search for: bld

Displaying 20 results from an estimated 211 matches for "bld".

Did you mean: bd
2015 Feb 23
2
[PATCH 1/2] nv50/ir: add fp64 support on G200 (NVA0)
...t computes the + * appropriate function. As a first guess, we use the "quake" style + * approximation for RSQ: + * + * 0x5fe6eb50c7b537a9 - num >> 1 + * + * For RCP, we will then square it. + */ + Value *abs, *guess, *parts[2], *input[2], *shr[4], *pred; + + bld.setPosition(i, false); + + abs = bld.mkOp1v(OP_ABS, TYPE_F64, bld.getSSA(8), i->getSrc(0)); + + parts[0] = bld.loadImm(NULL, 0xc7b537a9); + parts[1] = bld.loadImm(NULL, 0x5fe6eb50); + guess = bld.mkOp2v(OP_MERGE, TYPE_F64, bld.getSSA(8), parts[0], parts[1]); + + bld.mkSplit(input, 4, a...
2015 Feb 23
2
[Mesa-dev] [PATCH 2/2] nvc0/ir: improve precision of double RCP/RSQ results
...ouveau/codegen/nv50_ir_lowering_nvc0.cpp > index 87e75e1..9767566 100644 > --- a/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp > +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp > @@ -77,8 +77,9 @@ NVC0LegalizeSSA::handleRCPRSQ(Instruction *i) > bld.setPosition(i, false); > > // 1. Take the source and it up. > - Value *src[2], *dst[2], *def = i->getDef(0); > - bld.mkSplit(src, 4, i->getSrc(0)); > + Value *input = i->getSrc(0); > + Value *src[2], *dst[2], *guess, *def = i->getDef(0); > + bld.mkSp...
2015 Feb 23
0
[PATCH 2/2] nvc0/ir: improve precision of double RCP/RSQ results
...p b/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp index 87e75e1..9767566 100644 --- a/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp @@ -77,8 +77,9 @@ NVC0LegalizeSSA::handleRCPRSQ(Instruction *i) bld.setPosition(i, false); // 1. Take the source and it up. - Value *src[2], *dst[2], *def = i->getDef(0); - bld.mkSplit(src, 4, i->getSrc(0)); + Value *input = i->getSrc(0); + Value *src[2], *dst[2], *guess, *def = i->getDef(0); + bld.mkSplit(src, 4, input); // 2. We...
2015 Feb 23
0
[Mesa-dev] [PATCH 2/2] nvc0/ir: improve precision of double RCP/RSQ results
...ir_lowering_nvc0.cpp >> index 87e75e1..9767566 100644 >> --- a/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp >> +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp >> @@ -77,8 +77,9 @@ NVC0LegalizeSSA::handleRCPRSQ(Instruction *i) >> bld.setPosition(i, false); >> >> // 1. Take the source and it up. >> - Value *src[2], *dst[2], *def = i->getDef(0); >> - bld.mkSplit(src, 4, i->getSrc(0)); >> + Value *input = i->getSrc(0); >> + Value *src[2], *dst[2], *guess, *def = i->getDef...
2017 Dec 20
2
[PATCH] gm107/ir: use lane 0 for manual textureGrad handling
...s the same logic, but using SM50-friendly primitives. + static const uint8_t qOps[2] = + { QUADOP(MOV2, ADD, MOV2, ADD), QUADOP(MOV2, MOV2, ADD, ADD) }; Value *def[4][4]; - Value *crd[3]; + Value *crd[3], *arr, *shadow; Value *tmp; Instruction *tex, *add; - Value *zero = bld.loadImm(bld.getSSA(), 0); + Value *quad = bld.mkImm(SHFL_BOUND_QUAD); int l, c; const int dim = i->tex.target.getDim() + i->tex.target.isCube(); const int array = i->tex.target.isArray(); @@ -115,35 +112,40 @@ GM107LoweringPass::handleManualTXD(TexInstruction *i) for (c...
2014 May 18
1
[PATCH 1/2] nv50/ir: fix s32 x s32 -> high s32 multiply logic
...ing_nv50.cpp @@ -37,18 +37,25 @@ namespace nv50_ir { // ah*bl 00 // // fffe0001 + fffe0001 +// +// Note that this sort of splitting doesn't work for signed values, so we +// compute the sign on those manually and then perform an unsigned multiply. static bool expandIntegerMUL(BuildUtil *bld, Instruction *mul) { const bool highResult = mul->subOp == NV50_IR_SUBOP_MUL_HIGH; - DataType fTy = mul->sType; // full type - DataType hTy; + DataType fTy; // full type + switch (mul->sType) { + case TYPE_S32: fTy = TYPE_U32; break; + case TYPE_S64: fTy = TYPE_U64; brea...
2019 Oct 14
1
[PATCH] gm107/ir: fix loading z offset for layered 3d image bindings
...C0LoweringPass::loadSuInfo32(Value *ptr, int slot, uint32_t off, bool bindless { uint32_t base = slot * NVC0_SU_INFO__STRIDE; + // We don't upload surface info for bindless for GM107+ + assert(!bindless || targ->getChipset() < NVISA_GM107_CHIPSET); + if (ptr) { ptr = bld.mkOp2v(OP_ADD, TYPE_U32, bld.getSSA(), ptr, bld.mkImm(slot)); if (bindless) @@ -2204,7 +2207,7 @@ getDestType(const ImgType type) { } void -NVC0LoweringPass::convertSurfaceFormat(TexInstruction *su) +NVC0LoweringPass::convertSurfaceFormat(TexInstruction *su, Instruction **loaded) {...
2017 Dec 20
0
[PATCH] gm107/ir: use lane 0 for manual textureGrad handling
...y primitives. > + static const uint8_t qOps[2] = > + { QUADOP(MOV2, ADD, MOV2, ADD), QUADOP(MOV2, MOV2, ADD, ADD) }; > Value *def[4][4]; > - Value *crd[3]; > + Value *crd[3], *arr, *shadow; > Value *tmp; > Instruction *tex, *add; > - Value *zero = bld.loadImm(bld.getSSA(), 0); > + Value *quad = bld.mkImm(SHFL_BOUND_QUAD); > int l, c; > const int dim = i->tex.target.getDim() + i->tex.target.isCube(); > const int array = i->tex.target.isArray(); > @@ -115,35 +112,40 @@ GM107LoweringPass::handleManualTXD(TexIns...
2015 Jan 05
0
[PATCH] nv50/ir: change the way float face is returned
...uveau/codegen/nv50_ir_lowering_nv50.cpp index 0d7612e..1ad0860 100644 --- a/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nv50.cpp +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nv50.cpp @@ -1094,8 +1094,9 @@ NV50LoweringPreSSA::handleRDSV(Instruction *i) case SV_FACE: bld.mkInterp(NV50_IR_INTERP_FLAT, def, addr, NULL); if (i->dType == TYPE_F32) { - bld.mkOp2(OP_AND, TYPE_U32, def, def, bld.mkImm(0x80000000)); - bld.mkOp2(OP_XOR, TYPE_U32, def, def, bld.mkImm(0xbf800000)); + bld.mkOp2(OP_OR, TYPE_U32, def, def, bld.mkImm(0x00000001))...
2016 Mar 02
2
EH failures in MCJIT
After re-cmaking and rebuilding everything from scratch, I'm seeing failures in MCJIT. It this something known or expected? I build LLVM/clang with pre-packaged clang-3.7.0, with "-stdlib=libc++". Example failure: /w/bld/org/./bin/lli -remote-mcjit -mcjit-remote-process=/w/bld/org/./bin/lli-child-target /w/src/llvm.org/test/ExecutionEngine/MCJIT/remote/eh.ll -- Exit Code: 184 Command Output (stderr): -- warning: remote mcjit does not support lazy compilation terminating with uncaught exception of type int ERROR:...
2014 Mar 27
3
[LLVMdev] Lots of regtest failures on PPC64/Linux
...yes make check would suffice. Thanks! > > I see two reported failures: > > > FAIL: LLVM :: BugPoint/compile-custom.ll (459 of 9992) > ******************** TEST 'LLVM :: BugPoint/compile-custom.ll' FAILED ******************** > Script: > -- > /usr/home/kparzysz/bld.lv/Release+Asserts/bin/bugpoint -load /usr/home/kparzysz/bld.lv/Release+Asserts/lib/BugpointPasses.so --compile-custom --compile-command="/usr/home/kparzysz/llvm/test/BugPoint/compile-custom.ll.py arg1 arg2" --output-prefix /usr/home/kparzysz/bld.lv/test/BugPoint/Output/compile-custom.ll....
2016 Sep 27
2
[PATCH] nv50/ir: constant fold OP_SPLIT
...pp +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp @@ -920,6 +920,29 @@ ConstantFolding::opnd(Instruction *i, ImmediateValue &imm0, int s) Instruction *newi = i; switch (i->op) { + case OP_SPLIT: { + uint16_t shift = 0; + DataType type = TYPE_NONE; + bld.setPosition(i, false); + if (i->sType == TYPE_U64 || i->sType == TYPE_S64) { + shift = 32; + type = (i->sType == TYPE_U64) ? TYPE_U32 : TYPE_S32; + } + if (i->sType == TYPE_U32 || i->sType == TYPE_S32) { + shift = 16; + type = (i->sTy...
2016 Sep 30
2
[PATCH] nv50/ir: constant fold OP_SPLIT
...t; @@ -920,6 +920,29 @@ ConstantFolding::opnd(Instruction *i, ImmediateValue &imm0, int s) >> Instruction *newi = i; >> >> switch (i->op) { >> + case OP_SPLIT: { >> + uint16_t shift = 0; >> + DataType type = TYPE_NONE; >> + bld.setPosition(i, false); >> + if (i->sType == TYPE_U64 || i->sType == TYPE_S64) { >> + shift = 32; >> + type = (i->sType == TYPE_U64) ? TYPE_U32 : TYPE_S32; >> + } >> + if (i->sType == TYPE_U32 || i->sType == TYPE_S32) { >...
2013 Oct 22
2
[LLVMdev] How to use clang -intergrated-as to compile cross-(os/target) assembly file.
For example, I execute the following command sequences: H:\CI\bld\compilers\musl\src\math\i386>type sqrt.s #.globl _sqrt .global sqrt #.type sqrt, at function #_sqrt: sqrt: fldl 4(%esp) fsqrt fstsw %ax sub $12,%esp fld %st(0) fstpt (%esp) mov (%esp),%ecx and $0x7ff,%ecx cmp $0x400,%ecx j...
2014 Mar 26
2
[LLVMdev] Lots of regtest failures on PPC64/Linux
Hi, On Wed, Mar 26, 2014 at 6:27 PM, Krzysztof Parzyszek < kparzysz at codeaurora.org> wrote: > On 3/26/2014 8:04 AM, İsmail Dönmez wrote: > >> >> Recent trunk has a lot of failures on PPC64/Linux. One seems to be crash >> with a backtrace like: >> > > Is this with "make check"? I can try it on my G5/FreeBSD box when I get > home make
2014 May 13
1
[PATCH 1/2] nv50/ir: make sure that texprep/texquerylod's args get coalesced
Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu> Cc: "10.2" <mesa-stable at lists.freedesktop.org> --- Not 100% sure of the significance of this code, but this seems like the correct thing to do... will definitely run it through a full piglit run before pushing out. src/gallium/drivers/nouveau/codegen/nv50_ir_ra.cpp | 2 ++ 1 file changed, 2 insertions(+) diff --git
2016 Mar 03
2
EH failures in MCJIT
...: > > After re-cmaking and rebuilding everything from scratch, I'm seeing > failures in MCJIT. It this something known or expected? I build > LLVM/clang with pre-packaged clang-3.7.0, with "-stdlib=libc++". > > > Example failure: > > /w/bld/org/./bin/lli -remote-mcjit > -mcjit-remote-process=/w/bld/org/./bin/lli-child-target > /w/src/llvm.org/test/ExecutionEngine/MCJIT/remote/eh.ll > <http://llvm.org/test/ExecutionEngine/MCJIT/remote/eh.ll> > -- > Exit Code: 184 > > Command Output (st...
2015 Jan 09
3
[RESEND/PATCH] nv50/ir: Handle OP_CVT when folding constant expressions
...allium/drivers/nouveau/codegen/nv50_ir_peephole.cpp +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp @@ -997,6 +997,115 @@ ConstantFolding::opnd(Instruction *i, ImmediateValue &imm0, int s) i->op = OP_MOV; break; } + case OP_CVT: { + Storage res; + bld.setPosition(i, true); /* make sure bld is init'ed */ + switch(i->dType) { + case TYPE_U16: + switch (i->sType) { + case TYPE_F32: + if (i->saturate) + res.data.u16 = util_iround(CLAMP(imm0.reg.data.f32, 0, +...
2014 Aug 08
2
[PATCH 1/3] nvc0/ir: add base tex offset for fermi indirect tex case
...VC0LoweringPass::handleTEX(TexInstruction *i) Value *ticRel = i->getIndirectR(); Value *tscRel = i->getIndirectS(); - if (ticRel) + if (ticRel) { i->setSrc(i->tex.rIndirectSrc, NULL); - if (tscRel) + if (i->tex.r) + ticRel = bld.mkOp2v(OP_ADD, TYPE_U32, bld.getScratch(), + ticRel, bld.mkImm(i->tex.r)); + } + if (tscRel) { i->setSrc(i->tex.sIndirectSrc, NULL); + if (i->tex.s) + tscRel = bld.mkOp2v(OP_ADD, TYPE_U32, bld.getScratch(), +...
2016 Feb 25
2
asan link failure when configuring with -DBUILD_SHARED_LIBS=ON
...re when running ninja check on Linux. I'm configuring with shared libraries enabled, so I'm not sure why asan is looking for .a archives. $ cmake -G Ninja -DCMAKE_BUILD_TYPE=Debug -DLLVM_BUILD_TESTS=ON -DCLANG_INCLUDE_DOCS=ON -DBUILD_SHARED_LIBS=ON ../llvm $ ninja check-all FAILED: cd [..]bld/projects/compiler-rt/lib/asan/tests && [...]/bld/./bin/clang ASAN_INST_TEST_OBJECTS.gtest-all.cc.i386-with-calls.o ASAN_INST_TEST_OBJECTS.asan_asm_test.cc.i386-with-calls.o ASAN_INST_TEST_OBJECTS.asan_globals_test.cc.i386-with-calls.o ASAN_INST_TEST_OBJECTS.asan_interface_test.cc.i386-with-...