VEXTRACT(selectedPermute, C(lane)); 502 // shuffle enabled components into lower word of each 32bit lane, 0 extending to 32 bits 547 // after pshufb: group components together in each 128bit lane 553 // after PERMD: move and pack xy and zw components in low 64 bits of each 128bit lane 580 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:SwrJit::Builder::Shuffle8bpcGather4 [all...] |
| /xsrc/external/mit/MesaLib.old/dist/src/gallium/drivers/swr/rasterizer/jitter/ |
| H A D | fetch_jit.cpp | 368 // gather SIMD full pixels per lane then shift/mask to move each component to their 391 for (uint32_t lane = 0; lane < mVWidth; ++lane) local in function:FetchJit::CreateGatherOddFormats 394 Value* index = VEXTRACT(pOffsets, C(lane)); 395 Value* mask = VEXTRACT(pMask, C(lane)); 412 Value* pDst = BITCAST(GEP(pDstMem, C(lane)), PointerType::get(mInt8Ty, 0)); 420 Value* pDst = BITCAST(GEP(pDstMem, C(lane)), PointerType::get(mInt16Ty, 0)); 430 Value* pDst = BITCAST(GEP(pDstMem, C(lane)), PointerType::get(mInt16Ty, 0)); 1115 for (int64_t lane local in function:FetchJit::GetSimdValidIndicesHelper 1333 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:FetchJit::Shuffle8bpcGatherd16 1703 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:FetchJit::Shuffle16bpcGather16 1949 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:FetchJit::Shuffle16bpcGather [all...] |
| H A D | builder_mem.cpp | 234 /// @param pVecPassthru - SIMD wide vector of values to load when lane is inactive 442 // after pshufb: group components together in each 128bit lane 448 // after PERMD: move and pack xy components into each 128bit lane 474 // if x or z, extract 128bits from lane 0, else for y or w, extract from lane 1 475 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:SwrJit::Builder::Shuffle16bpcGather4 480 vGatherOutput[swizzleIndex] = VEXTRACT(selectedPermute, C(lane)); 497 // shuffle enabled components into lower word of each 32bit lane, 0 extending to 32 bits 542 // after pshufb: group components together in each 128bit lane 548 // after PERMD: move and pack xy and zw components in low 64 bits of each 128bit lane 575 uint32_t lane = ((i == 0) || (i == 2)) ? 0 : 1; local in function:SwrJit::Builder::Shuffle8bpcGather4 [all...] |
| /xsrc/external/mit/MesaLib.old/dist/src/gallium/drivers/nouveau/codegen/ |
| H A D | nv50_ir_lowering_gm107.cpp | 122 Value *lane = bld.mkImm(l); local in function:nv50_ir::GM107LoweringPass::handleManualTXD 124 // Make sure lane 0 has the appropriate array/depth compare values 127 bld.mkOp3(OP_SHFL, TYPE_F32, arr, i->getSrc(0), lane, quad); 129 bld.mkOp3(OP_SHFL, TYPE_F32, shadow, i->getSrc(array + dim + indirect), lane, quad); 132 // mov coordinates from lane l to all lanes 134 bld.mkOp3(OP_SHFL, TYPE_F32, crd[c], i->getSrc(c + array), lane, quad); 137 // add dPdx from lane l to lanes dx 139 bld.mkOp3(OP_SHFL, TYPE_F32, tmp, i->dPdx[c].get(), lane, quad); 145 // add dPdy from lane l to lanes dy 147 bld.mkOp3(OP_SHFL, TYPE_F32, tmp, i->dPdy[c].get(), lane, qua [all...] |
| /xsrc/external/mit/MesaLib/dist/src/gallium/drivers/nouveau/codegen/ |
| H A D | nv50_ir_lowering_gm107.cpp | 141 Value *lane = bld.mkImm(l); local in function:nv50_ir::GM107LoweringPass::handleManualTXD 143 // Make sure lane 0 has the appropriate array/depth compare values 146 bld.mkOp3(OP_SHFL, TYPE_F32, arr, i->getSrc(0), lane, quad); 148 bld.mkOp3(OP_SHFL, TYPE_F32, shadow, i->getSrc(array + dim + indirect), lane, quad); 151 // mov coordinates from lane l to all lanes 153 bld.mkOp3(OP_SHFL, TYPE_F32, crd[c], i->getSrc(c + array), lane, quad); 156 // add dPdx from lane l to lanes dx 158 bld.mkOp3(OP_SHFL, TYPE_F32, tmp, i->dPdx[c].get(), lane, quad); 164 // add dPdy from lane l to lanes dy 166 bld.mkOp3(OP_SHFL, TYPE_F32, tmp, i->dPdy[c].get(), lane, qua [all...] |
| /xsrc/external/mit/MesaLib.old/dist/src/gallium/drivers/swr/ |
| H A D | swr_shader.cpp | 416 Value *pTmpPtr = ALLOCA(mFP32Ty, C(4)); // used for dummy write for lane masking 438 for (uint32_t lane = 0; lane < mVWidth; ++lane) { local in function:BuilderSWR::swr_gs_llvm_emit_vertex 439 Value *pLaneOffset = VEXTRACT(pOutputOffset, C(lane)); 440 Value *pStream = LOAD(iface->pGsCtx, {0, SWR_GS_CONTEXT_pStreams, lane}); 444 Value *pLaneMask = VEXTRACT(vMask1, C(lane)); 457 vData = VEXTRACT(vData, C(lane)); 506 Value *pTmpPtr = ALLOCA(mInt8Ty, C(4)); // used for dummy read/write for lane masking 508 for (uint32_t lane local in function:BuilderSWR::swr_gs_llvm_end_primitive 535 for (uint32_t lane = 0; lane < mVWidth; ++lane) local in function:BuilderSWR::swr_gs_llvm_epilogue 660 for (uint32_t lane = 0; lane < mVWidth; ++lane) local in function:BuilderSWR::CompileGS [all...] |
| /xsrc/external/mit/MesaLib/dist/src/gallium/drivers/swr/ |
| H A D | swr_shader.cpp | 766 Value *pTmpPtr = ALLOCA(mFP32Ty, C(4)); // used for dummy write for lane masking 791 for (uint32_t lane = 0; lane < mVWidth; ++lane) { local in function:BuilderSWR::swr_gs_llvm_emit_vertex 792 Value *pLaneOffset = VEXTRACT(pOutputOffset, C(lane)); 793 Value *pStream = LOAD(iface->pGsCtx, {0, SWR_GS_CONTEXT_pStreams, lane}); 797 Value *pLaneMask = VEXTRACT(vMask1, C(lane)); 810 vData = VEXTRACT(vData, C(lane)); 842 for (uint32_t lane = 0; lane < mVWidt local in function:BuilderSWR::swr_gs_llvm_emit_vertex 910 for (uint32_t lane = 0; lane < mVWidth; ++lane) { local in function:BuilderSWR::swr_gs_llvm_end_primitive 936 for (uint32_t lane = 0; lane < mVWidth; ++lane) local in function:BuilderSWR::swr_gs_llvm_epilogue 1092 for (uint32_t lane = 0; lane < mVWidth; lane++) { local in function:BuilderSWR::swr_tcs_llvm_fetch_output 1213 for (uint32_t lane = 0; lane < mVWidth; lane++) { local in function:BuilderSWR::swr_tcs_llvm_store_output 1619 for (uint32_t lane = 0; lane < mVWidth; ++lane) local in function:BuilderSWR::CompileGS [all...] |
| /xsrc/external/mit/xf86-video-ati/dist/src/ |
| H A D | atombios_output.c | 585 5400, // 1 lane, 1.62 Ghz 586 9000, // 1 lane, 2.70 Ghz 587 10800, // 2 lane, 1.62 Ghz 588 18000, // 2 lane, 2.70 Ghz 589 21600, // 4 lane, 1.62 Ghz 590 36000, // 4 lane, 2.70 Ghz 2414 static uint8_t dp_get_lane_status(uint8_t link_status[DP_LINK_STATUS_SIZE], int lane) argument 2416 int i = DP_LANE0_1_STATUS + (lane >> 1); 2417 int s = (lane & 1) * 4; 2424 int lane; local in function:dp_clock_recovery_ok 2446 int lane; local in function:dp_channel_eq_ok 2483 dp_get_adjust_request_voltage(uint8_t link_status[DP_LINK_STATUS_SIZE],int lane) argument 2497 dp_get_adjust_request_pre_emphasis(uint8_t link_status[DP_LINK_STATUS_SIZE],int lane) argument 2561 int lane; local in function:dp_get_adjust_train [all...] |
| /xsrc/external/mit/MesaLib.old/dist/src/gallium/drivers/swr/rasterizer/core/core/ |
| H A D | pa_avx.cpp | 157 INLINE simd4scalar swizzleLaneN(const simdvector& v, int lane) argument 159 switch (lane) 311 INLINE simd4scalar swizzleLaneN(const simd16vector& v, int lane) argument 313 switch (lane) 505 for (uint32_t lane = 0; lane < KNOB_SIMD_WIDTH; ++lane) local in function:PaPatchListTerm 508 uint32_t input_cp = (lane + lane_offset) * TotalControlPoints + cp; 515 uint32_t input_cp = lane * TotalControlPoints + cp; 521 vec[lane] 564 for (uint32_t lane = 0; lane < KNOB_SIMD16_WIDTH; ++lane) local in function:PaPatchListTerm_simd16 1820 const int lane = pa.numPrims - pa.numPrimsComplete - 1; local in function:PaLineLoop1 1872 const int lane = pa.numPrims - pa.numPrimsComplete - 1; local in function:PaLineLoop1_simd16 [all...] |
| H A D | clip.h | 591 SIMD256::set_epi32(0 * sizeof(SIMDVERTEX_T<SIMD_T>), // unused lane 601 // @todo dynamic mask based on actual # of verts generated per lane 606 // transpose clipper output so that each lane's vertices are in SIMD order 626 // tranpose clipper output so that each lane's vertices are in SIMD order 854 // step to the lane 884 DWORD lane; local in function:Clipper::ScatterComponent 885 while (_BitScanForward(&lane, mask)) 887 mask &= ~(1 << lane); 888 const uint8_t* pBuf = reinterpret_cast<const uint8_t*>(pBuffer) + pOffsets[lane]; 889 *(float*)pBuf = pSrc[lane]; [all...] |
| /xsrc/external/mit/MesaLib.old/dist/src/amd/common/ |
| H A D | ac_llvm_build.h | 623 ac_build_readlane(struct ac_llvm_context *ctx, LLVMValueRef src, LLVMValueRef lane); 626 ac_build_writelane(struct ac_llvm_context *ctx, LLVMValueRef src, LLVMValueRef value, LLVMValueRef lane);
|
| /xsrc/external/mit/MesaLib/dist/docs/relnotes/ |
| H A D | 21.3.7.rst | 107 - aux/draw: fix llvm tcs lane vec generation
|
| /xsrc/external/mit/MesaLib/dist/src/amd/llvm/ |
| H A D | ac_llvm_build.h | 502 LLVMValueRef lane); 504 LLVMValueRef ac_build_readlane(struct ac_llvm_context *ctx, LLVMValueRef src, LLVMValueRef lane); 507 LLVMValueRef lane);
|
| H A D | ac_llvm_build.c | 3376 LLVMValueRef lane, bool with_opt_barrier) 3385 if (lane) 3386 lane = LLVMBuildZExt(ctx->builder, lane, ctx->i32, ""); 3389 ac_build_intrinsic(ctx, lane == NULL ? "llvm.amdgcn.readfirstlane" : "llvm.amdgcn.readlane", 3390 ctx->i32, (LLVMValueRef[]){src, lane}, lane == NULL ? 1 : 2, 3397 LLVMValueRef lane, bool with_opt_barrier) 3414 ret_comp = _ac_build_readlane(ctx, src, lane, with_opt_barrier); 3420 ret = _ac_build_readlane(ctx, src, lane, with_opt_barrie 3375 _ac_build_readlane(struct ac_llvm_context * ctx,LLVMValueRef src,LLVMValueRef lane,bool with_opt_barrier) argument 3396 ac_build_readlane_common(struct ac_llvm_context * ctx,LLVMValueRef src,LLVMValueRef lane,bool with_opt_barrier) argument 3439 ac_build_readlane_no_opt_barrier(struct ac_llvm_context * ctx,LLVMValueRef src,LLVMValueRef lane) argument 3445 ac_build_readlane(struct ac_llvm_context * ctx,LLVMValueRef src,LLVMValueRef lane) argument 3450 ac_build_writelane(struct ac_llvm_context * ctx,LLVMValueRef src,LLVMValueRef value,LLVMValueRef lane) argument [all...] |
| /xsrc/external/mit/MesaLib.old/dist/src/gallium/drivers/swr/rasterizer/common/ |
| H A D | simdlib_512_emu.inl | 236 template <int ImmT> // for each 128-bit lane: 534 Integer const& a, Integer const& swiz) // return a[swiz[i]] for each 32-bit lane i (int32) 540 permute_ps(Float const& a, Integer const& swiz) // return a[swiz[i]] for each 32-bit lane i (float)
|
| H A D | simdlib_128_avx.inl | 377 permute_epi32(Integer a, Integer swiz) // return a[swiz[i]] for each 32-bit lane i (float) 383 permute_ps(Float a, Integer swiz) // return a[swiz[i]] for each 32-bit lane i (float)
|
| H A D | simdlib_256_avx2.inl | 185 permute_ps(Float const& a, Integer const& swiz) // return a[swiz[i]] for each 32-bit lane i (float)
|
| /xsrc/external/mit/MesaLib/dist/src/gallium/drivers/swr/rasterizer/common/ |
| H A D | simdlib_512_emu.inl | 236 template <int ImmT> // for each 128-bit lane: 534 Integer const& a, Integer const& swiz) // return a[swiz[i]] for each 32-bit lane i (int32) 540 permute_ps(Float const& a, Integer const& swiz) // return a[swiz[i]] for each 32-bit lane i (float)
|
| H A D | simdlib_128_avx.inl | 377 permute_epi32(Integer a, Integer swiz) // return a[swiz[i]] for each 32-bit lane i (float) 383 permute_ps(Float a, Integer swiz) // return a[swiz[i]] for each 32-bit lane i (float)
|
| H A D | simdlib_256_avx2.inl | 193 permute_ps(Float const& a, Integer const& swiz) // return a[swiz[i]] for each 32-bit lane i (float)
|