[PATCH] D28979: [X86] Add demanded elts support for the inputs to pclmul intrinsic

[Craig Topper via Phabricator via llvm-commits]

craig.topper created this revision.

This intrinsic uses bit 0 and bit 4 of an immediate argument to determine which bits of its inputs to read. This patch uses this information to simplify the demanded elements of the input vectors.


https://reviews.llvm.org/D28979

Files:
  lib/Transforms/InstCombine/InstCombineCalls.cpp
  test/Transforms/InstCombine/x86-clmulqdq.ll


Index: test/Transforms/InstCombine/x86-clmulqdq.ll
===================================================================
--- /dev/null
+++ test/Transforms/InstCombine/x86-clmulqdq.ll
@@ -0,0 +1,48 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -instcombine -S | FileCheck %s
+
+declare <2 x i64> @llvm.x86.pclmulqdq(<2 x i64>, <2 x i64>, i8)
+
+define <2 x i64> @test_demanded_elts_pclmulqdq_0(<2 x i64> %a0, <2 x i64> %a1) {
+; CHECK-LABEL: @test_demanded_elts_pclmulqdq_0(
+; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x i64> @llvm.x86.pclmulqdq(<2 x i64> [[A0:%.*]], <2 x i64> [[A1:%.*]], i8 0)
+; CHECK-NEXT:    ret <2 x i64> [[TMP1]]
+;
+  %1 = insertelement <2 x i64> %a0, i64 1, i64 1
+  %2 = insertelement <2 x i64> %a1, i64 1, i64 1
+  %3 = call <2 x i64> @llvm.x86.pclmulqdq(<2 x i64> %1, <2 x i64> %2, i8 0)
+  ret <2 x i64> %3
+}
+
+define <2 x i64> @test_demanded_elts_pclmulqdq_1(<2 x i64> %a0, <2 x i64> %a1) {
+; CHECK-LABEL: @test_demanded_elts_pclmulqdq_1(
+; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x i64> @llvm.x86.pclmulqdq(<2 x i64> <i64 undef, i64 1>, <2 x i64> [[A1:%.*]], i8 1)
+; CHECK-NEXT:    ret <2 x i64> [[TMP1]]
+;
+  %1 = insertelement <2 x i64> %a0, i64 1, i64 1
+  %2 = insertelement <2 x i64> %a1, i64 1, i64 1
+  %3 = call <2 x i64> @llvm.x86.pclmulqdq(<2 x i64> %1, <2 x i64> %2, i8 1)
+  ret <2 x i64> %3
+}
+
+define <2 x i64> @test_demanded_elts_pclmulqdq_16(<2 x i64> %a0, <2 x i64> %a1) {
+; CHECK-LABEL: @test_demanded_elts_pclmulqdq_16(
+; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x i64> @llvm.x86.pclmulqdq(<2 x i64> [[A0:%.*]], <2 x i64> <i64 undef, i64 1>, i8 16)
+; CHECK-NEXT:    ret <2 x i64> [[TMP1]]
+;
+  %1 = insertelement <2 x i64> %a0, i64 1, i64 1
+  %2 = insertelement <2 x i64> %a1, i64 1, i64 1
+  %3 = call <2 x i64> @llvm.x86.pclmulqdq(<2 x i64> %1, <2 x i64> %2, i8 16)
+  ret <2 x i64> %3
+}
+
+define <2 x i64> @test_demanded_elts_pclmulqdq_17(<2 x i64> %a0, <2 x i64> %a1) {
+; CHECK-LABEL: @test_demanded_elts_pclmulqdq_17(
+; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x i64> @llvm.x86.pclmulqdq(<2 x i64> <i64 undef, i64 1>, <2 x i64> <i64 undef, i64 1>, i8 17)
+; CHECK-NEXT:    ret <2 x i64> [[TMP1]]
+;
+  %1 = insertelement <2 x i64> %a0, i64 1, i64 1
+  %2 = insertelement <2 x i64> %a1, i64 1, i64 1
+  %3 = call <2 x i64> @llvm.x86.pclmulqdq(<2 x i64> %1, <2 x i64> %2, i8 17)
+  ret <2 x i64> %3
+}
Index: lib/Transforms/InstCombine/InstCombineCalls.cpp
===================================================================
--- lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -2168,6 +2168,35 @@
     break;
   }
 
+  case Intrinsic::x86_pclmulqdq: {
+    if (auto *C = dyn_cast<ConstantInt>(II->getArgOperand(2))) {
+      unsigned Imm = C->getZExtValue();
+
+      bool MadeChange = false;
+      Value *Arg0 = II->getArgOperand(0);
+      Value *Arg1 = II->getArgOperand(1);
+      unsigned VWidth = Arg0->getType()->getVectorNumElements();
+      APInt UndefElts(VWidth, 0);
+      APInt DemandedElts(VWidth, 0);
+
+      DemandedElts = (Imm & 0x01) ? 2 : 1;
+      if (Value *V = SimplifyDemandedVectorElts(Arg0, DemandedElts,
+                                                UndefElts)) {
+        II->setArgOperand(0, V);
+        MadeChange = true;
+      }
+      DemandedElts = (Imm & 0x10) ? 2 : 1;
+      if (Value *V = SimplifyDemandedVectorElts(Arg1, DemandedElts,
+                                                UndefElts)) {
+        II->setArgOperand(1, V);
+        MadeChange = true;
+      }
+      if (MadeChange)
+        return II;
+    }
+    break;
+  }
+
   case Intrinsic::x86_sse41_insertps:
     if (Value *V = simplifyX86insertps(*II, *Builder))
       return replaceInstUsesWith(*II, V);


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D28979.85239.patch
Type: text/x-patch
Size: 3782 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170121/4cb9acc4/attachment.bin>