[llvm] [NVPTX] Don't propagate `ninf` and `nnan` in `lowerFREM` (PR #147125)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-nvptx

Author: None (paperchalice)

<details>
<summary>Changes</summary>

0/0 and 1/0 can produce nan and inf.

---
Full diff: https://github.com/llvm/llvm-project/pull/147125.diff


2 Files Affected:

- (modified) llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp (+7-3) 
- (added) llvm/test/CodeGen/NVPTX/frem-ninf-nnan.ll (+11) 


``````````diff
diff --git a/llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp b/llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp
index bb0aeb493ed48..9418ca3bf446c 100644
--- a/llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp
+++ b/llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp
@@ -2793,12 +2793,16 @@ static SDValue lowerFREM(SDValue Op, SelectionDAG &DAG,
   EVT Ty = Op.getValueType();
   SDNodeFlags Flags = Op->getFlags();
 
+  // fdiv can still generate inf and nan when nnan and ninf are set.
+  SDNodeFlags NewFlags = Flags;
+  NewFlags.setNoNaNs(false);
+  NewFlags.setNoInfs(false);
   SDValue Div = DAG.getNode(ISD::FDIV, DL, Ty, X, Y, Flags);
-  SDValue Trunc = DAG.getNode(ISD::FTRUNC, DL, Ty, Div, Flags);
+  SDValue Trunc = DAG.getNode(ISD::FTRUNC, DL, Ty, Div, NewFlags);
   SDValue Mul = DAG.getNode(ISD::FMUL, DL, Ty, Trunc, Y,
-                            Flags | SDNodeFlags::AllowContract);
+                            NewFlags | SDNodeFlags::AllowContract);
   SDValue Sub = DAG.getNode(ISD::FSUB, DL, Ty, X, Mul,
-                            Flags | SDNodeFlags::AllowContract);
+                            NewFlags | SDNodeFlags::AllowContract);
 
   if (AllowUnsafeFPMath || Flags.hasNoInfs())
     return Sub;
diff --git a/llvm/test/CodeGen/NVPTX/frem-ninf-nnan.ll b/llvm/test/CodeGen/NVPTX/frem-ninf-nnan.ll
new file mode 100644
index 0000000000000..639f9ab201ad0
--- /dev/null
+++ b/llvm/test/CodeGen/NVPTX/frem-ninf-nnan.ll
@@ -0,0 +1,11 @@
+; RUN: llc %s --stop-after=nvptx-isel -mcpu=sm_60 -o - | FileCheck %s
+
+target triple = "nvptx64-unknown-cuda"
+
+define float @frem_ninf_nnan(float %a, float %b) {
+  ; CHECK:     nnan ninf FDIV32rr_prec
+  ; CHECK-NOT: nnan ninf contract FNEGf32
+  ; CHECK:     contract FNEGf32
+  %r = frem ninf nnan float %a, %b
+  ret float %r
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/147125