[InstCombine] Fix ninf propagation for fcmp+sel -> minmax

[dtcxzyw]

Proof: https://alive2.llvm.org/ce/z/nCrvfr
Closes #136430

[llvmbot]

@llvm/pr-subscribers-llvm-transforms

Author: Yingwei Zheng (dtcxzyw)

Changes

Proof: https://alive2.llvm.org/ce/z/nCrvfr
Closes #136430

---

Full diff: https://github.com/llvm/llvm-project/pull/136433.diff

4 Files Affected:

• (modified) llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp (+6-2)
• (modified) llvm/test/Transforms/InstCombine/fcmp-fadd-select.ll (+1-1)
• (modified) llvm/test/Transforms/InstCombine/minmax-fp.ll (+12-2)
• (modified) llvm/test/Transforms/InstCombine/unordered-fcmp-select.ll (+1-1)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index 4bba2f406b4c1..3766ed699a192 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -3922,16 +3922,20 @@ Instruction *InstCombinerImpl::visitSelectInst(SelectInst &SI) {
       if (match(&SI, m_OrdOrUnordFMax(m_Value(X), m_Value(Y)))) {
         Value *BinIntr =
             Builder.CreateBinaryIntrinsic(Intrinsic::maxnum, X, Y, &SI);
-        if (auto *BinIntrInst = dyn_cast<Instruction>(BinIntr))
+        if (auto *BinIntrInst = dyn_cast<Instruction>(BinIntr)) {
           BinIntrInst->setHasNoNaNs(FCmp->hasNoNaNs());
+          BinIntrInst->setHasNoInfs(FCmp->hasNoInfs());
+        }
         return replaceInstUsesWith(SI, BinIntr);
       }
 
       if (match(&SI, m_OrdOrUnordFMin(m_Value(X), m_Value(Y)))) {
         Value *BinIntr =
             Builder.CreateBinaryIntrinsic(Intrinsic::minnum, X, Y, &SI);
-        if (auto *BinIntrInst = dyn_cast<Instruction>(BinIntr))
+        if (auto *BinIntrInst = dyn_cast<Instruction>(BinIntr)) {
           BinIntrInst->setHasNoNaNs(FCmp->hasNoNaNs());
+          BinIntrInst->setHasNoInfs(FCmp->hasNoInfs());
+        }
         return replaceInstUsesWith(SI, BinIntr);
       }
     }
diff --git a/llvm/test/Transforms/InstCombine/fcmp-fadd-select.ll b/llvm/test/Transforms/InstCombine/fcmp-fadd-select.ll
index 15fad55db8df1..e05ef6df1d41b 100644
--- a/llvm/test/Transforms/InstCombine/fcmp-fadd-select.ll
+++ b/llvm/test/Transforms/InstCombine/fcmp-fadd-select.ll
@@ -663,7 +663,7 @@ define float @test_fcmp_ogt_fadd_select_rewrite_flags2(float %in) {
 define float @test_fcmp_ogt_fadd_select_rewrite_and_fastmath(float %in) {
 ; CHECK-LABEL: define float @test_fcmp_ogt_fadd_select_rewrite_and_fastmath(
 ; CHECK-SAME: float [[IN:%.*]]) {
-; CHECK-NEXT:    [[SEL_NEW:%.*]] = call fast float @llvm.maxnum.f32(float [[IN]], float 0.000000e+00)
+; CHECK-NEXT:    [[SEL_NEW:%.*]] = call reassoc nnan nsz arcp contract afn float @llvm.maxnum.f32(float [[IN]], float 0.000000e+00)
 ; CHECK-NEXT:    [[ADD_NEW:%.*]] = fadd fast float [[SEL_NEW]], 1.000000e+00
 ; CHECK-NEXT:    ret float [[ADD_NEW]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/minmax-fp.ll b/llvm/test/Transforms/InstCombine/minmax-fp.ll
index 4fe8cf374344e..a8470a20365e9 100644
--- a/llvm/test/Transforms/InstCombine/minmax-fp.ll
+++ b/llvm/test/Transforms/InstCombine/minmax-fp.ll
@@ -331,7 +331,7 @@ define float @maxnum_ogt_fmf_on_select(float %a, float %b) {
 
 define <2 x float> @maxnum_oge_fmf_on_select(<2 x float> %a, <2 x float> %b) {
 ; CHECK-LABEL: @maxnum_oge_fmf_on_select(
-; CHECK-NEXT:    [[F:%.*]] = call ninf nsz <2 x float> @llvm.maxnum.v2f32(<2 x float> [[A:%.*]], <2 x float> [[B:%.*]])
+; CHECK-NEXT:    [[F:%.*]] = call nsz <2 x float> @llvm.maxnum.v2f32(<2 x float> [[A:%.*]], <2 x float> [[B:%.*]])
 ; CHECK-NEXT:    ret <2 x float> [[F]]
 ;
   %cond = fcmp oge <2 x float> %a, %b
@@ -383,6 +383,16 @@ define float @maxnum_no_nnan(float %a, float %b) {
   ret float %f
 }
 
+define float @minnum_olt_fmf_on_select_both_ninf(float %a, float %b) {
+; CHECK-LABEL: @minnum_olt_fmf_on_select_both_ninf(
+; CHECK-NEXT:    [[F:%.*]] = call ninf nsz float @llvm.minnum.f32(float [[A:%.*]], float [[B:%.*]])
+; CHECK-NEXT:    ret float [[F]]
+;
+  %cond = fcmp ninf olt float %a, %b
+  %f = select nnan ninf nsz i1 %cond, float %a, float %b
+  ret float %f
+}
+
 define float @minnum_olt_fmf_on_select(float %a, float %b) {
 ; CHECK-LABEL: @minnum_olt_fmf_on_select(
 ; CHECK-NEXT:    [[F:%.*]] = call nsz float @llvm.minnum.f32(float [[A:%.*]], float [[B:%.*]])
@@ -395,7 +405,7 @@ define float @minnum_olt_fmf_on_select(float %a, float %b) {
 
 define <2 x float> @minnum_ole_fmf_on_select(<2 x float> %a, <2 x float> %b) {
 ; CHECK-LABEL: @minnum_ole_fmf_on_select(
-; CHECK-NEXT:    [[F:%.*]] = call ninf nsz <2 x float> @llvm.minnum.v2f32(<2 x float> [[A:%.*]], <2 x float> [[B:%.*]])
+; CHECK-NEXT:    [[F:%.*]] = call nsz <2 x float> @llvm.minnum.v2f32(<2 x float> [[A:%.*]], <2 x float> [[B:%.*]])
 ; CHECK-NEXT:    ret <2 x float> [[F]]
 ;
   %cond = fcmp ole <2 x float> %a, %b
diff --git a/llvm/test/Transforms/InstCombine/unordered-fcmp-select.ll b/llvm/test/Transforms/InstCombine/unordered-fcmp-select.ll
index 178795f9f9a83..ab4c997014699 100644
--- a/llvm/test/Transforms/InstCombine/unordered-fcmp-select.ll
+++ b/llvm/test/Transforms/InstCombine/unordered-fcmp-select.ll
@@ -115,7 +115,7 @@ define float @select_max_ugt_2_use_cmp(float %a, float %b) {
 ; CHECK-LABEL: @select_max_ugt_2_use_cmp(
 ; CHECK-NEXT:    [[CMP:%.*]] = fcmp reassoc ugt float [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    call void @foo(i1 [[CMP]])
-; CHECK-NEXT:    [[SEL:%.*]] = call reassoc ninf nsz arcp contract afn float @llvm.maxnum.f32(float [[A]], float [[B]])
+; CHECK-NEXT:    [[SEL:%.*]] = call reassoc nsz arcp contract afn float @llvm.maxnum.f32(float [[A]], float [[B]])
 ; CHECK-NEXT:    ret float [[SEL]]
 ;
   %cmp = fcmp reassoc ugt float %a, %b

[nikic]

I'm a bit confused here. What are the semantics of ninf on select supposed to be? Normally it means that both the inputs and outputs have to be ninf. But if I understand correctly, for select we interpret this as only the output (and thus the chosen input) having to be ninf?

This makes some sense to me (esp as these are the only possible semantics for phis) but I can't find any wording to that effect in LangRef.

[dtcxzyw]

for select we interpret this as only the output (and thus the chosen input) having to be ninf

Yes. I found this while implementing llubi. I believe there are many places that interpret the select in this way...

[arsenm]

This makes some sense to me (esp as these are the only possible semantics for phis) but I can't find any wording to that effect in LangRef.

This is news to me. I have always assumed this applied to both inputs like every other operation

[dtcxzyw]

I have posted an RFC for further discussion: https://discourse.llvm.org/t/rfc-clarify-the-behavior-of-select-with-fp-poison-generating-flags/85974