[DirectX] Add lowering for fshl

Author: inbelic

llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp

@@ -15,6 +15,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/CodeGen/Passes.h"
+#include "llvm/IR/Constants.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/InstrTypes.h"
 #include "llvm/IR/Instruction.h"
@@ -200,6 +201,7 @@ static bool isIntrinsicExpansion(Function &F) {
   case Intrinsic::assume:
   case Intrinsic::abs:
   case Intrinsic::atan2:
+  case Intrinsic::fshl:
   case Intrinsic::exp:
   case Intrinsic::is_fpclass:
   case Intrinsic::log:
@@ -656,6 +658,29 @@ static Value *expandAtan2Intrinsic(CallInst *Orig) {
   return Result;
 }
 
+static Value *expandFunnelShiftIntrinsic(CallInst *Orig) {
+  Type *Ty = Orig->getType();
+  Value *A = Orig->getOperand(0);
+  Value *B = Orig->getOperand(1);
+  Value *Shift = Orig->getOperand(2);
+
+  IRBuilder<> Builder(Orig);
+
+  unsigned BitWidth = Ty->getScalarSizeInBits();
+  Constant *Mask = ConstantInt::get(Ty, BitWidth - 1);
+  Constant *Size = ConstantInt::get(Ty, BitWidth);
+
+  // The shift is not required to be masked as DXIL op will do so automatically
+  Value *Left = Builder.CreateShl(A, Shift);
+
+  Value *MaskedShift = Builder.CreateAnd(Shift, Mask);
+  Value *InverseShift = Builder.CreateSub(Size, MaskedShift);
+  Value *Right = Builder.CreateLShr(B, InverseShift);
+
+  Value *Result = Builder.CreateOr(Left, Right);
+  return Result;
+}
+
 static Value *expandPowIntrinsic(CallInst *Orig, Intrinsic::ID IntrinsicId) {
 
   Value *X = Orig->getOperand(0);
@@ -995,6 +1020,9 @@ static bool expandIntrinsic(Function &F, CallInst *Orig) {
   case Intrinsic::atan2:
     Result = expandAtan2Intrinsic(Orig);
     break;
+  case Intrinsic::fshl:
+    Result = expandFunnelShiftIntrinsic(Orig);
+    break;
   case Intrinsic::exp:
     Result = expandExpIntrinsic(Orig);
     break;

llvm/test/CodeGen/DirectX/fshl.ll

@@ -0,0 +1,82 @@
+; RUN: opt -S -scalarizer -dxil-intrinsic-expansion -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
+; RUN: opt -S -scalarizer -dxil-intrinsic-expansion -dxil-op-lower -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
+;
+; Make sure dxil operation function calls for funnel shifts left are generated.
+
+; CHECK-LABEL: define{{.*}}@fshl_i16(
+; CHECK-SAME: i16 %[[A:.*]], i16 %[[B:.*]], i16 %[[SHIFT:.*]])
+define noundef i16 @fshl_i16(i16 %a, i16 %b, i16 %shift) {
+entry:
+; CHECK: %[[LEFT:.*]] = shl i16 %[[A]], %[[SHIFT]]
+; CHECK: %[[MASKED_SHIFT:.*]] = and i16 %[[SHIFT]], 15
+; CHECK: %[[INVERSE_SHIFT:.*]] = sub i16 16, %[[MASKED_SHIFT]]
+; CHECK: %[[RIGHT:.*]] = lshr i16 %[[B]], %[[INVERSE_SHIFT]]
+; CHECK: %[[RES:.*]] = or i16 %[[LEFT]], %[[RIGHT]]
+; CHECK: ret i16 %[[RES]]
+  %fsh = call i16 @llvm.fshl.i16(i16 %a, i16 %b, i16 %shift)
+  ret i16 %fsh
+}
+
+declare i16 @llvm.fshl.i16(i16, i16, i16)
+
+; CHECK-LABEL: define{{.*}}@fshl_v1i32(
+; CHECK-SAME: <1 x i32> %[[A_VEC:.*]], <1 x i32> %[[B_VEC:.*]], <1 x i32> %[[SHIFT_VEC:.*]])
+define noundef <1 x i32> @fshl_v1i32(<1 x i32> %a, <1 x i32> %b, <1 x i32> %shift) {
+entry:
+; CHECK: %[[A:.*]] = extractelement <1 x i32> %[[A_VEC]], i64 0
+; CHECK: %[[B:.*]] = extractelement <1 x i32> %[[B_VEC]], i64 0
+; CHECK: %[[SHIFT:.*]] = extractelement <1 x i32> %[[SHIFT_VEC]], i64 0
+; CHECK: %[[LEFT:.*]] = shl i32 %[[A]], %[[SHIFT]]
+; CHECK: %[[MASKED_SHIFT:.*]] = and i32 %[[SHIFT]], 31
+; CHECK: %[[INVERSE_SHIFT:.*]] = sub i32 32, %[[MASKED_SHIFT]]
+; CHECK: %[[RIGHT:.*]] = lshr i32 %[[B]], %[[INVERSE_SHIFT]]
+; CHECK: %[[RES:.*]] = or i32 %[[LEFT]], %[[RIGHT]]
+; CHECK: %[[RES_VEC:.*]] = insertelement <1 x i32> poison, i32 %[[RES]], i64 0
+; CHECK: ret <1 x i32> %[[RES_VEC]]
+  %fsh = call <1 x i32> @llvm.fshl.v1i32(<1 x i32> %a, <1 x i32> %b, <1 x i32> %shift)
+  ret <1 x i32> %fsh
+}
+
+declare <1 x i32> @llvm.fshl.v1i32(<1 x i32>, <1 x i32>, <1 x i32>)
+
+; CHECK-LABEL: define{{.*}}@fshl_v1i64(
+; CHECK-SAME: <3 x i64> %[[A_VEC:.*]], <3 x i64> %[[B_VEC:.*]], <3 x i64> %[[SHIFT_VEC:.*]])
+define noundef <3 x i64> @fshl_v1i64(<3 x i64> %a, <3 x i64> %b, <3 x i64> %shift) {
+entry:
+; CHECK: %[[A0:.*]] = extractelement <3 x i64> %[[A_VEC]], i64 0
+; CHECK: %[[B0:.*]] = extractelement <3 x i64> %[[B_VEC]], i64 0
+; CHECK: %[[SHIFT0:.*]] = extractelement <3 x i64> %[[SHIFT_VEC]], i64 0
+; CHECK: %[[LEFT0:.*]] = shl i64 %[[A0]], %[[SHIFT0]]
+; CHECK: %[[MASKED_SHIFT0:.*]] = and i64 %[[SHIFT0]], 63
+; CHECK: %[[INVERSE_SHIFT0:.*]] = sub i64 64, %[[MASKED_SHIFT0]]
+; CHECK: %[[RIGHT0:.*]] = lshr i64 %[[B0]], %[[INVERSE_SHIFT0]]
+; CHECK: %[[RES0:.*]] = or i64 %[[LEFT0]], %[[RIGHT0]]
+; 
+; CHECK: %[[A1:.*]] = extractelement <3 x i64> %[[A_VEC]], i64 1
+; CHECK: %[[B1:.*]] = extractelement <3 x i64> %[[B_VEC]], i64 1
+; CHECK: %[[SHIFT1:.*]] = extractelement <3 x i64> %[[SHIFT_VEC]], i64 1
+; CHECK: %[[LEFT1:.*]] = shl i64 %[[A1]], %[[SHIFT1]]
+; CHECK: %[[MASKED_SHIFT1:.*]] = and i64 %[[SHIFT1]], 63
+; CHECK: %[[INVERSE_SHIFT1:.*]] = sub i64 64, %[[MASKED_SHIFT1]]
+; CHECK: %[[RIGHT1:.*]] = lshr i64 %[[B1]], %[[INVERSE_SHIFT1]]
+; CHECK: %[[RES1:.*]] = or i64 %[[LEFT1]], %[[RIGHT1]]
+; 
+; CHECK: %[[A2:.*]] = extractelement <3 x i64> %[[A_VEC]], i64 2
+; CHECK: %[[B2:.*]] = extractelement <3 x i64> %[[B_VEC]], i64 2
+; CHECK: %[[SHIFT2:.*]] = extractelement <3 x i64> %[[SHIFT_VEC]], i64 2
+; CHECK: %[[LEFT2:.*]] = shl i64 %[[A2]], %[[SHIFT2]]
+; CHECK: %[[MASKED_SHIFT2:.*]] = and i64 %[[SHIFT2]], 63
+; CHECK: %[[INVERSE_SHIFT2:.*]] = sub i64 64, %[[MASKED_SHIFT2]]
+; CHECK: %[[RIGHT2:.*]] = lshr i64 %[[B2]], %[[INVERSE_SHIFT2]]
+; CHECK: %[[RES2:.*]] = or i64 %[[LEFT2]], %[[RIGHT2]]
+; 
+; CHECK: %[[INSERT0:.*]] = insertelement <3 x i64> poison, i64 %[[RES0]], i64 0
+; CHECK: %[[INSERT1:.*]] = insertelement <3 x i64> %[[INSERT0]], i64 %[[RES1]], i64 1
+; CHECK: %[[RES_VEC:.*]] = insertelement <3 x i64> %[[INSERT1]], i64 %[[RES2]], i64 2
+; 
+; CHECK: ret <3 x i64> %[[RES_VEC]]
+  %fsh = call <3 x i64> @llvm.fshl.v1i64(<3 x i64> %a, <3 x i64> %b, <3 x i64> %shift)
+  ret <3 x i64> %fsh
+}
+
+declare <3 x i64> @llvm.fshl.v1i64(<3 x i64>, <3 x i64>, <3 x i64>)