[llvm] r309604 - [X86][InstCombine] Add some simplifications for BZHI intrinsics

[Craig Topper via llvm-commits]

Author: ctopper
Date: Mon Jul 31 11:52:15 2017
New Revision: 309604

URL: http://llvm.org/viewvc/llvm-project?rev=309604&view=rev
Log:
[X86][InstCombine] Add some simplifications for BZHI intrinsics

This intrinsic clears the upper bits starting at a specified index. If the index is a constant we can do some simplifications.

This could be in InstSimplify, but we don't handle any target specific intrinsics there today.

Differential Revision: https://reviews.llvm.org/D36069

Modified:
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
    llvm/trunk/test/Transforms/InstCombine/X86/x86-bmi-tbm.ll

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=309604&r1=309603&r2=309604&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Mon Jul 31 11:52:15 2017
@@ -2274,6 +2274,26 @@ Instruction *InstCombiner::visitCallInst
     }
     break;
 
+  case Intrinsic::x86_bmi_bzhi_32:
+  case Intrinsic::x86_bmi_bzhi_64:
+    // If the RHS is a constant we can try some simplifications.
+    if (auto *C = dyn_cast<ConstantInt>(II->getArgOperand(1))) {
+      uint64_t Index = C->getZExtValue() & 0xff;
+      unsigned BitWidth = II->getType()->getIntegerBitWidth();
+      if (Index >= BitWidth)
+        return replaceInstUsesWith(CI, II->getArgOperand(0));
+      if (Index == 0)
+        return replaceInstUsesWith(CI, ConstantInt::get(II->getType(), 0));
+      // If the LHS is also a constant, we can completely constant fold this.
+      if (auto *InC = dyn_cast<ConstantInt>(II->getArgOperand(0))) {
+        uint64_t Result = InC->getZExtValue();
+        Result &= maskTrailingOnes<uint64_t>(Index);
+        return replaceInstUsesWith(CI, ConstantInt::get(II->getType(), Result));
+      }
+      // TODO should we convert this to an AND if the RHS is constant?
+    }
+    break;
+
   case Intrinsic::x86_vcvtph2ps_128:
   case Intrinsic::x86_vcvtph2ps_256: {
     auto Arg = II->getArgOperand(0);

Modified: llvm/trunk/test/Transforms/InstCombine/X86/x86-bmi-tbm.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/X86/x86-bmi-tbm.ll?rev=309604&r1=309603&r2=309604&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/X86/x86-bmi-tbm.ll (original)
+++ llvm/trunk/test/Transforms/InstCombine/X86/x86-bmi-tbm.ll Mon Jul 31 11:52:15 2017
@@ -5,6 +5,8 @@ declare i32 @llvm.x86.tbm.bextri.u32(i32
 declare i64 @llvm.x86.tbm.bextri.u64(i64, i64) nounwind readnone
 declare i32 @llvm.x86.bmi.bextr.32(i32, i32) nounwind readnone
 declare i64 @llvm.x86.bmi.bextr.64(i64, i64) nounwind readnone
+declare i32 @llvm.x86.bmi.bzhi.32(i32, i32) nounwind readnone
+declare i64 @llvm.x86.bmi.bzhi.64(i64, i64) nounwind readnone
 
 define i32 @test_x86_tbm_bextri_u32(i32 %a) nounwind readnone {
 ; CHECK-LABEL: @test_x86_tbm_bextri_u32(
@@ -201,3 +203,69 @@ define i64 @test_x86_bmi_bextri_64_const
   %1 = tail call i64 @llvm.x86.bmi.bextr.64(i64 3735928559, i64 32772) ; extract bits 131:4 from 0xDEADBEEF
   ret i64 %1
 }
+
+define i32 @test_x86_bmi_bzhi_32(i32 %a) nounwind readnone {
+; CHECK-LABEL: @test_x86_bmi_bzhi_32(
+; CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.x86.bmi.bzhi.32(i32 [[A:%.*]], i32 31)
+; CHECK-NEXT:    ret i32 [[TMP1]]
+;
+  %1 = tail call i32 @llvm.x86.bmi.bzhi.32(i32 %a, i32 31)
+  ret i32 %1
+}
+
+define i32 @test_x86_bmi_bzhi_32_zero(i32 %a) nounwind readnone {
+; CHECK-LABEL: @test_x86_bmi_bzhi_32_zero(
+; CHECK-NEXT:    ret i32 0
+;
+  %1 = tail call i32 @llvm.x86.bmi.bzhi.32(i32 %a, i32 0)
+  ret i32 %1
+}
+
+define i32 @test_x86_bmi_bzhi_32_max(i32 %a) nounwind readnone {
+; CHECK-LABEL: @test_x86_bmi_bzhi_32_max(
+; CHECK-NEXT:    ret i32 [[A:%.*]]
+;
+  %1 = tail call i32 @llvm.x86.bmi.bzhi.32(i32 %a, i32 32)
+  ret i32 %1
+}
+
+define i32 @test_x86_bmi_bzhi_32_constfold() nounwind readnone {
+; CHECK-LABEL: @test_x86_bmi_bzhi_32_constfold(
+; CHECK-NEXT:    ret i32 1
+;
+  %1 = tail call i32 @llvm.x86.bmi.bzhi.32(i32 5, i32 1)
+  ret i32 %1
+}
+
+define i64 @test_x86_bmi_bzhi_64(i64 %a) nounwind readnone {
+; CHECK-LABEL: @test_x86_bmi_bzhi_64(
+; CHECK-NEXT:    [[TMP1:%.*]] = tail call i64 @llvm.x86.bmi.bzhi.64(i64 [[A:%.*]], i64 63)
+; CHECK-NEXT:    ret i64 [[TMP1]]
+;
+  %1 = tail call i64 @llvm.x86.bmi.bzhi.64(i64 %a, i64 63)
+  ret i64 %1
+}
+
+define i64 @test_x86_bmi_bzhi_64_zero(i64 %a) nounwind readnone {
+; CHECK-LABEL: @test_x86_bmi_bzhi_64_zero(
+; CHECK-NEXT:    ret i64 0
+;
+  %1 = tail call i64 @llvm.x86.bmi.bzhi.64(i64 %a, i64 0)
+  ret i64 %1
+}
+
+define i64 @test_x86_bmi_bzhi_64_max(i64 %a) nounwind readnone {
+; CHECK-LABEL: @test_x86_bmi_bzhi_64_max(
+; CHECK-NEXT:    ret i64 [[A:%.*]]
+;
+  %1 = tail call i64 @llvm.x86.bmi.bzhi.64(i64 %a, i64 64)
+  ret i64 %1
+}
+
+define i64 @test_x86_bmi_bzhi_64_constfold() nounwind readnone {
+; CHECK-LABEL: @test_x86_bmi_bzhi_64_constfold(
+; CHECK-NEXT:    ret i64 1
+;
+  %1 = tail call i64 @llvm.x86.bmi.bzhi.64(i64 5, i64 1)
+  ret i64 %1
+}