[llvm] r333995 - [X86][SSE] Add basic PACKUS support to X86TargetLowering::computeKnownBitsForTargetNode

[Simon Pilgrim via llvm-commits]

Author: rksimon
Date: Tue Jun  5 02:45:03 2018
New Revision: 333995

URL: http://llvm.org/viewvc/llvm-project?rev=333995&view=rev
Log:
[X86][SSE] Add basic PACKUS support to X86TargetLowering::computeKnownBitsForTargetNode

Helps improve analysis of saturation ops

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/test/CodeGen/X86/vector-trunc-usat.ll

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=333995&r1=333994&r2=333995&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Tue Jun  5 02:45:03 2018
@@ -28744,6 +28744,19 @@ void X86TargetLowering::computeKnownBits
     }
     break;
   }
+  case X86ISD::PACKUS: {
+    // PACKUS is just a truncation if the upper half is zero.
+    // TODO: Add DemandedElts support.
+    KnownBits Known2;
+    DAG.computeKnownBits(Op.getOperand(0), Known, Depth + 1);
+    DAG.computeKnownBits(Op.getOperand(1), Known2, Depth + 1);
+    Known.One &= Known2.One;
+    Known.Zero &= Known2.Zero;
+    if (Known.countMinLeadingZeros() < BitWidth)
+      Known.resetAll();
+    Known = Known.trunc(BitWidth);
+    break;
+  }
   case X86ISD::VZEXT: {
     // TODO: Add DemandedElts support.
     SDValue N0 = Op.getOperand(0);

Modified: llvm/trunk/test/CodeGen/X86/vector-trunc-usat.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-trunc-usat.ll?rev=333995&r1=333994&r2=333995&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vector-trunc-usat.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vector-trunc-usat.ll Tue Jun  5 02:45:03 2018
@@ -2076,7 +2076,7 @@ define void @trunc_usat_v8i32_v8i8_store
 ; AVX1-NEXT:    vpminud %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpminud %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
@@ -2087,7 +2087,7 @@ define void @trunc_usat_v8i32_v8i8_store
 ; AVX2-NEXT:    vpminud %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq