search for: op_set

...degen/nv50_ir_peephole.cpp index c89e1c3..cdae3c8 100644 --- a/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp @@ -187,7 +187,8 @@ LoadPropagation::checkSwapSrc01(Instruction *insn) return; } - if (insn->op == OP_SET) + if (insn->op == OP_SET || insn->op == OP_SET_AND || + insn->op == OP_SET_OR || insn->op == OP_SET_XOR) insn->asCmp()->setCond = reverseCondCode(insn->asCmp()->setCond); else if (insn->op == OP_SLCT) -- 1.8.5.5

Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu> --- Pretty sure there's nothing wrong with it, but it looks odd in the code. src/gallium/drivers/nouveau/codegen/nv50_ir_emit_gk110.cpp | 2 ++ src/gallium/drivers/nouveau/codegen/nv50_ir_emit_gm107.cpp | 7 +++++-- src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp | 2 ++ 3 files changed, 9 insertions(+), 2 deletions(-)

...vers/nouveau/codegen/nv50_ir_target_nv50.cpp > +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_target_nv50.cpp > @@ -413,6 +413,8 @@ TargetNV50::isOpSupported(operation op, DataType ty) const > return false; > case OP_SAD: > return ty == TYPE_S32; > + case OP_SET: > + return !isFloatType(ty); > default: > return true; > }

...su->tex.bindless) + v = loadSuInfo32(ind, slot, NVC0_SU_INFO_UNK1C, su->tex.bindless); + else + v = bld.mkOp2v(OP_SHR, TYPE_U32, bld.getSSA(), ind, bld.mkImm(11)); + Value *is_3d = bld.mkOp2v(OP_AND, TYPE_U32, bld.getSSA(), v, bld.mkImm(1)); + pred2d = bld.mkCmp(OP_SET, CC_EQ, TYPE_U32, bld.getSSA(1, FILE_PREDICATE), + TYPE_U32, bld.mkImm(0), is_3d); + + bld.mkOp2(OP_SHR, TYPE_U32, v, v, bld.loadImm(NULL, 16)); + su->moveSources(dim, 1); + su->setSrc(dim, v); + su->tex.target = nv50_ir::TEX_TARGET_3D; + po...

...X come up with a way not to do this for EVERY little access but @@ -1978,6 +1979,7 @@ NVC0LoweringPass::visit(Instruction *i) } i->setIndirect(0, 1, NULL); i->setIndirect(0, 0, ptr); + i->getSrc(0)->reg.file = FILE_MEMORY_GLOBAL; bld.mkCmp(OP_SET, CC_GT, TYPE_U32, pred, TYPE_U32, offset, length); i->setPredicate(CC_NOT_P, pred); if (i->defExists(0)) { @@ -1987,7 +1989,7 @@ NVC0LoweringPass::visit(Instruction *i) break; case OP_ATOM: { - const bool cctl = i->src(0).getFile() == FILE_MEMORY_GLO...

Retrieving the high 32 bits of a signed multiply is rather annoying. It appears that the simplest way to do this is to compute the absolute value of the arguments, and perform a u32 x u32 -> u64 operation. If the arguments' signs differ, then negate the result. Since there is no u64 support in the cvt instruction, we have the perform the 2's complement negation "by hand".

Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu> --- .../drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp | 66 +++++++++++++++++++++- 1 file changed, 63 insertions(+), 3 deletions(-) diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp index dfb093c..e38a3b8 100644 ---

...s but >> @@ -1978,6 +1979,7 @@ NVC0LoweringPass::visit(Instruction *i) >> } >> i->setIndirect(0, 1, NULL); >> i->setIndirect(0, 0, ptr); >> + i->getSrc(0)->reg.file = FILE_MEMORY_GLOBAL; >> bld.mkCmp(OP_SET, CC_GT, TYPE_U32, pred, TYPE_U32, offset, length); >> i->setPredicate(CC_NOT_P, pred); >> if (i->defExists(0)) { >> @@ -1987,7 +1989,7 @@ NVC0LoweringPass::visit(Instruction *i) >> break; >> case OP_ATOM: >> { >&gt...

...s for EVERY little access but > @@ -1978,6 +1979,7 @@ NVC0LoweringPass::visit(Instruction *i) > } > i->setIndirect(0, 1, NULL); > i->setIndirect(0, 0, ptr); > + i->getSrc(0)->reg.file = FILE_MEMORY_GLOBAL; > bld.mkCmp(OP_SET, CC_GT, TYPE_U32, pred, TYPE_U32, offset, length); > i->setPredicate(CC_NOT_P, pred); > if (i->defExists(0)) { > @@ -1987,7 +1989,7 @@ NVC0LoweringPass::visit(Instruction *i) > break; > case OP_ATOM: > { > - const bool cctl = i...

...0LoweringPass::visit(Instruction *i) >>>> } >>>> i->setIndirect(0, 1, NULL); >>>> i->setIndirect(0, 0, ptr); >>>> + i->getSrc(0)->reg.file = FILE_MEMORY_GLOBAL; >>>> bld.mkCmp(OP_SET, CC_GT, TYPE_U32, pred, TYPE_U32, offset, >>>> length); >>>> i->setPredicate(CC_NOT_P, pred); >>>> if (i->defExists(0)) { >>>> @@ -1987,7 +1989,7 @@ NVC0LoweringPass::visit(Instruction *i) >>>> break; &g...

I went through the gallium-nine tree and picked out nouveau patches that are general bug-fixes. The first bunch I'd like to also get into 10.2. I've reviewed all of them and they make sense to me, but sending them out for public review as well in case there are any objections. Unless I hear objections, I'd like to push this by Friday. Christoph Bumiller (11): nv50,nvc0: always pull

...978,6 +1979,7 @@ NVC0LoweringPass::visit(Instruction *i) >>> } >>> i->setIndirect(0, 1, NULL); >>> i->setIndirect(0, 0, ptr); >>> + i->getSrc(0)->reg.file = FILE_MEMORY_GLOBAL; >>> bld.mkCmp(OP_SET, CC_GT, TYPE_U32, pred, TYPE_U32, offset, >>> length); >>> i->setPredicate(CC_NOT_P, pred); >>> if (i->defExists(0)) { >>> @@ -1987,7 +1989,7 @@ NVC0LoweringPass::visit(Instruction *i) >>> break; >>> case...

...(Instruction *i) >>>>> } >>>>> i->setIndirect(0, 1, NULL); >>>>> i->setIndirect(0, 0, ptr); >>>>> + i->getSrc(0)->reg.file = FILE_MEMORY_GLOBAL; >>>>> bld.mkCmp(OP_SET, CC_GT, TYPE_U32, pred, TYPE_U32, offset, >>>>> length); >>>>> i->setPredicate(CC_NOT_P, pred); >>>>> if (i->defExists(0)) { >>>>> @@ -1987,7 +1989,7 @@ NVC0LoweringPass::visit(Instruction *i) >>>>&g...