[llvm] b0eea72 - [X86] Support load shrinking for strict fp nodes in combineCVTPH2PS

[Craig Topper via llvm-commits]

Author: Craig Topper
Date: 2020-06-07T21:09:55-07:00
New Revision: b0eea7213b50970ea2a499eff398d0cc75113a56

URL: https://github.com/llvm/llvm-project/commit/b0eea7213b50970ea2a499eff398d0cc75113a56
DIFF: https://github.com/llvm/llvm-project/commit/b0eea7213b50970ea2a499eff398d0cc75113a56.diff

LOG: [X86] Support load shrinking for strict fp nodes in combineCVTPH2PS

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86ISelLowering.cpp
    llvm/test/CodeGen/X86/vector-half-conversions.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 3d4807884ae6..3bba9212b893 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -44865,18 +44865,22 @@ static SDValue combineCVTPH2PS(SDNode *N, SelectionDAG &DAG,
       return SDValue(N, 0);
     }
 
-    // FIXME: Shrink vector loads.
-    if (IsStrict)
-      return SDValue();
-
     // Convert a full vector load into vzload when not all bits are needed.
     if (ISD::isNormalLoad(Src.getNode()) && Src.hasOneUse()) {
-      LoadSDNode *LN = cast<LoadSDNode>(N->getOperand(0));
+      LoadSDNode *LN = cast<LoadSDNode>(N->getOperand(IsStrict ? 1 : 0));
       if (SDValue VZLoad = narrowLoadToVZLoad(LN, MVT::i64, MVT::v2i64, DAG)) {
         SDLoc dl(N);
-        SDValue Convert = DAG.getNode(N->getOpcode(), dl, MVT::v4f32,
-                                      DAG.getBitcast(MVT::v8i16, VZLoad));
-        DCI.CombineTo(N, Convert);
+        if (IsStrict) {
+          SDValue Convert = DAG.getNode(
+              N->getOpcode(), dl, {MVT::v4f32, MVT::Other},
+              {N->getOperand(0), DAG.getBitcast(MVT::v8i16, VZLoad)});
+          DCI.CombineTo(N, Convert, Convert.getValue(1));
+        } else {
+          SDValue Convert = DAG.getNode(N->getOpcode(), dl, MVT::v4f32,
+                                        DAG.getBitcast(MVT::v8i16, VZLoad));
+          DCI.CombineTo(N, Convert);
+        }
+
         DAG.ReplaceAllUsesOfValueWith(SDValue(LN, 1), VZLoad.getValue(1));
         DCI.recursivelyDeleteUnusedNodes(LN);
         return SDValue(N, 0);

diff  --git a/llvm/test/CodeGen/X86/vector-half-conversions.ll b/llvm/test/CodeGen/X86/vector-half-conversions.ll
index 85d5f580974c..a4be7337f145 100644
--- a/llvm/test/CodeGen/X86/vector-half-conversions.ll
+++ b/llvm/test/CodeGen/X86/vector-half-conversions.ll
@@ -1,9 +1,9 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+f16c -verify-machineinstrs | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+f16c -verify-machineinstrs | FileCheck %s --check-prefixes=ALL,AVX,AVX2,AVX2-SLOW
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+f16c,+fast-variable-shuffle -verify-machineinstrs | FileCheck %s --check-prefixes=ALL,AVX,AVX2,AVX2-FAST
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f -verify-machineinstrs | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vl,+fast-variable-shuffle -verify-machineinstrs | FileCheck %s --check-prefixes=ALL,AVX,AVX512,AVX512VL
+; RUN: llc < %s -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx,+f16c -verify-machineinstrs | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx2,+f16c -verify-machineinstrs | FileCheck %s --check-prefixes=ALL,AVX,AVX2,AVX2-SLOW
+; RUN: llc < %s -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx2,+f16c,+fast-variable-shuffle -verify-machineinstrs | FileCheck %s --check-prefixes=ALL,AVX,AVX2,AVX2-FAST
+; RUN: llc < %s -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx512f -verify-machineinstrs | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512F
+; RUN: llc < %s -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vl,+fast-variable-shuffle -verify-machineinstrs | FileCheck %s --check-prefixes=ALL,AVX,AVX512,AVX512VL
 
 ;
 ; Half to Float