RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 18:54:02 +01:00
Code Issues Projects Releases Wiki Activity

[x86] Fix a pretty horrible bug and inconsistency in the x86 asm

Browse Source 
parsing (and latent bug in the instruction definitions).

This is effectively a revert of r136287 which tried to address
a specific and narrow case of immediate operands failing to be accepted
by x86 instructions with a pretty heavy hammer: it introduced a new kind
of operand that behaved differently. All of that is removed with this
commit, but the test cases are both preserved and enhanced.

The core problem that r136287 and this commit are trying to handle is
that gas accepts both of the following instructions:

  insertps $192, %xmm0, %xmm1
  insertps $-64, %xmm0, %xmm1

These will encode to the same byte sequence, with the immediate
occupying an 8-bit entry. The first form was fixed by r136287 but that
broke the prior handling of the second form! =[ Ironically, we would
still emit the second form in some cases and then be unable to
re-assemble the output.

The reason why the first instruction failed to be handled is because
prior to r136287 the operands ere marked 'i32i8imm' which forces them to
be sign-extenable. Clearly, that won't work for 192 in a single byte.
However, making thim zero-extended or "unsigned" doesn't really address
the core issue either because it breaks negative immediates. The correct
fix is to make these operands 'i8imm' reflecting that they can be either
signed or unsigned but must be 8-bit immediates. This patch backs out
r136287 and then changes those places as well as some others to use
'i8imm' rather than one of the extended variants.

Naturally, this broke something else. The custom DAG nodes had to be
updated to have a much more accurate type constraint of an i8 node, and
a bunch of Pat immediates needed to be specified as i8 values.

The fallout didn't end there though. We also then ceased to be able to
match the instruction-specific intrinsics to the instructions so
modified. Digging, this is because they too used i32 rather than i8 in
their signature. So I've also switched those intrinsics to i8 arguments
in line with the instructions.

In order to make the intrinsic adjustments of course, I also had to add
auto upgrading for the intrinsics.

I suspect that the intrinsic argument types may have led everything down
this rabbit hole. Pretty happy with the result.

llvm-svn: 217310
This commit is contained in:
Chandler Carruth 2014-09-06 10:00:01 +00:00
parent 820d92269f
commit 5b09348e8e
18 changed files with 303 additions and 115 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
include/llvm/IR/IntrinsicsX86.td
View File

@ -886,7 +886,7 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
// Vector insert
let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
def int_x86_sse41_insertps : GCCBuiltin<"__builtin_ia32_insertps128">,
Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty,llvm_i32_ty],
Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty, llvm_i8_ty],
[IntrNoMem]>;
}
@ -896,13 +896,13 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
Intrinsic<[llvm_v16i8_ty], [llvm_v16i8_ty, llvm_v16i8_ty,llvm_v16i8_ty],
[IntrNoMem]>;
def int_x86_sse41_pblendw : GCCBuiltin<"__builtin_ia32_pblendw128">,
Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty, llvm_v8i16_ty, llvm_i32_ty],
Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty, llvm_v8i16_ty, llvm_i8_ty],
[IntrNoMem]>;
def int_x86_sse41_blendpd : GCCBuiltin<"__builtin_ia32_blendpd">,
Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_v2f64_ty, llvm_i32_ty],
Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_v2f64_ty, llvm_i8_ty],
[IntrNoMem]>;
def int_x86_sse41_blendps : GCCBuiltin<"__builtin_ia32_blendps">,
Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty, llvm_i32_ty],
Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty, llvm_i8_ty],
[IntrNoMem]>;
def int_x86_sse41_blendvpd : GCCBuiltin<"__builtin_ia32_blendvpd">,
Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_v2f64_ty,llvm_v2f64_ty],
@ -915,17 +915,17 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
// Vector dot product
let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
def int_x86_sse41_dppd : GCCBuiltin<"__builtin_ia32_dppd">,
Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_v2f64_ty,llvm_i32_ty],
Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_v2f64_ty, llvm_i8_ty],
[IntrNoMem, Commutative]>;
def int_x86_sse41_dpps : GCCBuiltin<"__builtin_ia32_dpps">,
Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty,llvm_i32_ty],
Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty, llvm_i8_ty],
[IntrNoMem, Commutative]>;
}
// Vector sum of absolute differences
let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
def int_x86_sse41_mpsadbw : GCCBuiltin<"__builtin_ia32_mpsadbw128">,
Intrinsic<[llvm_v8i16_ty], [llvm_v16i8_ty, llvm_v16i8_ty,llvm_i32_ty],
Intrinsic<[llvm_v8i16_ty], [llvm_v16i8_ty, llvm_v16i8_ty,llvm_i8_ty],
[IntrNoMem, Commutative]>;
}
@ -1171,10 +1171,10 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
def int_x86_avx_blend_pd_256 : GCCBuiltin<"__builtin_ia32_blendpd256">,
Intrinsic<[llvm_v4f64_ty], [llvm_v4f64_ty,
llvm_v4f64_ty, llvm_i32_ty], [IntrNoMem]>;
llvm_v4f64_ty, llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx_blend_ps_256 : GCCBuiltin<"__builtin_ia32_blendps256">,
Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty,
llvm_v8f32_ty, llvm_i32_ty], [IntrNoMem]>;
llvm_v8f32_ty, llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx_blendv_pd_256 : GCCBuiltin<"__builtin_ia32_blendvpd256">,
Intrinsic<[llvm_v4f64_ty], [llvm_v4f64_ty,
llvm_v4f64_ty, llvm_v4f64_ty], [IntrNoMem]>;
@ -1187,7 +1187,7 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
def int_x86_avx_dp_ps_256 : GCCBuiltin<"__builtin_ia32_dpps256">,
Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty,
llvm_v8f32_ty, llvm_i32_ty], [IntrNoMem]>;
llvm_v8f32_ty, llvm_i8_ty], [IntrNoMem]>;
}
// Vector compare
@ -1710,13 +1710,13 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
llvm_v32i8_ty], [IntrNoMem]>;
def int_x86_avx2_pblendw : GCCBuiltin<"__builtin_ia32_pblendw256">,
Intrinsic<[llvm_v16i16_ty], [llvm_v16i16_ty, llvm_v16i16_ty,
llvm_i32_ty], [IntrNoMem]>;
llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx2_pblendd_128 : GCCBuiltin<"__builtin_ia32_pblendd128">,
Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty,
llvm_i32_ty], [IntrNoMem]>;
llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx2_pblendd_256 : GCCBuiltin<"__builtin_ia32_pblendd256">,
Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty, llvm_v8i32_ty,
llvm_i32_ty], [IntrNoMem]>;
llvm_i8_ty], [IntrNoMem]>;
}
// Vector load with broadcast
@ -1955,7 +1955,7 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
llvm_v32i8_ty], [IntrNoMem]>;
def int_x86_avx2_mpsadbw : GCCBuiltin<"__builtin_ia32_mpsadbw256">,
Intrinsic<[llvm_v16i16_ty], [llvm_v32i8_ty, llvm_v32i8_ty,
llvm_i32_ty], [IntrNoMem, Commutative]>;
llvm_i8_ty], [IntrNoMem, Commutative]>;
def int_x86_avx2_movntdqa : GCCBuiltin<"__builtin_ia32_movntdqa256">,
Intrinsic<[llvm_v4i64_ty], [llvm_ptr_ty], [IntrReadMem]>;
}

89
lib/IR/AutoUpgrade.cpp
View File

@ -43,6 +43,22 @@ static bool UpgradeSSE41Function(Function* F, Intrinsic::ID IID,
return true;
}
// Upgrade the declarations of intrinsic functions whose 8-bit immediate mask
// arguments have changed their type from i32 to i8.
static bool UpgradeX86IntrinsicsWith8BitMask(Function *F, Intrinsic::ID IID,
Function *&NewFn) {
// Check that the last argument is an i32.
Type *LastArgType = F->getFunctionType()->getParamType(
F->getFunctionType()->getNumParams() - 1);
if (!LastArgType->isIntegerTy(32))
return false;
// Move this function aside and map down.
F->setName(F->getName() + ".old");
NewFn = Intrinsic::getDeclaration(F->getParent(), IID);
return true;
}
static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
assert(F && "Illegal to upgrade a non-existent Function.");
@ -130,6 +146,51 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
if (Name == "x86.sse41.ptestnzc")
return UpgradeSSE41Function(F, Intrinsic::x86_sse41_ptestnzc, NewFn);
}
// Several blend and other instructions with maskes used the wrong number of
// bits.
if (Name == "x86.sse41.pblendw")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_sse41_pblendw,
NewFn);
if (Name == "x86.sse41.blendpd")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_sse41_blendpd,
NewFn);
if (Name == "x86.sse41.blendps")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_sse41_blendps,
NewFn);
if (Name == "x86.sse41.insertps")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_sse41_insertps,
NewFn);
if (Name == "x86.sse41.dppd")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_sse41_dppd,
NewFn);
if (Name == "x86.sse41.dpps")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_sse41_dpps,
NewFn);
if (Name == "x86.sse41.mpsadbw")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_sse41_mpsadbw,
NewFn);
if (Name == "x86.avx.blend.pd.256")
return UpgradeX86IntrinsicsWith8BitMask(
F, Intrinsic::x86_avx_blend_pd_256, NewFn);
if (Name == "x86.avx.blend.ps.256")
return UpgradeX86IntrinsicsWith8BitMask(
F, Intrinsic::x86_avx_blend_ps_256, NewFn);
if (Name == "x86.avx.dp.ps.256")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_avx_dp_ps_256,
NewFn);
if (Name == "x86.avx2.pblendw")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_avx2_pblendw,
NewFn);
if (Name == "x86.avx2.pblendd.128")
return UpgradeX86IntrinsicsWith8BitMask(
F, Intrinsic::x86_avx2_pblendd_128, NewFn);
if (Name == "x86.avx2.pblendd.256")
return UpgradeX86IntrinsicsWith8BitMask(
F, Intrinsic::x86_avx2_pblendd_256, NewFn);
if (Name == "x86.avx2.mpsadbw")
return UpgradeX86IntrinsicsWith8BitMask(F, Intrinsic::x86_avx2_mpsadbw,
NewFn);
// frcz.ss/sd may need to have an argument dropped
if (Name.startswith("x86.xop.vfrcz.ss") && F->arg_size() == 2) {
F->setName(Name + ".old");
@ -413,6 +474,34 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
CI->eraseFromParent();
return;
}
case Intrinsic::x86_sse41_pblendw:
case Intrinsic::x86_sse41_blendpd:
case Intrinsic::x86_sse41_blendps:
case Intrinsic::x86_sse41_insertps:
case Intrinsic::x86_sse41_dppd:
case Intrinsic::x86_sse41_dpps:
case Intrinsic::x86_sse41_mpsadbw:
case Intrinsic::x86_avx_blend_pd_256:
case Intrinsic::x86_avx_blend_ps_256:
case Intrinsic::x86_avx_dp_ps_256:
case Intrinsic::x86_avx2_pblendw:
case Intrinsic::x86_avx2_pblendd_128:
case Intrinsic::x86_avx2_pblendd_256:
case Intrinsic::x86_avx2_mpsadbw: {
// Need to truncate the last argument from i32 to i8 -- this argument models
// an inherently 8-bit immediate operand to these x86 instructions.
SmallVector<Value *, 4> Args(CI->arg_operands().begin(),
CI->arg_operands().end());
// Replace the last argument with a trunc.
Args.back() = Builder.CreateTrunc(Args.back(), Type::getInt8Ty(C), "trunc");
CallInst *NewCall = Builder.CreateCall(NewFn, Args);
CI->replaceAllUsesWith(NewCall);
CI->eraseFromParent();
return;
}
}
}

4
lib/Target/X86/AsmParser/X86AsmParserCommon.h
View File

@ -24,10 +24,6 @@ inline bool isImmSExti32i8Value(uint64_t Value) {
(0xFFFFFFFFFFFFFF80ULL <= Value && Value <= 0xFFFFFFFFFFFFFFFFULL));
}
inline bool isImmZExtu32u8Value(uint64_t Value) {
return (Value <= 0x00000000000000FFULL);
}
inline bool isImmSExti64i8Value(uint64_t Value) {
return (( Value <= 0x000000000000007FULL)||
(0xFFFFFFFFFFFFFF80ULL <= Value && Value <= 0xFFFFFFFFFFFFFFFFULL));

14
lib/Target/X86/AsmParser/X86Operand.h
View File

@ -153,20 +153,6 @@ struct X86Operand : public MCParsedAsmOperand {
// extension.
return isImmSExti32i8Value(CE->getValue());
}
bool isImmZExtu32u8() const {
if (!isImm())
return false;
// If this isn't a constant expr, just assume it fits and let relaxation
// handle it.
const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(getImm());
if (!CE)
return true;
// Otherwise, check the value is in a range that makes sense for this
// extension.
return isImmZExtu32u8Value(CE->getValue());
}
bool isImmSExti64i8() const {
if (!isImm())
return false;

8
lib/Target/X86/X86InstrAVX512.td
View File

@ -393,12 +393,12 @@ def : Pat<(vinsert256_insert:$ins (v16i32 VR512:$src1),
// vinsertps - insert f32 to XMM
def VINSERTPSzrr : AVX512AIi8<0x21, MRMSrcReg, (outs VR128X:$dst),
(ins VR128X:$src1, VR128X:$src2, u32u8imm:$src3),
(ins VR128X:$src1, VR128X:$src2, i8imm:$src3),
"vinsertps\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
[(set VR128X:$dst, (X86insertps VR128X:$src1, VR128X:$src2, imm:$src3))]>,
EVEX_4V;
def VINSERTPSzrm: AVX512AIi8<0x21, MRMSrcMem, (outs VR128X:$dst),
(ins VR128X:$src1, f32mem:$src2, u32u8imm:$src3),
(ins VR128X:$src1, f32mem:$src2, i8imm:$src3),
"vinsertps\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
[(set VR128X:$dst, (X86insertps VR128X:$src1,
(v4f32 (scalar_to_vector (loadf32 addr:$src2))),
@ -538,13 +538,13 @@ def : Pat<(insert_subvector undef, (v8f32 VR256X:$src), (iPTR 0)),
// vextractps - extract 32 bits from XMM
def VEXTRACTPSzrr : AVX512AIi8<0x17, MRMDestReg, (outs GR32:$dst),
(ins VR128X:$src1, u32u8imm:$src2),
(ins VR128X:$src1, i32i8imm:$src2),
"vextractps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set GR32:$dst, (extractelt (bc_v4i32 (v4f32 VR128X:$src1)), imm:$src2))]>,
EVEX;
def VEXTRACTPSzmr : AVX512AIi8<0x17, MRMDestMem, (outs),
(ins f32mem:$dst, VR128X:$src1, u32u8imm:$src2),
(ins f32mem:$dst, VR128X:$src1, i32i8imm:$src2),
"vextractps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(store (extractelt (bc_v4i32 (v4f32 VR128X:$src1)), imm:$src2),
addr:$dst)]>, EVEX, EVEX_CD8<32, CD8VT1>;

4
lib/Target/X86/X86InstrFragmentsSIMD.td
View File

@ -83,7 +83,7 @@ def X86pinsrw : SDNode<"X86ISD::PINSRW",
SDTCisVT<2, i32>, SDTCisPtrTy<3>]>>;
def X86insertps : SDNode<"X86ISD::INSERTPS",
SDTypeProfile<1, 3, [SDTCisVT<0, v4f32>, SDTCisSameAs<0,1>,
SDTCisVT<2, v4f32>, SDTCisPtrTy<3>]>>;
SDTCisVT<2, v4f32>, SDTCisVT<3, i8>]>>;
def X86vzmovl : SDNode<"X86ISD::VZEXT_MOVL",
SDTypeProfile<1, 1, [SDTCisSameAs<0,1>]>>;
@ -197,7 +197,7 @@ def SDTVBroadcast : SDTypeProfile<1, 1, [SDTCisVec<0>]>;
def SDTVBroadcastm : SDTypeProfile<1, 2, [SDTCisVec<0>, SDTCisVec<1>]>;
def SDTBlend : SDTypeProfile<1, 3, [SDTCisVec<0>, SDTCisSameAs<0,1>,
SDTCisSameAs<1,2>, SDTCisVT<3, i32>]>;
SDTCisSameAs<1,2>, SDTCisVT<3, i8>]>;
def SDTFma : SDTypeProfile<1, 3, [SDTCisSameAs<0,1>,
SDTCisSameAs<1,2>, SDTCisSameAs<1,3>]>;

17
lib/Target/X86/X86InstrInfo.td
View File

@ -551,11 +551,6 @@ class ImmSExtAsmOperandClass : AsmOperandClass {
let RenderMethod = "addImmOperands";
}
class ImmZExtAsmOperandClass : AsmOperandClass {
let SuperClasses = [ImmAsmOperand];
let RenderMethod = "addImmOperands";
}
def X86GR32orGR64AsmOperand : AsmOperandClass {
let Name = "GR32orGR64";
}
@ -568,6 +563,7 @@ def AVX512RC : Operand<i32> {
let PrintMethod = "printRoundingControl";
let OperandType = "OPERAND_IMMEDIATE";
}
// Sign-extended immediate classes. We don't need to define the full lattice
// here because there is no instruction with an ambiguity between ImmSExti64i32
// and ImmSExti32i8.
@ -595,12 +591,6 @@ def ImmSExti32i8AsmOperand : ImmSExtAsmOperandClass {
let Name = "ImmSExti32i8";
}
// [0, 0x000000FF]
def ImmZExtu32u8AsmOperand : ImmZExtAsmOperandClass {
let Name = "ImmZExtu32u8";
}
// [0, 0x0000007F] |
// [0xFFFFFFFFFFFFFF80, 0xFFFFFFFFFFFFFFFF]
def ImmSExti64i8AsmOperand : ImmSExtAsmOperandClass {
@ -620,11 +610,6 @@ def i32i8imm : Operand<i32> {
let ParserMatchClass = ImmSExti32i8AsmOperand;
let OperandType = "OPERAND_IMMEDIATE";
}
// 32-bits but only 8 bits are significant, and those 8 bits are unsigned.
def u32u8imm : Operand<i32> {
let ParserMatchClass = ImmZExtu32u8AsmOperand;
let OperandType = "OPERAND_IMMEDIATE";
}
// 64-bits but only 32 bits are significant.
def i64i32imm : Operand<i64> {

26
lib/Target/X86/X86InstrSSE.td
View File

@ -5379,7 +5379,7 @@ let Predicates = [HasAVX] in {
// the corresponding elements in the second input vector.
def : Pat<(v8f32 (X86Blendi (v8f32 (fsub VR256:$lhs, VR256:$rhs)),
(v8f32 (fadd VR256:$lhs, VR256:$rhs)), (i32 170))),
(v8f32 (fadd VR256:$lhs, VR256:$rhs)), (i8 170))),
(VADDSUBPSYrr VR256:$lhs, VR256:$rhs)>;
// Constant 10 corresponds to the binary mask '1010'.
@ -5388,16 +5388,16 @@ let Predicates = [HasAVX] in {
// - the 2nd and 4th element from the second input vector (the 'fadd' node).
def : Pat<(v4f64 (X86Blendi (v4f64 (fsub VR256:$lhs, VR256:$rhs)),
(v4f64 (fadd VR256:$lhs, VR256:$rhs)), (i32 10))),
(v4f64 (fadd VR256:$lhs, VR256:$rhs)), (i8 10))),
(VADDSUBPDYrr VR256:$lhs, VR256:$rhs)>;
def : Pat<(v4f64 (X86Blendi (v4f64 (fsub VR256:$lhs, VR256:$rhs)),
(v4f64 (fadd VR256:$lhs, VR256:$rhs)), (i32 10))),
(v4f64 (fadd VR256:$lhs, VR256:$rhs)), (i8 10))),
(VADDSUBPDYrr VR256:$lhs, VR256:$rhs)>;
def : Pat<(v4f32 (X86Blendi (v4f32 (fsub VR128:$lhs, VR128:$rhs)),
(v4f32 (fadd VR128:$lhs, VR128:$rhs)), (i32 10))),
(v4f32 (fadd VR128:$lhs, VR128:$rhs)), (i8 10))),
(VADDSUBPSrr VR128:$lhs, VR128:$rhs)>;
def : Pat<(v2f64 (X86Blendi (v2f64 (fsub VR128:$lhs, VR128:$rhs)),
(v2f64 (fadd VR128:$lhs, VR128:$rhs)), (i32 2))),
(v2f64 (fadd VR128:$lhs, VR128:$rhs)), (i8 2))),
(VADDSUBPDrr VR128:$lhs, VR128:$rhs)>;
def : Pat<(v2f64 (X86Movsd (v2f64 (fadd VR128:$lhs, VR128:$rhs)),
(v2f64 (fsub VR128:$lhs, VR128:$rhs)))),
@ -5411,11 +5411,11 @@ let Predicates = [UseSSE3] in {
// - the 2nd and 4th element from the second input vector (the fadd node).
def : Pat<(v4f32 (X86Blendi (v4f32 (fsub VR128:$lhs, VR128:$rhs)),
(v4f32 (fadd VR128:$lhs, VR128:$rhs)), (i32 10))),
(v4f32 (fadd VR128:$lhs, VR128:$rhs)), (i8 10))),
(ADDSUBPSrr VR128:$lhs, VR128:$rhs)>;
def : Pat<(v2f64 (X86Blendi (v2f64 (fsub VR128:$lhs, VR128:$rhs)),
(v2f64 (fadd VR128:$lhs, VR128:$rhs)), (i32 2))),
(v2f64 (fadd VR128:$lhs, VR128:$rhs)), (i8 2))),
(ADDSUBPDrr VR128:$lhs, VR128:$rhs)>;
def : Pat<(v2f64 (X86Movsd (v2f64 (fadd VR128:$lhs, VR128:$rhs)),
(v2f64 (fsub VR128:$lhs, VR128:$rhs)))),
@ -6705,7 +6705,7 @@ let Constraints = "$src1 = $dst" in
multiclass SS41I_insertf32<bits<8> opc, string asm, bit Is2Addr = 1,
OpndItins itins = DEFAULT_ITINS> {
def rr : SS4AIi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, u32u8imm:$src3),
(ins VR128:$src1, VR128:$src2, i8imm:$src3),
!if(Is2Addr,
!strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
!strconcat(asm,
@ -6714,7 +6714,7 @@ multiclass SS41I_insertf32<bits<8> opc, string asm, bit Is2Addr = 1,
(X86insertps VR128:$src1, VR128:$src2, imm:$src3))], itins.rr>,
Sched<[WriteFShuffle]>;
def rm : SS4AIi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, f32mem:$src2, u32u8imm:$src3),
(ins VR128:$src1, f32mem:$src2, i8imm:$src3),
!if(Is2Addr,
!strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
!strconcat(asm,
@ -7350,7 +7350,7 @@ multiclass SS41I_binop_rmi_int<bits<8> opc, string OpcodeStr,
OpndItins itins = DEFAULT_ITINS> {
let isCommutable = 1 in
def rri : SS4AIi8<opc, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2, u32u8imm:$src3),
(ins RC:$src1, RC:$src2, i8imm:$src3),
!if(Is2Addr,
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
@ -7359,7 +7359,7 @@ multiclass SS41I_binop_rmi_int<bits<8> opc, string OpcodeStr,
[(set RC:$dst, (IntId RC:$src1, RC:$src2, imm:$src3))], itins.rr>,
Sched<[itins.Sched]>;
def rmi : SS4AIi8<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, x86memop:$src2, u32u8imm:$src3),
(ins RC:$src1, x86memop:$src2, i8imm:$src3),
!if(Is2Addr,
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
@ -8579,13 +8579,13 @@ multiclass AVX2_binop_rmi_int<bits<8> opc, string OpcodeStr,
X86MemOperand x86memop> {
let isCommutable = 1 in
def rri : AVX2AIi8<opc, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2, u32u8imm:$src3),
(ins RC:$src1, RC:$src2, i8imm:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set RC:$dst, (IntId RC:$src1, RC:$src2, imm:$src3))]>,
Sched<[WriteBlend]>, VEX_4V;
def rmi : AVX2AIi8<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, x86memop:$src2, u32u8imm:$src3),
(ins RC:$src1, x86memop:$src2, i8imm:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set RC:$dst,

26
test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll Normal file
View File

@ -0,0 +1,26 @@
; RUN: llc < %s -mtriple=x86_64-apple-darwin -march=x86 -mcpu=corei7-avx | FileCheck %s
define <4 x double> @test_x86_avx_blend_pd_256(<4 x double> %a0, <4 x double> %a1) {
; CHECK: vblendpd
%res = call <4 x double> @llvm.x86.avx.blend.pd.256(<4 x double> %a0, <4 x double> %a1, i32 7) ; <<4 x double>> [#uses=1]
ret <4 x double> %res
}
declare <4 x double> @llvm.x86.avx.blend.pd.256(<4 x double>, <4 x double>, i32) nounwind readnone
define <8 x float> @test_x86_avx_blend_ps_256(<8 x float> %a0, <8 x float> %a1) {
; CHECK: vblendps
%res = call <8 x float> @llvm.x86.avx.blend.ps.256(<8 x float> %a0, <8 x float> %a1, i32 7) ; <<8 x float>> [#uses=1]
ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx.blend.ps.256(<8 x float>, <8 x float>, i32) nounwind readnone
define <8 x float> @test_x86_avx_dp_ps_256(<8 x float> %a0, <8 x float> %a1) {
; CHECK: vdpps
%res = call <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float> %a0, <8 x float> %a1, i32 7) ; <<8 x float>> [#uses=1]
ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float>, <8 x float>, i32) nounwind readnone

40
test/CodeGen/X86/avx-intrinsics-x86.ll
View File

@ -818,18 +818,18 @@ declare <16 x i8> @llvm.x86.sse3.ldu.dq(i8*) nounwind readonly
define <2 x double> @test_x86_sse41_blendpd(<2 x double> %a0, <2 x double> %a1) {
; CHECK: vblendpd
%res = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
%res = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a1, i8 7) ; <<2 x double>> [#uses=1]
ret <2 x double> %res
}
declare <2 x double> @llvm.x86.sse41.blendpd(<2 x double>, <2 x double>, i32) nounwind readnone
declare <2 x double> @llvm.x86.sse41.blendpd(<2 x double>, <2 x double>, i8) nounwind readnone
define <4 x float> @test_x86_sse41_blendps(<4 x float> %a0, <4 x float> %a1) {
; CHECK: vblendps
%res = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
%res = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.sse41.blendps(<4 x float>, <4 x float>, i32) nounwind readnone
declare <4 x float> @llvm.x86.sse41.blendps(<4 x float>, <4 x float>, i8) nounwind readnone
define <2 x double> @test_x86_sse41_blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
@ -850,35 +850,35 @@ declare <4 x float> @llvm.x86.sse41.blendvps(<4 x float>, <4 x float>, <4 x floa
define <2 x double> @test_x86_sse41_dppd(<2 x double> %a0, <2 x double> %a1) {
; CHECK: vdppd
%res = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
%res = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %a0, <2 x double> %a1, i8 7) ; <<2 x double>> [#uses=1]
ret <2 x double> %res
}
declare <2 x double> @llvm.x86.sse41.dppd(<2 x double>, <2 x double>, i32) nounwind readnone
declare <2 x double> @llvm.x86.sse41.dppd(<2 x double>, <2 x double>, i8) nounwind readnone
define <4 x float> @test_x86_sse41_dpps(<4 x float> %a0, <4 x float> %a1) {
; CHECK: vdpps
%res = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
%res = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.sse41.dpps(<4 x float>, <4 x float>, i32) nounwind readnone
declare <4 x float> @llvm.x86.sse41.dpps(<4 x float>, <4 x float>, i8) nounwind readnone
define <4 x float> @test_x86_sse41_insertps(<4 x float> %a0, <4 x float> %a1) {
; CHECK: vinsertps
%res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
%res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i32) nounwind readnone
declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i8) nounwind readnone
define <8 x i16> @test_x86_sse41_mpsadbw(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK: vmpsadbw
%res = call <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8> %a0, <16 x i8> %a1, i32 7) ; <<8 x i16>> [#uses=1]
%res = call <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <<8 x i16>> [#uses=1]
ret <8 x i16> %res
}
declare <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8>, <16 x i8>, i32) nounwind readnone
declare <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8>, <16 x i8>, i8) nounwind readnone
define <8 x i16> @test_x86_sse41_packusdw(<4 x i32> %a0, <4 x i32> %a1) {
@ -899,10 +899,10 @@ declare <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8>, <16 x i8>, <16 x i8>) noun
define <8 x i16> @test_x86_sse41_pblendw(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK: vpblendw
%res = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a1, i32 7) ; <<8 x i16>> [#uses=1]
%res = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a1, i8 7) ; <<8 x i16>> [#uses=1]
ret <8 x i16> %res
}
declare <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16>, <8 x i16>, i32) nounwind readnone
declare <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16>, <8 x i16>, i8) nounwind readnone
define <8 x i16> @test_x86_sse41_phminposuw(<8 x i16> %a0) {
@ -1770,18 +1770,18 @@ declare <8 x float> @llvm.x86.avx.addsub.ps.256(<8 x float>, <8 x float>) nounwi
define <4 x double> @test_x86_avx_blend_pd_256(<4 x double> %a0, <4 x double> %a1) {
; CHECK: vblendpd
%res = call <4 x double> @llvm.x86.avx.blend.pd.256(<4 x double> %a0, <4 x double> %a1, i32 7) ; <<4 x double>> [#uses=1]
%res = call <4 x double> @llvm.x86.avx.blend.pd.256(<4 x double> %a0, <4 x double> %a1, i8 7) ; <<4 x double>> [#uses=1]
ret <4 x double> %res
}
declare <4 x double> @llvm.x86.avx.blend.pd.256(<4 x double>, <4 x double>, i32) nounwind readnone
declare <4 x double> @llvm.x86.avx.blend.pd.256(<4 x double>, <4 x double>, i8) nounwind readnone
define <8 x float> @test_x86_avx_blend_ps_256(<8 x float> %a0, <8 x float> %a1) {
; CHECK: vblendps
%res = call <8 x float> @llvm.x86.avx.blend.ps.256(<8 x float> %a0, <8 x float> %a1, i32 7) ; <<8 x float>> [#uses=1]
%res = call <8 x float> @llvm.x86.avx.blend.ps.256(<8 x float> %a0, <8 x float> %a1, i8 7) ; <<8 x float>> [#uses=1]
ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx.blend.ps.256(<8 x float>, <8 x float>, i32) nounwind readnone
declare <8 x float> @llvm.x86.avx.blend.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone
define <4 x double> @test_x86_avx_blendv_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) {
@ -1950,10 +1950,10 @@ declare <8 x i32> @llvm.x86.avx.cvtt.ps2dq.256(<8 x float>) nounwind readnone
define <8 x float> @test_x86_avx_dp_ps_256(<8 x float> %a0, <8 x float> %a1) {
; CHECK: vdpps
%res = call <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float> %a0, <8 x float> %a1, i32 7) ; <<8 x float>> [#uses=1]
%res = call <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float> %a0, <8 x float> %a1, i8 7) ; <<8 x float>> [#uses=1]
ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float>, <8 x float>, i32) nounwind readnone
declare <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone
define <4 x double> @test_x86_avx_hadd_pd_256(<4 x double> %a0, <4 x double> %a1) {

2
test/CodeGen/X86/avx.ll
View File

@ -60,7 +60,7 @@ define <4 x float> @insertps_from_vector_load_offset_2(<4 x float> %a, <4 x floa
; X32: movl 8(%esp), %ecx
; CHECK-NOT: mov
;; Try to match a bit more of the instr, since we need the load's offset.
; CHECK: vinsertps $192, 12(%{{...}},%{{...}}), %
; CHECK: vinsertps $-64, 12(%{{...}},%{{...}}), %
; CHECK-NEXT: ret
%1 = getelementptr inbounds <4 x float>* %pb, i64 %index
%2 = load <4 x float>* %1, align 16

33
test/CodeGen/X86/avx2-intrinsics-x86-upgrade.ll Normal file
View File

@ -0,0 +1,33 @@
; RUN: llc < %s -mtriple=x86_64-apple-darwin -march=x86 -mcpu=core-avx2 -mattr=avx2 | FileCheck %s
define <16 x i16> @test_x86_avx2_pblendw(<16 x i16> %a0, <16 x i16> %a1) {
; CHECK: vpblendw
%res = call <16 x i16> @llvm.x86.avx2.pblendw(<16 x i16> %a0, <16 x i16> %a1, i32 7) ; <<16 x i16>> [#uses=1]
ret <16 x i16> %res
}
declare <16 x i16> @llvm.x86.avx2.pblendw(<16 x i16>, <16 x i16>, i32) nounwind readnone
define <4 x i32> @test_x86_avx2_pblendd_128(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK: vpblendd
%res = call <4 x i32> @llvm.x86.avx2.pblendd.128(<4 x i32> %a0, <4 x i32> %a1, i32 7) ; <<4 x i32>> [#uses=1]
ret <4 x i32> %res
}
declare <4 x i32> @llvm.x86.avx2.pblendd.128(<4 x i32>, <4 x i32>, i32) nounwind readnone
define <8 x i32> @test_x86_avx2_pblendd_256(<8 x i32> %a0, <8 x i32> %a1) {
; CHECK: vpblendd
%res = call <8 x i32> @llvm.x86.avx2.pblendd.256(<8 x i32> %a0, <8 x i32> %a1, i32 7) ; <<8 x i32>> [#uses=1]
ret <8 x i32> %res
}
declare <8 x i32> @llvm.x86.avx2.pblendd.256(<8 x i32>, <8 x i32>, i32) nounwind readnone
define <16 x i16> @test_x86_avx2_mpsadbw(<32 x i8> %a0, <32 x i8> %a1) {
; CHECK: vmpsadbw
%res = call <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8> %a0, <32 x i8> %a1, i32 7) ; <<16 x i16>> [#uses=1]
ret <16 x i16> %res
}
declare <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8>, <32 x i8>, i32) nounwind readnone

16
test/CodeGen/X86/avx2-intrinsics-x86.ll
View File

@ -475,10 +475,10 @@ declare <4 x i64> @llvm.x86.avx2.movntdqa(i8*) nounwind readonly
define <16 x i16> @test_x86_avx2_mpsadbw(<32 x i8> %a0, <32 x i8> %a1) {
; CHECK: vmpsadbw
%res = call <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8> %a0, <32 x i8> %a1, i32 7) ; <<16 x i16>> [#uses=1]
%res = call <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8> %a0, <32 x i8> %a1, i8 7) ; <<16 x i16>> [#uses=1]
ret <16 x i16> %res
}
declare <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8>, <32 x i8>, i32) nounwind readnone
declare <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8>, <32 x i8>, i8) nounwind readnone
define <16 x i16> @test_x86_avx2_packusdw(<8 x i32> %a0, <8 x i32> %a1) {
@ -499,10 +499,10 @@ declare <32 x i8> @llvm.x86.avx2.pblendvb(<32 x i8>, <32 x i8>, <32 x i8>) nounw
define <16 x i16> @test_x86_avx2_pblendw(<16 x i16> %a0, <16 x i16> %a1) {
; CHECK: vpblendw
%res = call <16 x i16> @llvm.x86.avx2.pblendw(<16 x i16> %a0, <16 x i16> %a1, i32 7) ; <<16 x i16>> [#uses=1]
%res = call <16 x i16> @llvm.x86.avx2.pblendw(<16 x i16> %a0, <16 x i16> %a1, i8 7) ; <<16 x i16>> [#uses=1]
ret <16 x i16> %res
}
declare <16 x i16> @llvm.x86.avx2.pblendw(<16 x i16>, <16 x i16>, i32) nounwind readnone
declare <16 x i16> @llvm.x86.avx2.pblendw(<16 x i16>, <16 x i16>, i8) nounwind readnone
define <32 x i8> @test_x86_avx2_pmaxsb(<32 x i8> %a0, <32 x i8> %a1) {
@ -706,18 +706,18 @@ declare <8 x float> @llvm.x86.avx2.vbroadcast.ss.ps.256(<4 x float>) nounwind re
define <4 x i32> @test_x86_avx2_pblendd_128(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK: vpblendd
%res = call <4 x i32> @llvm.x86.avx2.pblendd.128(<4 x i32> %a0, <4 x i32> %a1, i32 7) ; <<4 x i32>> [#uses=1]
%res = call <4 x i32> @llvm.x86.avx2.pblendd.128(<4 x i32> %a0, <4 x i32> %a1, i8 7) ; <<4 x i32>> [#uses=1]
ret <4 x i32> %res
}
declare <4 x i32> @llvm.x86.avx2.pblendd.128(<4 x i32>, <4 x i32>, i32) nounwind readnone
declare <4 x i32> @llvm.x86.avx2.pblendd.128(<4 x i32>, <4 x i32>, i8) nounwind readnone
define <8 x i32> @test_x86_avx2_pblendd_256(<8 x i32> %a0, <8 x i32> %a1) {
; CHECK: vpblendd
%res = call <8 x i32> @llvm.x86.avx2.pblendd.256(<8 x i32> %a0, <8 x i32> %a1, i32 7) ; <<8 x i32>> [#uses=1]
%res = call <8 x i32> @llvm.x86.avx2.pblendd.256(<8 x i32> %a0, <8 x i32> %a1, i8 7) ; <<8 x i32>> [#uses=1]
ret <8 x i32> %res
}
declare <8 x i32> @llvm.x86.avx2.pblendd.256(<8 x i32>, <8 x i32>, i32) nounwind readnone
declare <8 x i32> @llvm.x86.avx2.pblendd.256(<8 x i32>, <8 x i32>, i8) nounwind readnone
define <16 x i8> @test_x86_avx2_pbroadcastb_128(<16 x i8> %a0) {

61
test/CodeGen/X86/sse41-intrinsics-x86-upgrade.ll Normal file
View File

@ -0,0 +1,61 @@
; RUN: llc < %s -mtriple=i386-apple-darwin -mattr=-avx,+sse4.1 | FileCheck %s
; This test works just like the non-upgrade one except that it only checks
; forms which require auto-upgrading.
define <2 x double> @test_x86_sse41_blendpd(<2 x double> %a0, <2 x double> %a1) {
; CHECK: blendpd
%res = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
ret <2 x double> %res
}
declare <2 x double> @llvm.x86.sse41.blendpd(<2 x double>, <2 x double>, i32) nounwind readnone
define <4 x float> @test_x86_sse41_blendps(<4 x float> %a0, <4 x float> %a1) {
; CHECK: blendps
%res = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.sse41.blendps(<4 x float>, <4 x float>, i32) nounwind readnone
define <2 x double> @test_x86_sse41_dppd(<2 x double> %a0, <2 x double> %a1) {
; CHECK: dppd
%res = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
ret <2 x double> %res
}
declare <2 x double> @llvm.x86.sse41.dppd(<2 x double>, <2 x double>, i32) nounwind readnone
define <4 x float> @test_x86_sse41_dpps(<4 x float> %a0, <4 x float> %a1) {
; CHECK: dpps
%res = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.sse41.dpps(<4 x float>, <4 x float>, i32) nounwind readnone
define <4 x float> @test_x86_sse41_insertps(<4 x float> %a0, <4 x float> %a1) {
; CHECK: insertps
%res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i32) nounwind readnone
define <8 x i16> @test_x86_sse41_mpsadbw(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK: mpsadbw
%res = call <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8> %a0, <16 x i8> %a1, i32 7) ; <<8 x i16>> [#uses=1]
ret <8 x i16> %res
}
declare <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8>, <16 x i8>, i32) nounwind readnone
define <8 x i16> @test_x86_sse41_pblendw(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK: pblendw
%res = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a1, i32 7) ; <<8 x i16>> [#uses=1]
ret <8 x i16> %res
}
declare <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16>, <8 x i16>, i32) nounwind readnone

28
test/CodeGen/X86/sse41-intrinsics-x86.ll
View File

@ -2,18 +2,18 @@
define <2 x double> @test_x86_sse41_blendpd(<2 x double> %a0, <2 x double> %a1) {
; CHECK: blendpd
%res = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
%res = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a1, i8 7) ; <<2 x double>> [#uses=1]
ret <2 x double> %res
}
declare <2 x double> @llvm.x86.sse41.blendpd(<2 x double>, <2 x double>, i32) nounwind readnone
declare <2 x double> @llvm.x86.sse41.blendpd(<2 x double>, <2 x double>, i8) nounwind readnone
define <4 x float> @test_x86_sse41_blendps(<4 x float> %a0, <4 x float> %a1) {
; CHECK: blendps
%res = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
%res = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.sse41.blendps(<4 x float>, <4 x float>, i32) nounwind readnone
declare <4 x float> @llvm.x86.sse41.blendps(<4 x float>, <4 x float>, i8) nounwind readnone
define <2 x double> @test_x86_sse41_blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
@ -34,35 +34,35 @@ declare <4 x float> @llvm.x86.sse41.blendvps(<4 x float>, <4 x float>, <4 x floa
define <2 x double> @test_x86_sse41_dppd(<2 x double> %a0, <2 x double> %a1) {
; CHECK: dppd
%res = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
%res = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %a0, <2 x double> %a1, i8 7) ; <<2 x double>> [#uses=1]
ret <2 x double> %res
}
declare <2 x double> @llvm.x86.sse41.dppd(<2 x double>, <2 x double>, i32) nounwind readnone
declare <2 x double> @llvm.x86.sse41.dppd(<2 x double>, <2 x double>, i8) nounwind readnone
define <4 x float> @test_x86_sse41_dpps(<4 x float> %a0, <4 x float> %a1) {
; CHECK: dpps
%res = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
%res = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.sse41.dpps(<4 x float>, <4 x float>, i32) nounwind readnone
declare <4 x float> @llvm.x86.sse41.dpps(<4 x float>, <4 x float>, i8) nounwind readnone
define <4 x float> @test_x86_sse41_insertps(<4 x float> %a0, <4 x float> %a1) {
; CHECK: insertps
%res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
%res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i32) nounwind readnone
declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i8) nounwind readnone
define <8 x i16> @test_x86_sse41_mpsadbw(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK: mpsadbw
%res = call <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8> %a0, <16 x i8> %a1, i32 7) ; <<8 x i16>> [#uses=1]
%res = call <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <<8 x i16>> [#uses=1]
ret <8 x i16> %res
}
declare <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8>, <16 x i8>, i32) nounwind readnone
declare <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8>, <16 x i8>, i8) nounwind readnone
define <8 x i16> @test_x86_sse41_packusdw(<4 x i32> %a0, <4 x i32> %a1) {
@ -83,10 +83,10 @@ declare <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8>, <16 x i8>, <16 x i8>) noun
define <8 x i16> @test_x86_sse41_pblendw(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK: pblendw
%res = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a1, i32 7) ; <<8 x i16>> [#uses=1]
%res = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a1, i8 7) ; <<8 x i16>> [#uses=1]
ret <8 x i16> %res
}
declare <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16>, <8 x i16>, i32) nounwind readnone
declare <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16>, <8 x i16>, i8) nounwind readnone
define <8 x i16> @test_x86_sse41_phminposuw(<8 x i16> %a0) {

2
test/CodeGen/X86/sse41.ll
View File

@ -618,7 +618,7 @@ define <4 x float> @insertps_from_vector_load_offset_2(<4 x float> %a, <4 x floa
; X32: movl 8(%esp), %ecx
; CHECK-NOT: mov
;; Try to match a bit more of the instr, since we need the load's offset.
; CHECK: insertps $192, 12(%{{...}},%{{...}}), %
; CHECK: insertps $-64, 12(%{{...}},%{{...}}), %
; CHECK-NEXT: ret
%1 = getelementptr inbounds <4 x float>* %pb, i64 %index
%2 = load <4 x float>* %1, align 16

18
test/MC/X86/x86-32-coverage.s
View File

@ -19618,22 +19618,36 @@
// CHECK: blendvps %xmm2, %xmm1 # encoding: [0x66,0x0f,0x38,0x14,0xca]
blendvps %xmm2, %xmm1
// rdar://9795008
// These instructions take a mask not an 8-bit sign extended value.
// These instructions can take an unsigned 8-bit mask as well as a signed 8-bit
// immediate. Check both forms here.
// CHECK: blendps $129, %xmm2, %xmm1
blendps $0x81, %xmm2, %xmm1
// CHECK: blendps $-64, %xmm2, %xmm1
blendps $-64, %xmm2, %xmm1
// CHECK: blendpd $129, %xmm2, %xmm1
blendpd $0x81, %xmm2, %xmm1
// CHECK: blendpd $-64, %xmm2, %xmm1
blendpd $-64, %xmm2, %xmm1
// CHECK: pblendw $129, %xmm2, %xmm1
pblendw $0x81, %xmm2, %xmm1
// CHECK: pblendw $-64, %xmm2, %xmm1
pblendw $-64, %xmm2, %xmm1
// CHECK: mpsadbw $129, %xmm2, %xmm1
mpsadbw $0x81, %xmm2, %xmm1
// CHECK: mpsadbw $-64, %xmm2, %xmm1
mpsadbw $-64, %xmm2, %xmm1
// CHECK: dpps $129, %xmm2, %xmm1
dpps $0x81, %xmm2, %xmm1
// CHECK: dpps $-64, %xmm2, %xmm1
dpps $-64, %xmm2, %xmm1
// CHECK: dppd $129, %xmm2, %xmm1
dppd $0x81, %xmm2, %xmm1
// CHECK: dppd $-64, %xmm2, %xmm1
dppd $-64, %xmm2, %xmm1
// CHECK: insertps $129, %xmm2, %xmm1
insertps $0x81, %xmm2, %xmm1
// CHECK: insertps $-64, %xmm2, %xmm1
insertps $-64, %xmm2, %xmm1
// PR13253 handle implicit optional third argument that must always be xmm0
// CHECK: pblendvb %xmm2, %xmm1

2
utils/TableGen/X86RecognizableInstr.cpp
View File

@ -912,7 +912,6 @@ OperandType RecognizableInstr::typeFromString(const std::string &s,
TYPE("i32mem", TYPE_Mv)
TYPE("i32imm", TYPE_IMMv)
TYPE("i32i8imm", TYPE_IMM32)
TYPE("u32u8imm", TYPE_IMM32)
TYPE("GR32", TYPE_R32)
TYPE("GR32orGR64", TYPE_R32)
TYPE("i64mem", TYPE_Mv)
@ -1015,7 +1014,6 @@ RecognizableInstr::immediateEncodingFromString(const std::string &s,
ENCODING("i16imm", ENCODING_IW)
}
ENCODING("i32i8imm", ENCODING_IB)
ENCODING("u32u8imm", ENCODING_IB)
ENCODING("SSECC", ENCODING_IB)
ENCODING("AVXCC", ENCODING_IB)
ENCODING("AVX512RC", ENCODING_IB)