[Transforms] Add cos(fabs(x)) -> cos(x) to SimplifyLibCalls

We have this for InstCombine, but forgot to add it for SimplifyLibCalls.

Author: AZero13

llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp

@@ -1934,8 +1934,10 @@ static Value *optimizeTrigReflections(CallInst *Call, LibFunc Func,
   case LibFunc_cos:
   case LibFunc_cosf:
   case LibFunc_cosl:
-    // cos(-X) --> cos(X)
-    if (match(Call->getArgOperand(0), m_FNeg(m_Value(X))))
+    // cos(-x) --> cos(x)
+    // cos(fabs(x)) --> cos(x)
+    if (match(Call->getArgOperand(0), m_FNeg(m_Value(X))) ||
+        match(Call->getArgOperand(0), m_FAbs(m_Value(X))))
       return copyFlags(*Call,
                        B.CreateCall(Call->getCalledFunction(), X, "cos"));
     break;

llvm/test/Transforms/InstCombine/cos-1.ll

@@ -109,9 +109,8 @@ define float @cosf_unary_negated_arg_FMF(float %x) {
 
 define double @cos_unary_fabs_arg(double %x) {
 ; ANY-LABEL: @cos_unary_fabs_arg(
-; ANY-NEXT:    [[FABS:%.*]] = tail call double @llvm.fabs.f64(double [[X:%.*]])
-; ANY-NEXT:    [[R:%.*]] = call double @cos(double [[FABS]])
-; ANY-NEXT:    ret double [[R]]
+; ANY-NEXT:    [[COS:%.*]] = call double @cos(double [[X:%.*]])
+; ANY-NEXT:    ret double [[COS]]
 ;
   %fabs = tail call double @llvm.fabs.f64(double %x)
   %r = call double @cos(double %fabs)
@@ -120,9 +119,8 @@ define double @cos_unary_fabs_arg(double %x) {
 
 define float @cosf_unary_fabs_arg(float %x) {
 ; ANY-LABEL: @cosf_unary_fabs_arg(
-; ANY-NEXT:    [[FABS:%.*]] = tail call float @llvm.fabs.f32(float [[X:%.*]])
-; ANY-NEXT:    [[R:%.*]] = call float @cosf(float [[FABS]])
-; ANY-NEXT:    ret float [[R]]
+; ANY-NEXT:    [[COS:%.*]] = call float @cosf(float [[X:%.*]])
+; ANY-NEXT:    ret float [[COS]]
 ;
   %fabs = tail call float @llvm.fabs.f32(float %x)
   %r = call float @cosf(float %fabs)
@@ -131,9 +129,8 @@ define float @cosf_unary_fabs_arg(float %x) {
 
 define float @cosf_unary_fabs_arg_FMF(float %x) {
 ; ANY-LABEL: @cosf_unary_fabs_arg_FMF(
-; ANY-NEXT:    [[FABS:%.*]] = tail call float @llvm.fabs.f32(float [[X:%.*]])
-; ANY-NEXT:    [[R:%.*]] = call reassoc nnan float @cosf(float [[FABS]])
-; ANY-NEXT:    ret float [[R]]
+; ANY-NEXT:    [[COS:%.*]] = call reassoc nnan float @cosf(float [[X:%.*]])
+; ANY-NEXT:    ret float [[COS]]
 ;
   %fabs = tail call float @llvm.fabs.f32(float %x)
   %r = call nnan reassoc float @cosf(float %fabs)