[llvm] [X86] movsd/movss/movd/movq - add support for constant comments (PR #78601)
via llvm-commits
llvm-commits at lists.llvm.org
Thu Jan 18 07:42:31 PST 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-x86
Author: Simon Pilgrim (RKSimon)
<details>
<summary>Changes</summary>
If we're loading a constant value, print the constant (and the zero upper elements) instead of just the shuffle mask
This did require me to move the shuffle mask handling into addConstantComments as we can't handle this in the MC layer
---
Patch is 668.38 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/78601.diff
109 Files Affected:
- (modified) llvm/lib/Target/X86/MCTargetDesc/X86InstComments.cpp (+2-30)
- (modified) llvm/lib/Target/X86/X86MCInstLower.cpp (+90)
- (modified) llvm/test/CodeGen/X86/2008-09-25-sseregparm-1.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/GlobalISel/fconstant.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/asm-reg-type-mismatch-avx512.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/atomic-fp.ll (+18-18)
- (modified) llvm/test/CodeGen/X86/avx512-cmp.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/avx512-fma-intrinsics.ll (+6-13)
- (modified) llvm/test/CodeGen/X86/avx512-insert-extract.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll (+24-24)
- (modified) llvm/test/CodeGen/X86/avx512-mov.ll (+12-12)
- (modified) llvm/test/CodeGen/X86/avx512-regcall-NoMask.ll (+6-6)
- (modified) llvm/test/CodeGen/X86/avx512-vec-cmp.ll (+6-6)
- (modified) llvm/test/CodeGen/X86/bc-extract.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/bfloat.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/buildvec-insertvec.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/cmov-fp.ll (+24-24)
- (modified) llvm/test/CodeGen/X86/cmovcmov.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/combine-fabs.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/combineIncDecVector-crash.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/cvtv2f32.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/dagcombine-select.ll (+6-6)
- (modified) llvm/test/CodeGen/X86/deopt-intrinsic.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/extract-fp.ll (+6-6)
- (modified) llvm/test/CodeGen/X86/extractelement-fp.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/extractelement-load.ll (+4-4)
- (modified) llvm/test/CodeGen/X86/fadd-combines.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/fast-isel-constpool.ll (+16-16)
- (modified) llvm/test/CodeGen/X86/fdiv-combine-vec.ll (+12-12)
- (modified) llvm/test/CodeGen/X86/fdiv-combine.ll (+6-6)
- (modified) llvm/test/CodeGen/X86/fma-intrinsics-canonical.ll (+16-16)
- (modified) llvm/test/CodeGen/X86/fma-intrinsics-x86-upgrade.ll (+32-32)
- (modified) llvm/test/CodeGen/X86/fma-intrinsics-x86.ll (+34-34)
- (modified) llvm/test/CodeGen/X86/fma-scalar-memfold.ll (+64-64)
- (modified) llvm/test/CodeGen/X86/fma.ll (+66-66)
- (modified) llvm/test/CodeGen/X86/fma_patterns.ll (+6-6)
- (modified) llvm/test/CodeGen/X86/fmf-flags.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/fminimum-fmaximum.ll (+12-12)
- (modified) llvm/test/CodeGen/X86/fold-int-pow2-with-fmul-or-fdiv.ll (+56-56)
- (modified) llvm/test/CodeGen/X86/fp-intrinsics-fma.ll (+12-12)
- (modified) llvm/test/CodeGen/X86/fp-intrinsics.ll (+58-58)
- (modified) llvm/test/CodeGen/X86/fp-logic.ll (+6-6)
- (modified) llvm/test/CodeGen/X86/fp-strict-scalar-fptoint-fp16.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/fp-strict-scalar-fptoint.ll (+12-12)
- (modified) llvm/test/CodeGen/X86/fp-undef.ll (+25-25)
- (modified) llvm/test/CodeGen/X86/fpclamptosat.ll (+8-8)
- (modified) llvm/test/CodeGen/X86/fpclamptosat_vec.ll (+10-10)
- (modified) llvm/test/CodeGen/X86/fptosi-sat-scalar.ll (+28-28)
- (modified) llvm/test/CodeGen/X86/fptosi-sat-vector-128.ll (+15-15)
- (modified) llvm/test/CodeGen/X86/fptoui-sat-scalar.ll (+23-23)
- (modified) llvm/test/CodeGen/X86/fptoui-sat-vector-128.ll (+12-12)
- (modified) llvm/test/CodeGen/X86/ftrunc.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/half.ll (+7-7)
- (modified) llvm/test/CodeGen/X86/insertelement-ones.ll (+3-3)
- (modified) llvm/test/CodeGen/X86/ldexp.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/load-scalar-as-vector.ll (+8-8)
- (modified) llvm/test/CodeGen/X86/logical-load-fold.ll (+4-4)
- (modified) llvm/test/CodeGen/X86/lsr-static-addr.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/machine-trace-metrics-crash.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/masked-iv-safe.ll (+24-24)
- (modified) llvm/test/CodeGen/X86/masked-iv-unsafe.ll (+39-39)
- (modified) llvm/test/CodeGen/X86/merge-consecutive-loads-128.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/neg_fp.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/nontemporal-4.ll (+3-3)
- (modified) llvm/test/CodeGen/X86/oss-fuzz-25184.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/peep-test-0.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/pow.ll (+10-10)
- (modified) llvm/test/CodeGen/X86/powi-int32min.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/pr23103.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/pr37879.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/pr40539.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/pr44749.ll (+3-3)
- (modified) llvm/test/CodeGen/X86/pr59258.ll (+8-8)
- (modified) llvm/test/CodeGen/X86/pr59305.ll (+3-3)
- (modified) llvm/test/CodeGen/X86/pseudo_cmov_lower2.ll (+4-4)
- (modified) llvm/test/CodeGen/X86/recip-fastmath.ll (+16-16)
- (modified) llvm/test/CodeGen/X86/recip-fastmath2.ll (+32-32)
- (modified) llvm/test/CodeGen/X86/recip-pic.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/scalar-fp-to-i64.ll (+12-12)
- (modified) llvm/test/CodeGen/X86/scalarize-fp.ll (+24-24)
- (modified) llvm/test/CodeGen/X86/select-of-fp-constants.ll (+3-3)
- (modified) llvm/test/CodeGen/X86/select.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/select_const.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/setcc-combine.ll (+1-1)
- (modified) llvm/test/CodeGen/X86/sqrt-fastmath-tune.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/sqrt-fastmath-tunecpu-attr.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/sqrt-fastmath.ll (+6-6)
- (modified) llvm/test/CodeGen/X86/sse-fcopysign.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/sse-intrinsics-fast-isel.ll (+92-92)
- (modified) llvm/test/CodeGen/X86/sse-load-ret.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/sse-minmax.ll (+33-33)
- (modified) llvm/test/CodeGen/X86/sse1.ll (+8-8)
- (modified) llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll (+192-192)
- (modified) llvm/test/CodeGen/X86/sse2-intrinsics-x86-upgrade.ll (+12-12)
- (modified) llvm/test/CodeGen/X86/sse3-avx-addsub-2.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/sse41.ll (+28-28)
- (modified) llvm/test/CodeGen/X86/stack-folding-int-avx2.ll (+3-3)
- (modified) llvm/test/CodeGen/X86/swifterror.ll (+16-16)
- (modified) llvm/test/CodeGen/X86/vec-strict-fptoint-128.ll (+28-28)
- (modified) llvm/test/CodeGen/X86/vec-strict-fptoint-256.ll (+8-8)
- (modified) llvm/test/CodeGen/X86/vec-strict-fptoint-512.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/vec_fp_to_int.ll (+17-17)
- (modified) llvm/test/CodeGen/X86/vector-constrained-fp-intrinsics-fma.ll (+10-10)
- (modified) llvm/test/CodeGen/X86/vector-constrained-fp-intrinsics.ll (+625-625)
- (modified) llvm/test/CodeGen/X86/vector-reduce-fadd.ll (+3-3)
- (modified) llvm/test/CodeGen/X86/vector-shuffle-combining.ll (+4-4)
- (modified) llvm/test/CodeGen/X86/vselect-zero.ll (+8-8)
- (modified) llvm/test/CodeGen/X86/widen-load-of-small-alloca.ll (+2-2)
- (modified) llvm/test/CodeGen/X86/x86-64-varargs.ll (+2-2)
``````````diff
diff --git a/llvm/lib/Target/X86/MCTargetDesc/X86InstComments.cpp b/llvm/lib/Target/X86/MCTargetDesc/X86InstComments.cpp
index 20b37d5a99902be..619328af12719d1 100644
--- a/llvm/lib/Target/X86/MCTargetDesc/X86InstComments.cpp
+++ b/llvm/lib/Target/X86/MCTargetDesc/X86InstComments.cpp
@@ -1212,15 +1212,7 @@ bool llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
case X86::VMOVSDZrr:
Src2Name = getRegName(MI->getOperand(2).getReg());
Src1Name = getRegName(MI->getOperand(1).getReg());
- [[fallthrough]];
-
- case X86::MOVSDrm_alt:
- case X86::MOVSDrm:
- case X86::VMOVSDrm_alt:
- case X86::VMOVSDrm:
- case X86::VMOVSDZrm:
- case X86::VMOVSDZrm_alt:
- DecodeScalarMoveMask(2, nullptr == Src2Name, ShuffleMask);
+ DecodeScalarMoveMask(2, false, ShuffleMask);
DestName = getRegName(MI->getOperand(0).getReg());
break;
@@ -1229,15 +1221,7 @@ bool llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
case X86::VMOVSSZrr:
Src2Name = getRegName(MI->getOperand(2).getReg());
Src1Name = getRegName(MI->getOperand(1).getReg());
- [[fallthrough]];
-
- case X86::MOVSSrm:
- case X86::MOVSSrm_alt:
- case X86::VMOVSSrm:
- case X86::VMOVSSrm_alt:
- case X86::VMOVSSZrm:
- case X86::VMOVSSZrm_alt:
- DecodeScalarMoveMask(4, nullptr == Src2Name, ShuffleMask);
+ DecodeScalarMoveMask(4, false, ShuffleMask);
DestName = getRegName(MI->getOperand(0).getReg());
break;
@@ -1248,22 +1232,10 @@ bool llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
case X86::VMOVZPQILo2PQIrr:
case X86::VMOVZPQILo2PQIZrr:
Src1Name = getRegName(MI->getOperand(1).getReg());
- [[fallthrough]];
-
- case X86::MOVQI2PQIrm:
- case X86::VMOVQI2PQIrm:
- case X86::VMOVQI2PQIZrm:
DecodeZeroMoveLowMask(2, ShuffleMask);
DestName = getRegName(MI->getOperand(0).getReg());
break;
- case X86::MOVDI2PDIrm:
- case X86::VMOVDI2PDIrm:
- case X86::VMOVDI2PDIZrm:
- DecodeZeroMoveLowMask(4, ShuffleMask);
- DestName = getRegName(MI->getOperand(0).getReg());
- break;
-
case X86::EXTRQI:
if (MI->getOperand(2).isImm() &&
MI->getOperand(3).isImm())
diff --git a/llvm/lib/Target/X86/X86MCInstLower.cpp b/llvm/lib/Target/X86/X86MCInstLower.cpp
index 2d5ccbfdfc765fa..cc615a3fac34fd3 100644
--- a/llvm/lib/Target/X86/X86MCInstLower.cpp
+++ b/llvm/lib/Target/X86/X86MCInstLower.cpp
@@ -1806,6 +1806,96 @@ static void addConstantComments(const MachineInstr *MI,
break;
}
+ case X86::MOVSDrm:
+ case X86::MOVSSrm:
+ case X86::VMOVSDrm:
+ case X86::VMOVSSrm:
+ case X86::VMOVSDZrm:
+ case X86::VMOVSSZrm:
+ case X86::MOVSDrm_alt:
+ case X86::MOVSSrm_alt:
+ case X86::VMOVSDrm_alt:
+ case X86::VMOVSSrm_alt:
+ case X86::VMOVSDZrm_alt:
+ case X86::VMOVSSZrm_alt:
+ case X86::MOVDI2PDIrm:
+ case X86::MOVQI2PQIrm:
+ case X86::VMOVDI2PDIrm:
+ case X86::VMOVQI2PQIrm:
+ case X86::VMOVDI2PDIZrm:
+ case X86::VMOVQI2PQIZrm: {
+ assert(MI->getNumOperands() >= (1 + X86::AddrNumOperands) &&
+ "Unexpected number of operands!");
+ int SclWidth = 32;
+ int VecWidth = 128;
+
+ switch (MI->getOpcode()) {
+ default:
+ llvm_unreachable("Invalid opcode");
+ case X86::MOVSDrm:
+ case X86::VMOVSDrm:
+ case X86::VMOVSDZrm:
+ case X86::MOVSDrm_alt:
+ case X86::VMOVSDrm_alt:
+ case X86::VMOVSDZrm_alt:
+ case X86::MOVQI2PQIrm:
+ case X86::VMOVQI2PQIrm:
+ case X86::VMOVQI2PQIZrm:
+ SclWidth = 64;
+ VecWidth = 128;
+ break;
+ case X86::MOVSSrm:
+ case X86::VMOVSSrm:
+ case X86::VMOVSSZrm:
+ case X86::MOVSSrm_alt:
+ case X86::VMOVSSrm_alt:
+ case X86::VMOVSSZrm_alt:
+ case X86::MOVDI2PDIrm:
+ case X86::VMOVDI2PDIrm:
+ case X86::VMOVDI2PDIZrm:
+ SclWidth = 32;
+ VecWidth = 128;
+ break;
+ }
+ std::string Comment;
+ raw_string_ostream CS(Comment);
+ const MachineOperand &DstOp = MI->getOperand(0);
+ CS << X86ATTInstPrinter::getRegisterName(DstOp.getReg()) << " = ";
+
+ if (auto *C =
+ X86::getConstantFromPool(*MI, MI->getOperand(1 + X86::AddrDisp))) {
+ if (SclWidth == C->getType()->getScalarSizeInBits()) {
+ if (auto *CI = dyn_cast<ConstantInt>(C)) {
+ CS << "[";
+ printConstant(CI->getValue(), CS);
+ for (int I = 1, E = VecWidth / SclWidth; I < E; ++I) {
+ CS << ",0";
+ }
+ CS << "]";
+ OutStreamer.AddComment(CS.str());
+ break; // early-out
+ }
+ if (auto *CF = dyn_cast<ConstantFP>(C)) {
+ CS << "[";
+ printConstant(CF->getValue(), CS);
+ APFloat ZeroFP = APFloat::getZero(CF->getValue().getSemantics());
+ for (int I = 1, E = VecWidth / SclWidth; I < E; ++I) {
+ CS << ",";
+ printConstant(ZeroFP, CS);
+ }
+ CS << "]";
+ OutStreamer.AddComment(CS.str());
+ break; // early-out
+ }
+ }
+ }
+
+ // We didn't find a constant load, fallback to a shuffle mask decode.
+ CS << (SclWidth == 32 ? "mem[0],zero,zero,zero" : "mem[0],zero");
+ OutStreamer.AddComment(CS.str());
+ break;
+ }
+
#define MOV_CASE(Prefix, Suffix) \
case X86::Prefix##MOVAPD##Suffix##rm: \
case X86::Prefix##MOVAPS##Suffix##rm: \
diff --git a/llvm/test/CodeGen/X86/2008-09-25-sseregparm-1.ll b/llvm/test/CodeGen/X86/2008-09-25-sseregparm-1.ll
index a2dd55767a7efae..6288f7e1d039c9c 100644
--- a/llvm/test/CodeGen/X86/2008-09-25-sseregparm-1.ll
+++ b/llvm/test/CodeGen/X86/2008-09-25-sseregparm-1.ll
@@ -5,7 +5,7 @@
define inreg double @foo1() nounwind {
; CHECK-LABEL: foo1:
; CHECK: # %bb.0:
-; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
+; CHECK-NEXT: movsd {{.*#+}} xmm0 = [1.0E+0,0.0E+0]
; CHECK-NEXT: retl
ret double 1.0
}
@@ -13,7 +13,7 @@ define inreg double @foo1() nounwind {
define inreg float @foo2() nounwind {
; CHECK-LABEL: foo2:
; CHECK: # %bb.0:
-; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-NEXT: movss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; CHECK-NEXT: retl
ret float 1.0
}
diff --git a/llvm/test/CodeGen/X86/GlobalISel/fconstant.ll b/llvm/test/CodeGen/X86/GlobalISel/fconstant.ll
index 1e08c804af586e3..a9b2037e9947a1b 100644
--- a/llvm/test/CodeGen/X86/GlobalISel/fconstant.ll
+++ b/llvm/test/CodeGen/X86/GlobalISel/fconstant.ll
@@ -8,7 +8,7 @@
define void @test_float(ptr %a , float %b) {
; CHECK64_SMALL-LABEL: test_float:
; CHECK64_SMALL: # %bb.0: # %entry
-; CHECK64_SMALL-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK64_SMALL-NEXT: movss {{.*#+}} xmm1 = [5.5E+0,0.0E+0,0.0E+0,0.0E+0]
; CHECK64_SMALL-NEXT: addss %xmm0, %xmm1
; CHECK64_SMALL-NEXT: movd %xmm1, %eax
; CHECK64_SMALL-NEXT: movl %eax, (%rdi)
@@ -26,7 +26,7 @@ define void @test_float(ptr %a , float %b) {
; CHECK32: # %bb.0: # %entry
; CHECK32-NEXT: movl {{[0-9]+}}(%esp), %eax
; CHECK32-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; CHECK32-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK32-NEXT: movss {{.*#+}} xmm0 = [5.5E+0,0.0E+0,0.0E+0,0.0E+0]
; CHECK32-NEXT: movd %ecx, %xmm1
; CHECK32-NEXT: addss %xmm0, %xmm1
; CHECK32-NEXT: movd %xmm1, %ecx
diff --git a/llvm/test/CodeGen/X86/asm-reg-type-mismatch-avx512.ll b/llvm/test/CodeGen/X86/asm-reg-type-mismatch-avx512.ll
index 053ca11b95a50c7..56b05418afa9464 100644
--- a/llvm/test/CodeGen/X86/asm-reg-type-mismatch-avx512.ll
+++ b/llvm/test/CodeGen/X86/asm-reg-type-mismatch-avx512.ll
@@ -5,7 +5,7 @@ define i64 @test1() nounwind {
; CHECK-LABEL: test1:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: #APP
-; CHECK-NEXT: vmovq {{.*#+}} xmm16 = mem[0],zero
+; CHECK-NEXT: vmovq 0, %xmm16
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovq %xmm16, %rax
; CHECK-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/atomic-fp.ll b/llvm/test/CodeGen/X86/atomic-fp.ll
index d933ffec623b945..1094edd19af4380 100644
--- a/llvm/test/CodeGen/X86/atomic-fp.ll
+++ b/llvm/test/CodeGen/X86/atomic-fp.ll
@@ -207,28 +207,28 @@ define dso_local void @fadd_32g() nounwind {
;
; X86-SSE2-LABEL: fadd_32g:
; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE2-NEXT: movss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X86-SSE2-NEXT: addss glob32, %xmm0
; X86-SSE2-NEXT: movss %xmm0, glob32
; X86-SSE2-NEXT: retl
;
; X86-AVX-LABEL: fadd_32g:
; X86-AVX: # %bb.0:
-; X86-AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX-NEXT: vmovss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X86-AVX-NEXT: vaddss glob32, %xmm0, %xmm0
; X86-AVX-NEXT: vmovss %xmm0, glob32
; X86-AVX-NEXT: retl
;
; X64-SSE-LABEL: fadd_32g:
; X64-SSE: # %bb.0:
-; X64-SSE-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X64-SSE-NEXT: movss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X64-SSE-NEXT: addss glob32(%rip), %xmm0
; X64-SSE-NEXT: movss %xmm0, glob32(%rip)
; X64-SSE-NEXT: retq
;
; X64-AVX-LABEL: fadd_32g:
; X64-AVX: # %bb.0:
-; X64-AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X64-AVX-NEXT: vmovss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X64-AVX-NEXT: vaddss glob32(%rip), %xmm0, %xmm0
; X64-AVX-NEXT: vmovss %xmm0, glob32(%rip)
; X64-AVX-NEXT: retq
@@ -319,14 +319,14 @@ define dso_local void @fadd_64g() nounwind {
;
; X64-SSE-LABEL: fadd_64g:
; X64-SSE: # %bb.0:
-; X64-SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
+; X64-SSE-NEXT: movsd {{.*#+}} xmm0 = [1.0E+0,0.0E+0]
; X64-SSE-NEXT: addsd glob64(%rip), %xmm0
; X64-SSE-NEXT: movsd %xmm0, glob64(%rip)
; X64-SSE-NEXT: retq
;
; X64-AVX-LABEL: fadd_64g:
; X64-AVX: # %bb.0:
-; X64-AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
+; X64-AVX-NEXT: vmovsd {{.*#+}} xmm0 = [1.0E+0,0.0E+0]
; X64-AVX-NEXT: vaddsd glob64(%rip), %xmm0, %xmm0
; X64-AVX-NEXT: vmovsd %xmm0, glob64(%rip)
; X64-AVX-NEXT: retq
@@ -368,14 +368,14 @@ define dso_local void @fadd_32imm() nounwind {
;
; X86-SSE2-LABEL: fadd_32imm:
; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE2-NEXT: movss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X86-SSE2-NEXT: addss -559038737, %xmm0
; X86-SSE2-NEXT: movss %xmm0, -559038737
; X86-SSE2-NEXT: retl
;
; X86-AVX-LABEL: fadd_32imm:
; X86-AVX: # %bb.0:
-; X86-AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX-NEXT: vmovss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X86-AVX-NEXT: vaddss -559038737, %xmm0, %xmm0
; X86-AVX-NEXT: vmovss %xmm0, -559038737
; X86-AVX-NEXT: retl
@@ -383,7 +383,7 @@ define dso_local void @fadd_32imm() nounwind {
; X64-SSE-LABEL: fadd_32imm:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: movl $3735928559, %eax # imm = 0xDEADBEEF
-; X64-SSE-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X64-SSE-NEXT: movss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X64-SSE-NEXT: addss (%rax), %xmm0
; X64-SSE-NEXT: movss %xmm0, (%rax)
; X64-SSE-NEXT: retq
@@ -391,7 +391,7 @@ define dso_local void @fadd_32imm() nounwind {
; X64-AVX-LABEL: fadd_32imm:
; X64-AVX: # %bb.0:
; X64-AVX-NEXT: movl $3735928559, %eax # imm = 0xDEADBEEF
-; X64-AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X64-AVX-NEXT: vmovss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X64-AVX-NEXT: vaddss (%rax), %xmm0, %xmm0
; X64-AVX-NEXT: vmovss %xmm0, (%rax)
; X64-AVX-NEXT: retq
@@ -483,7 +483,7 @@ define dso_local void @fadd_64imm() nounwind {
; X64-SSE-LABEL: fadd_64imm:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: movl $3735928559, %eax # imm = 0xDEADBEEF
-; X64-SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
+; X64-SSE-NEXT: movsd {{.*#+}} xmm0 = [1.0E+0,0.0E+0]
; X64-SSE-NEXT: addsd (%rax), %xmm0
; X64-SSE-NEXT: movsd %xmm0, (%rax)
; X64-SSE-NEXT: retq
@@ -491,7 +491,7 @@ define dso_local void @fadd_64imm() nounwind {
; X64-AVX-LABEL: fadd_64imm:
; X64-AVX: # %bb.0:
; X64-AVX-NEXT: movl $3735928559, %eax # imm = 0xDEADBEEF
-; X64-AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
+; X64-AVX-NEXT: vmovsd {{.*#+}} xmm0 = [1.0E+0,0.0E+0]
; X64-AVX-NEXT: vaddsd (%rax), %xmm0, %xmm0
; X64-AVX-NEXT: vmovsd %xmm0, (%rax)
; X64-AVX-NEXT: retq
@@ -534,7 +534,7 @@ define dso_local void @fadd_32stack() nounwind {
; X86-SSE2-LABEL: fadd_32stack:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %eax
-; X86-SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE2-NEXT: movss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X86-SSE2-NEXT: addss (%esp), %xmm0
; X86-SSE2-NEXT: movss %xmm0, (%esp)
; X86-SSE2-NEXT: popl %eax
@@ -543,7 +543,7 @@ define dso_local void @fadd_32stack() nounwind {
; X86-AVX-LABEL: fadd_32stack:
; X86-AVX: # %bb.0:
; X86-AVX-NEXT: pushl %eax
-; X86-AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX-NEXT: vmovss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X86-AVX-NEXT: vaddss (%esp), %xmm0, %xmm0
; X86-AVX-NEXT: vmovss %xmm0, (%esp)
; X86-AVX-NEXT: popl %eax
@@ -551,14 +551,14 @@ define dso_local void @fadd_32stack() nounwind {
;
; X64-SSE-LABEL: fadd_32stack:
; X64-SSE: # %bb.0:
-; X64-SSE-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X64-SSE-NEXT: movss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X64-SSE-NEXT: addss -{{[0-9]+}}(%rsp), %xmm0
; X64-SSE-NEXT: movss %xmm0, -{{[0-9]+}}(%rsp)
; X64-SSE-NEXT: retq
;
; X64-AVX-LABEL: fadd_32stack:
; X64-AVX: # %bb.0:
-; X64-AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X64-AVX-NEXT: vmovss {{.*#+}} xmm0 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; X64-AVX-NEXT: vaddss -{{[0-9]+}}(%rsp), %xmm0, %xmm0
; X64-AVX-NEXT: vmovss %xmm0, -{{[0-9]+}}(%rsp)
; X64-AVX-NEXT: retq
@@ -650,14 +650,14 @@ define dso_local void @fadd_64stack() nounwind {
;
; X64-SSE-LABEL: fadd_64stack:
; X64-SSE: # %bb.0:
-; X64-SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
+; X64-SSE-NEXT: movsd {{.*#+}} xmm0 = [1.0E+0,0.0E+0]
; X64-SSE-NEXT: addsd -{{[0-9]+}}(%rsp), %xmm0
; X64-SSE-NEXT: movsd %xmm0, -{{[0-9]+}}(%rsp)
; X64-SSE-NEXT: retq
;
; X64-AVX-LABEL: fadd_64stack:
; X64-AVX: # %bb.0:
-; X64-AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
+; X64-AVX-NEXT: vmovsd {{.*#+}} xmm0 = [1.0E+0,0.0E+0]
; X64-AVX-NEXT: vaddsd -{{[0-9]+}}(%rsp), %xmm0, %xmm0
; X64-AVX-NEXT: vmovsd %xmm0, -{{[0-9]+}}(%rsp)
; X64-AVX-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/avx512-cmp.ll b/llvm/test/CodeGen/X86/avx512-cmp.ll
index 919edb334b3671b..0c3d9d6f7277c40 100644
--- a/llvm/test/CodeGen/X86/avx512-cmp.ll
+++ b/llvm/test/CodeGen/X86/avx512-cmp.ll
@@ -70,7 +70,7 @@ define float @test5(float %p) #0 {
; ALL-NEXT: retq
; ALL-NEXT: LBB3_1: ## %if.end
; ALL-NEXT: vcmpltss %xmm0, %xmm1, %k1
-; ALL-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; ALL-NEXT: vmovss {{.*#+}} xmm0 = [-1.0E+0,0.0E+0,0.0E+0,0.0E+0]
; ALL-NEXT: vmovss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 {%k1}
; ALL-NEXT: retq
entry:
diff --git a/llvm/test/CodeGen/X86/avx512-fma-intrinsics.ll b/llvm/test/CodeGen/X86/avx512-fma-intrinsics.ll
index def5ad51d732c29..c5a994e6846a455 100644
--- a/llvm/test/CodeGen/X86/avx512-fma-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512-fma-intrinsics.ll
@@ -1150,19 +1150,12 @@ define <16 x float>@test_int_x86_avx512_mask_vfnmadd_ps_512(<16 x float> %x0, <1
; This test case used to crash due to combineFMA not bitcasting results of isFNEG.
define <4 x float> @foo() {
-; X86-LABEL: foo:
-; X86: # %bb.0: # %entry
-; X86-NEXT: vmovss (%eax), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x00]
-; X86-NEXT: # xmm0 = mem[0],zero,zero,zero
-; X86-NEXT: vfmsub213ss {rd-sae}, %xmm0, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x38,0xab,0xc0]
-; X86-NEXT: retl # encoding: [0xc3]
-;
-; X64-LABEL: foo:
-; X64: # %bb.0: # %entry
-; X64-NEXT: vmovss (%rax), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x00]
-; X64-NEXT: # xmm0 = mem[0],zero,zero,zero
-; X64-NEXT: vfmsub213ss {rd-sae}, %xmm0, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x38,0xab,0xc0]
-; X64-NEXT: retq # encoding: [0xc3]
+; CHECK-LABEL: foo:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-NEXT: # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x00]
+; CHECK-NEXT: vfmsub213ss {rd-sae}, %xmm0, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x38,0xab,0xc0]
+; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
entry:
%0 = load <4 x float>, ptr undef, align 16
%sub = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %0
diff --git a/llvm/test/CodeGen/X86/avx512-insert-extract.ll b/llvm/test/CodeGen/X86/avx512-insert-extract.ll
index 4a2dd7673f4e767..abfe3e6428e6632 100644
--- a/llvm/test/CodeGen/X86/avx512-insert-extract.ll
+++ b/llvm/test/CodeGen/X86/avx512-insert-extract.ll
@@ -2175,7 +2175,7 @@ define void @test_concat_v2i1(ptr %arg, ptr %arg1, ptr %arg2) nounwind {
; KNL-NEXT: movzwl %ax, %eax
; KNL-NEXT: vmovd %eax, %xmm1
; KNL-NEXT: vcvtph2ps %xmm1, %xmm1
-; KNL-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; KNL-NEXT: vmovss {{.*#+}} xmm2 = [6.0E+0,0.0E+0,0.0E+0,0.0E+0]
; KNL-NEXT: vucomiss %xmm2, %xmm1
; KNL-NEXT: setb %al
; KNL-NEXT: andl $1, %eax
@@ -2217,7 +2217,7 @@ define void @test_concat_v2i1(ptr %arg, ptr %arg1, ptr %arg2) nounwind {
; SKX-NEXT: movzwl %ax, %eax
; SKX-NEXT: vmovd %eax, %xmm1
; SKX-NEXT: vcvtph2ps %xmm1, %xmm1
-; SKX-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; SKX-NEXT: vmovss {{.*#+}} xmm2 = [6.0E+0,0.0E+0,0.0E+0,0.0E+0]
; SKX-NEXT: vucomiss %xmm2, %xmm1
; SKX-NEXT: setb %al
; SKX-NEXT: kmovd %eax, %k0
diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
index a5a4bf1e53631ee..6c9c28bc9e55e11 100644
--- a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
@@ -10104,10 +10104,10 @@ define void @fmadd_ss_mask_memfold(ptr %a, ptr %b, i8 %c) {
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x0c]
; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx ## encoding: [0x8b,0x4c,0x24,0x08]
; X86-NEXT: movl {{[0-9]+}}(%esp), %edx ## encoding: [0x8b,0x54,0x24,0x04]
-; X86-NEXT: vmovss (%edx), %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x02]
-; X86-NEXT: ## xmm0 = mem[0],zero,zero,zero
-; X86-NEXT: vmovss (%ecx), %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x09]
-; X86-NEXT: ## xmm1 = mem[0],zero,zero,zero
+; X86-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-NEXT: ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x02]
+; X86-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; X86-NEXT: ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x09]
; X86-NEXT: vfmadd213ss %xmm0, %xmm0, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0xa9,0xc8]
; X86-NEXT: ## xmm1 = (xmm0 * xmm1) + xmm0
; X86-NEXT: kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
@@ -10117,10 +10117,10 @@ define void @fmadd_ss_mask_memfold(ptr %a, ptr %b, i8 %c) {
;
; X64-LABEL: fmadd_ss_mask_memfold:
; X64: ## %bb.0:
-; X64-NEXT: vmovss (%rdi), %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x07]
-; X64-NEXT: ## xmm0 = mem[0],zero,zero,zero
-; X64-NEXT: vmovss (%rsi), %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x0e]
-; X64-NEXT: ## xmm1 = mem[0],zero,zero,zero
+; X64-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X64-NEXT: ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x07]
+; X64-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; X64-NEXT: ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x0e]
; X64-NEXT: vfmadd213ss %xmm0, %xmm0, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0xa9,0xc8]
; X64-NEXT: ## xmm1 = (xmm0 * xmm1) + xmm0
; X64-NEXT: kmovw %ed...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/78601
More information about the llvm-commits
mailing list