[AArch64] Add abs intrinsic costs

[stelios-arm]

Summary

This patch adds cost-modelling for abs vector intrinsic.

[david-arm]

LGTM!

[dmgreen]

Yeah LGTM. Thanks for the patch.

[stelios-arm]

Closed by git revision 30cb9c03b53ee03af2cdf16f4ee645e5dcff7e21.

llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp

case Intrinsic::usub_sat: {
     // This is a base cost of 1 for the vadd, plus 3 extract shifts if we
     // need to extend the type, as it uses shr(qadd(shl, shl)).
     unsigned Instrs =
         LT.second.getScalarSizeInBits() == RetTy->getScalarSizeInBits() ? 1 : 4;
     if (any_of(ValidSatTys, [&LT](MVT M) { return M == LT.second; }))
       return LT.first * Instrs;
     break;
   }
+  case Intrinsic::abs: {
+    static const auto ValidAbsTys = {MVT::v8i8,  MVT::v16i8, MVT::v4i16,
+                                     MVT::v8i16, MVT::v2i32, MVT::v4i32,
+                                     MVT::v2i64};
+    auto LT = TLI->getTypeLegalizationCost(DL, RetTy);
+    if (any_of(ValidAbsTys, [&LT](MVT M) { return M == LT.second; }))
+      return LT.first;
+    break;
+  }
   default:
     break;
   }
   return BaseT::getIntrinsicInstrCost(ICA, CostKind);
 }
 
 bool AArch64TTIImpl::isWideningInstruction(Type *DstTy, unsigned Opcode,
                                            ArrayRef<const Value *> Args) {

llvm/test/Analysis/CostModel/AArch64/abs.ll

+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt -cost-model -analyze -mtriple=aarch64-none-eabi < %s | FileCheck %s --check-prefix=RECIP
+; RUN: opt -cost-model -analyze -cost-kind=code-size -mtriple=aarch64-none-eabi < %s | FileCheck %s --check-prefix=SIZE
+
+declare <2 x i64>  @llvm.abs.v2i64(<2 x i64>, i1)
+declare <4 x i64>  @llvm.abs.v4i64(<4 x i64>, i1)
+declare <8 x i64>  @llvm.abs.v8i64(<8 x i64>, i1)
+
+declare <2 x i32>  @llvm.abs.v2i32(<2 x i32>, i1)
+declare <4 x i32>  @llvm.abs.v4i32(<4 x i32>, i1)
+declare <8 x i32>  @llvm.abs.v8i32(<8 x i32>, i1)
+declare <16 x i32> @llvm.abs.v16i32(<16 x i32>, i1)
+
+declare <2 x i16>  @llvm.abs.v2i16(<2 x i16>, i1)
+declare <4 x i16>  @llvm.abs.v4i16(<4 x i16>, i1)
+declare <8 x i16>  @llvm.abs.v8i16(<8 x i16>, i1)
+declare <16 x i16> @llvm.abs.v16i16(<16 x i16>, i1)
+declare <32 x i16> @llvm.abs.v32i16(<32 x i16>, i1)
+
+declare <2 x i8>   @llvm.abs.v2i8(<2 x i8>, i1)
+declare <4 x i8>   @llvm.abs.v4i8(<4 x i8>, i1)
+declare <8 x i8>   @llvm.abs.v8i8(<8 x i8>, i1)
+declare <16 x i8>  @llvm.abs.v16i8(<16 x i8>, i1)
+declare <32 x i8>  @llvm.abs.v32i8(<32 x i8>, i1)
+declare <64 x i8>  @llvm.abs.v64i8(<64 x i8>, i1)
+
+define i32 @abs(i32 %arg) {
+; RECIP-LABEL: 'abs'
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <2 x i32> @llvm.abs.v2i32(<2 x i32> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I8 = call <8 x i8> @llvm.abs.v8i8(<8 x i8> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.abs.v16i8(<16 x i8> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I8 = call <32 x i8> @llvm.abs.v32i8(<32 x i8> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = call <64 x i8> @llvm.abs.v64i8(<64 x i8> undef, i1 false)
+; RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SIZE-LABEL: 'abs'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <2 x i32> @llvm.abs.v2i32(<2 x i32> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I8 = call <8 x i8> @llvm.abs.v8i8(<8 x i8> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.abs.v16i8(<16 x i8> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I8 = call <32 x i8> @llvm.abs.v32i8(<32 x i8> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = call <64 x i8> @llvm.abs.v64i8(<64 x i8> undef, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+  %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> undef, i1 false)
+  %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> undef, i1 false)
+  %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> undef, i1 false)
+
+  %V2I32  = call <2 x i32>  @llvm.abs.v2i32(<2 x i32> undef, i1 false)
+  %V4I32  = call <4 x i32>  @llvm.abs.v4i32(<4 x i32> undef, i1 false)
+  %V8I32  = call <8 x i32>  @llvm.abs.v8i32(<8 x i32> undef, i1 false)
+  %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 false)
+
+  %V2I16  = call <2 x i16>  @llvm.abs.v2i16(<2 x i16> undef, i1 false)
+  %V4I16  = call <4 x i16>  @llvm.abs.v4i16(<4 x i16> undef, i1 false)
+  %V8I16  = call <8 x i16>  @llvm.abs.v8i16(<8 x i16> undef, i1 false)
+  %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
+  %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
+
+  %V8I8  = call <8 x i8>  @llvm.abs.v8i8(<8 x i8> undef, i1 false)
+  %V16I8 = call <16 x i8> @llvm.abs.v16i8(<16 x i8> undef, i1 false)
+  %V32I8 = call <32 x i8> @llvm.abs.v32i8(<32 x i8> undef, i1 false)
+  %V64I8 = call <64 x i8> @llvm.abs.v64i8(<64 x i8> undef, i1 false)
+
+  ret i32 undef
+}