3 files changed, 52 insertions, 4 deletions
diff --git a/llvm/lib/Target/X86/X86InstrInfo.cpp b/llvm/lib/Target/X86/X86InstrInfo.cpp
index d79806e664b..efa972ad9a9 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.cpp
+++ b/llvm/lib/Target/X86/X86InstrInfo.cpp
@@ -6395,12 +6395,16 @@ static bool isAssociativeAndCommutative(const MachineInstr &Inst) {
   case X86::ADDPSrr:
   case X86::ADDSDrr:
   case X86::ADDSSrr:
+  case X86::MULPDrr:
+  case X86::MULPSrr:
+  case X86::MULSDrr:
+  case X86::MULSSrr:
   case X86::VADDPDrr:
   case X86::VADDPSrr:
   case X86::VADDSDrr:
   case X86::VADDSSrr:
-  case X86::MULSDrr:
-  case X86::MULSSrr:
+  case X86::VMULPDrr:
+  case X86::VMULPSrr:
   case X86::VMULSDrr:
   case X86::VMULSSrr:
     return Inst.getParent()->getParent()->getTarget().Options.UnsafeFPMath;
diff --git a/llvm/test/CodeGen/X86/machine-combiner.ll b/llvm/test/CodeGen/X86/machine-combiner.ll
index af1c661417c..fb21c19744f 100644
--- a/llvm/test/CodeGen/X86/machine-combiner.ll
+++ b/llvm/test/CodeGen/X86/machine-combiner.ll
@@ -254,3 +254,47 @@ define <2 x double> @reassociate_adds_v2f64(<2 x double> %x0, <2 x double> %x1,
   ret <2 x double> %t2
 }
 
+; Verify that SSE and AVX 128-bit vector single-precison multiplies are reassociated.
+
+define <4 x float> @reassociate_muls_v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, <4 x float> %x3) {
+; SSE-LABEL: reassociate_muls_v4f32:
+; SSE:       # BB#0:
+; SSE-NEXT:    addps %xmm1, %xmm0
+; SSE-NEXT:    mulps %xmm3, %xmm2
+; SSE-NEXT:    mulps %xmm2, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: reassociate_muls_v4f32:
+; AVX:       # BB#0:
+; AVX-NEXT:    vaddps %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vmulps %xmm3, %xmm2, %xmm1
+; AVX-NEXT:    vmulps %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retq
+  %t0 = fadd <4 x float> %x0, %x1
+  %t1 = fmul <4 x float> %x2, %t0
+  %t2 = fmul <4 x float> %x3, %t1
+  ret <4 x float> %t2
+}
+
+; Verify that SSE and AVX 128-bit vector double-precison multiplies are reassociated.
+
+define <2 x double> @reassociate_muls_v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, <2 x double> %x3) {
+; SSE-LABEL: reassociate_muls_v2f64:
+; SSE:       # BB#0:
+; SSE-NEXT:    addpd %xmm1, %xmm0
+; SSE-NEXT:    mulpd %xmm3, %xmm2
+; SSE-NEXT:    mulpd %xmm2, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: reassociate_muls_v2f64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vaddpd %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vmulpd %xmm3, %xmm2, %xmm1
+; AVX-NEXT:    vmulpd %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retq
+  %t0 = fadd <2 x double> %x0, %x1
+  %t1 = fmul <2 x double> %x2, %t0
+  %t2 = fmul <2 x double> %x3, %t1
+  ret <2 x double> %t2
+}
+
diff --git a/llvm/test/CodeGen/X86/sqrt-fastmath.ll b/llvm/test/CodeGen/X86/sqrt-fastmath.ll
index 9b851db8121..daa6d7292e7 100644
--- a/llvm/test/CodeGen/X86/sqrt-fastmath.ll
+++ b/llvm/test/CodeGen/X86/sqrt-fastmath.ll
@@ -99,8 +99,8 @@ define <4 x float> @reciprocal_square_root_v4f32(<4 x float> %x) #0 {
 ; ESTIMATE-LABEL: reciprocal_square_root_v4f32:
 ; ESTIMATE:       # BB#0:
 ; ESTIMATE-NEXT:    vrsqrtps %xmm0, %xmm1
-; ESTIMATE-NEXT:    vmulps %xmm1, %xmm1, %xmm2
-; ESTIMATE-NEXT:    vmulps %xmm0, %xmm2, %xmm0
+; ESTIMATE-NEXT:    vmulps %xmm0, %xmm1, %xmm0
+; ESTIMATE-NEXT:    vmulps %xmm0, %xmm1, %xmm0
 ; ESTIMATE-NEXT:    vaddps {{.*}}(%rip), %xmm0, %xmm0
 ; ESTIMATE-NEXT:    vmulps {{.*}}(%rip), %xmm1, %xmm1
 ; ESTIMATE-NEXT:    vmulps %xmm1, %xmm0, %xmm0