[PATCH] D119629: [DAG] computeKnownBits - add ISD::AVGCEILU handling

[Simon Pilgrim via Phabricator via llvm-commits]

This revision was landed with ongoing or failed builds.
This revision was automatically updated to reflect the committed changes.
Closed by commit rG30e9cdd1aa2e: [DAG] computeKnownBits - add ISD::AVGCEILU handling (authored by RKSimon).

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D119629/new/

https://reviews.llvm.org/D119629

Files:
  llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
  llvm/test/CodeGen/X86/combine-pavg.ll


Index: llvm/test/CodeGen/X86/combine-pavg.ll
===================================================================
--- llvm/test/CodeGen/X86/combine-pavg.ll
+++ llvm/test/CodeGen/X86/combine-pavg.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX
 
 declare <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8>, <16 x i8>) nounwind readnone
 declare <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16>, <8 x i16>) nounwind readnone
@@ -21,7 +21,6 @@
   ret <16 x i8> %1
 }
 
-; TODO: Failure to remove masks as we know the upper bits are zero
 define <16 x i8> @combine_pavgw_knownbits(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> %a2, <8 x i16> %a3) {
 ; SSE-LABEL: combine_pavgw_knownbits:
 ; SSE:       # %bb.0:
@@ -32,41 +31,20 @@
 ; SSE-NEXT:    pand %xmm4, %xmm2
 ; SSE-NEXT:    pand %xmm4, %xmm3
 ; SSE-NEXT:    pavgw %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
-; SSE-NEXT:    pand %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    packuswb %xmm3, %xmm0
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: combine_pavgw_knownbits:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [31,31,31,31,31,31,31,31]
-; AVX1-NEXT:    vpand %xmm4, %xmm0, %xmm0
-; AVX1-NEXT:    vpand %xmm4, %xmm1, %xmm1
-; AVX1-NEXT:    vpavgw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpand %xmm4, %xmm2, %xmm1
-; AVX1-NEXT:    vpand %xmm4, %xmm3, %xmm2
-; AVX1-NEXT:    vpavgw %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
-; AVX1-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: combine_pavgw_knownbits:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm4 = [31,31,31,31,31,31,31,31]
-; AVX2-NEXT:    vpand %xmm4, %xmm0, %xmm0
-; AVX2-NEXT:    vpand %xmm4, %xmm1, %xmm1
-; AVX2-NEXT:    vpavgw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpand %xmm4, %xmm2, %xmm1
-; AVX2-NEXT:    vpand %xmm4, %xmm3, %xmm2
-; AVX2-NEXT:    vpavgw %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpbroadcastw {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    retq
+; AVX-LABEL: combine_pavgw_knownbits:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm4 = [31,31,31,31,31,31,31,31]
+; AVX-NEXT:    vpand %xmm4, %xmm0, %xmm0
+; AVX-NEXT:    vpand %xmm4, %xmm1, %xmm1
+; AVX-NEXT:    vpavgw %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpand %xmm4, %xmm2, %xmm1
+; AVX-NEXT:    vpand %xmm4, %xmm3, %xmm2
+; AVX-NEXT:    vpavgw %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retq
   %m0 = and <8 x i16> %a0, <i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31>
   %m1 = and <8 x i16> %a1, <i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31>
   %m2 = and <8 x i16> %a2, <i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31>
Index: llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
===================================================================
--- llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -3129,6 +3129,16 @@
     Known = KnownBits::udiv(Known, Known2);
     break;
   }
+  case ISD::AVGCEILU: {
+    Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
+    Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
+    Known = Known.zext(BitWidth + 1);
+    Known2 = Known2.zext(BitWidth + 1);
+    KnownBits One = KnownBits::makeConstant(APInt(1, 1));
+    Known = KnownBits::computeForAddCarry(Known, Known2, One);
+    Known = Known.extractBits(BitWidth, 1);
+    break;
+  }
   case ISD::SELECT:
   case ISD::VSELECT:
     Known = computeKnownBits(Op.getOperand(2), DemandedElts, Depth+1);


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D119629.409215.patch
Type: text/x-patch
Size: 4431 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220216/b703b263/attachment-0001.bin>