Index: llvm/trunk/include/llvm/IR/IntrinsicsWebAssembly.td =================================================================== --- llvm/trunk/include/llvm/IR/IntrinsicsWebAssembly.td +++ llvm/trunk/include/llvm/IR/IntrinsicsWebAssembly.td @@ -91,6 +91,22 @@ // SIMD intrinsics //===----------------------------------------------------------------------===// +def int_wasm_add_saturate_signed : + Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem, IntrSpeculatable]>; +def int_wasm_add_saturate_unsigned : + Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem, IntrSpeculatable]>; +def int_wasm_sub_saturate_signed : + Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem, IntrSpeculatable]>; +def int_wasm_sub_saturate_unsigned : + Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem, IntrSpeculatable]>; def int_wasm_bitselect : Intrinsic<[llvm_anyvector_ty], [LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>], Index: llvm/trunk/lib/Target/WebAssembly/WebAssemblyISD.def =================================================================== --- llvm/trunk/lib/Target/WebAssembly/WebAssemblyISD.def +++ llvm/trunk/lib/Target/WebAssembly/WebAssemblyISD.def @@ -25,5 +25,9 @@ HANDLE_NODETYPE(ANYTRUE) HANDLE_NODETYPE(ALLTRUE) HANDLE_NODETYPE(BITSELECT) +HANDLE_NODETYPE(ADD_SAT_S) +HANDLE_NODETYPE(ADD_SAT_U) +HANDLE_NODETYPE(SUB_SAT_S) +HANDLE_NODETYPE(SUB_SAT_U) // add memory opcodes starting at ISD::FIRST_TARGET_MEMORY_OPCODE here... Index: llvm/trunk/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp =================================================================== --- llvm/trunk/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp +++ llvm/trunk/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp @@ -966,6 +966,32 @@ default: return {}; // Don't custom lower most intrinsics. + case Intrinsic::wasm_add_saturate_signed: + case Intrinsic::wasm_add_saturate_unsigned: + case Intrinsic::wasm_sub_saturate_signed: + case Intrinsic::wasm_sub_saturate_unsigned: { + unsigned OpCode; + switch (IntNo) { + case Intrinsic::wasm_add_saturate_signed: + OpCode = WebAssemblyISD::ADD_SAT_S; + break; + case Intrinsic::wasm_add_saturate_unsigned: + OpCode = WebAssemblyISD::ADD_SAT_U; + break; + case Intrinsic::wasm_sub_saturate_signed: + OpCode = WebAssemblyISD::SUB_SAT_S; + break; + case Intrinsic::wasm_sub_saturate_unsigned: + OpCode = WebAssemblyISD::SUB_SAT_U; + break; + default: + llvm_unreachable("unexpected intrinsic id"); + break; + } + return DAG.getNode(OpCode, DL, Op.getValueType(), Op.getOperand(1), + Op.getOperand(2)); + } + case Intrinsic::wasm_bitselect: return DAG.getNode(WebAssemblyISD::BITSELECT, DL, Op.getValueType(), Op.getOperand(1), Op.getOperand(2), Op.getOperand(3)); Index: llvm/trunk/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td =================================================================== --- llvm/trunk/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td +++ llvm/trunk/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td @@ -20,11 +20,18 @@ // Custom nodes for custom operations def wasm_shuffle_t : SDTypeProfile<1, 18, []>; +def wasm_saturate_t : SDTypeProfile<1, 2, + [SDTCisVec<0>, SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>] +>; def wasm_bitselect_t : SDTypeProfile<1, 3, [SDTCisVec<0>, SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>, SDTCisSameAs<0, 3>] >; def wasm_reduce_t : SDTypeProfile<1, 1, [SDTCisVT<0, i32>, SDTCisVec<1>]>; def wasm_shuffle : SDNode<"WebAssemblyISD::SHUFFLE", wasm_shuffle_t>; +def wasm_add_sat_s : SDNode<"WebAssemblyISD::ADD_SAT_S", wasm_saturate_t>; +def wasm_add_sat_u : SDNode<"WebAssemblyISD::ADD_SAT_U", wasm_saturate_t>; +def wasm_sub_sat_s : SDNode<"WebAssemblyISD::SUB_SAT_S", wasm_saturate_t>; +def wasm_sub_sat_u : SDNode<"WebAssemblyISD::SUB_SAT_U", wasm_saturate_t>; def wasm_bitselect : SDNode<"WebAssemblyISD::BITSELECT", wasm_bitselect_t>; def wasm_anytrue : SDNode<"WebAssemblyISD::ANYTRUE", wasm_reduce_t>; def wasm_alltrue : SDNode<"WebAssemblyISD::ALLTRUE", wasm_reduce_t>; @@ -139,6 +146,10 @@ defm "" : SIMDBinary; defm "" : SIMDBinary; } +multiclass SIMDBinarySat baseInst> { + defm "" : SIMDBinary; + defm "" : SIMDBinary; +} multiclass SIMDShift simdop> { defm _#vec_t : SIMD_I<(outs V128:$dst), (ins V128:$vec, I32:$x), @@ -366,10 +377,14 @@ defm ADD : SIMDBinaryFP; defm MUL : SIMDBinaryIntNoI64x2; defm MUL : SIMDBinaryFP; +defm ADD_SAT_S : SIMDBinarySat; +defm ADD_SAT_U : SIMDBinarySat; } // isCommutable = 1 defm SUB : SIMDBinaryInt; defm SUB : SIMDBinaryFP; +defm SUB_SAT_S : SIMDBinarySat; +defm SUB_SAT_U : SIMDBinarySat; defm DIV : SIMDBinaryFP; defm "" : SIMDNegInt; Index: llvm/trunk/test/CodeGen/WebAssembly/simd-arith.ll =================================================================== --- llvm/trunk/test/CodeGen/WebAssembly/simd-arith.ll +++ llvm/trunk/test/CodeGen/WebAssembly/simd-arith.ll @@ -1,9 +1,9 @@ ; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -wasm-enable-unimplemented-simd -mattr=+simd128 | FileCheck %s --check-prefixes CHECK,SIMD128,SIMD128-SLOW ; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -wasm-enable-unimplemented-simd -mattr=+simd128 -fast-isel | FileCheck %s --check-prefixes CHECK,SIMD128,SIMD128-FAST -; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -mattr=+simd128 | FileCheck %s --check-prefixes CHECK,SIMD128-VM,SIMD128-VM-SLOW -; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -mattr=+simd128 -fast-isel | FileCheck %s --check-prefixes CHECK,SIMD128-VM,SIMD128-VM-FAST -; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -mattr=-simd128 | FileCheck %s --check-prefixes CHECK,NO-SIMD128,NO-SIMD128-SLOW -; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -mattr=-simd128 -fast-isel | FileCheck %s --check-prefixes CHECK,NO-SIMD128,NO-SIMD128-FAST +; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -mattr=+simd128 | FileCheck %s --check-prefixes CHECK,SIMD128-VM +; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -mattr=+simd128 -fast-isel | FileCheck %s --check-prefixes CHECK,SIMD128-VM +; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -mattr=-simd128 | FileCheck %s --check-prefixes CHECK,NO-SIMD128 +; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -mattr=-simd128 -fast-isel | FileCheck %s --check-prefixes CHECK,NO-SIMD128 ; Test that basic SIMD128 arithmetic operations assemble as expected. Index: llvm/trunk/test/CodeGen/WebAssembly/simd-intrinsics.ll =================================================================== --- llvm/trunk/test/CodeGen/WebAssembly/simd-intrinsics.ll +++ llvm/trunk/test/CodeGen/WebAssembly/simd-intrinsics.ll @@ -11,6 +11,58 @@ ; ============================================================================== ; 16 x i8 ; ============================================================================== +; CHECK-LABEL: add_sat_s_v16i8: +; SIMD128-NEXT: .param v128, v128{{$}} +; SIMD128-NEXT: .result v128{{$}} +; SIMD128-NEXT: i8x16.add_saturate_s $push[[R:[0-9]+]]=, $0, $1{{$}} +; SIMD128-NEXT: return $pop[[R]]{{$}} +declare <16 x i8> @llvm.wasm.add.saturate.signed.v16i8(<16 x i8>, <16 x i8>) +define <16 x i8> @add_sat_s_v16i8(<16 x i8> %x, <16 x i8> %y) { + %a = call <16 x i8> @llvm.wasm.add.saturate.signed.v16i8( + <16 x i8> %x, <16 x i8> %y + ) + ret <16 x i8> %a +} + +; CHECK-LABEL: add_sat_u_v16i8: +; SIMD128-NEXT: .param v128, v128{{$}} +; SIMD128-NEXT: .result v128{{$}} +; SIMD128-NEXT: i8x16.add_saturate_u $push[[R:[0-9]+]]=, $0, $1{{$}} +; SIMD128-NEXT: return $pop[[R]]{{$}} +declare <16 x i8> @llvm.wasm.add.saturate.unsigned.v16i8(<16 x i8>, <16 x i8>) +define <16 x i8> @add_sat_u_v16i8(<16 x i8> %x, <16 x i8> %y) { + %a = call <16 x i8> @llvm.wasm.add.saturate.unsigned.v16i8( + <16 x i8> %x, <16 x i8> %y + ) + ret <16 x i8> %a +} + +; CHECK-LABEL: sub_sat_s_v16i8: +; SIMD128-NEXT: .param v128, v128{{$}} +; SIMD128-NEXT: .result v128{{$}} +; SIMD128-NEXT: i8x16.sub_saturate_s $push[[R:[0-9]+]]=, $0, $1{{$}} +; SIMD128-NEXT: return $pop[[R]]{{$}} +declare <16 x i8> @llvm.wasm.sub.saturate.signed.v16i8(<16 x i8>, <16 x i8>) +define <16 x i8> @sub_sat_s_v16i8(<16 x i8> %x, <16 x i8> %y) { + %a = call <16 x i8> @llvm.wasm.sub.saturate.signed.v16i8( + <16 x i8> %x, <16 x i8> %y + ) + ret <16 x i8> %a +} + +; CHECK-LABEL: sub_sat_u_v16i8: +; SIMD128-NEXT: .param v128, v128{{$}} +; SIMD128-NEXT: .result v128{{$}} +; SIMD128-NEXT: i8x16.sub_saturate_u $push[[R:[0-9]+]]=, $0, $1{{$}} +; SIMD128-NEXT: return $pop[[R]]{{$}} +declare <16 x i8> @llvm.wasm.sub.saturate.unsigned.v16i8(<16 x i8>, <16 x i8>) +define <16 x i8> @sub_sat_u_v16i8(<16 x i8> %x, <16 x i8> %y) { + %a = call <16 x i8> @llvm.wasm.sub.saturate.unsigned.v16i8( + <16 x i8> %x, <16 x i8> %y + ) + ret <16 x i8> %a +} + ; CHECK-LABEL: any_v16i8: ; SIMD128-NEXT: .param v128{{$}} ; SIMD128-NEXT: .result i32{{$}} @@ -49,6 +101,58 @@ ; ============================================================================== ; 8 x i16 ; ============================================================================== +; CHECK-LABEL: add_sat_s_v8i16: +; SIMD128-NEXT: .param v128, v128{{$}} +; SIMD128-NEXT: .result v128{{$}} +; SIMD128-NEXT: i16x8.add_saturate_s $push[[R:[0-9]+]]=, $0, $1{{$}} +; SIMD128-NEXT: return $pop[[R]]{{$}} +declare <8 x i16> @llvm.wasm.add.saturate.signed.v8i16(<8 x i16>, <8 x i16>) +define <8 x i16> @add_sat_s_v8i16(<8 x i16> %x, <8 x i16> %y) { + %a = call <8 x i16> @llvm.wasm.add.saturate.signed.v8i16( + <8 x i16> %x, <8 x i16> %y + ) + ret <8 x i16> %a +} + +; CHECK-LABEL: add_sat_u_v8i16: +; SIMD128-NEXT: .param v128, v128{{$}} +; SIMD128-NEXT: .result v128{{$}} +; SIMD128-NEXT: i16x8.add_saturate_u $push[[R:[0-9]+]]=, $0, $1{{$}} +; SIMD128-NEXT: return $pop[[R]]{{$}} +declare <8 x i16> @llvm.wasm.add.saturate.unsigned.v8i16(<8 x i16>, <8 x i16>) +define <8 x i16> @add_sat_u_v8i16(<8 x i16> %x, <8 x i16> %y) { + %a = call <8 x i16> @llvm.wasm.add.saturate.unsigned.v8i16( + <8 x i16> %x, <8 x i16> %y + ) + ret <8 x i16> %a +} + +; CHECK-LABEL: sub_sat_s_v8i16: +; SIMD128-NEXT: .param v128, v128{{$}} +; SIMD128-NEXT: .result v128{{$}} +; SIMD128-NEXT: i16x8.sub_saturate_s $push[[R:[0-9]+]]=, $0, $1{{$}} +; SIMD128-NEXT: return $pop[[R]]{{$}} +declare <8 x i16> @llvm.wasm.sub.saturate.signed.v8i16(<8 x i16>, <8 x i16>) +define <8 x i16> @sub_sat_s_v8i16(<8 x i16> %x, <8 x i16> %y) { + %a = call <8 x i16> @llvm.wasm.sub.saturate.signed.v8i16( + <8 x i16> %x, <8 x i16> %y + ) + ret <8 x i16> %a +} + +; CHECK-LABEL: sub_sat_u_v8i16: +; SIMD128-NEXT: .param v128, v128{{$}} +; SIMD128-NEXT: .result v128{{$}} +; SIMD128-NEXT: i16x8.sub_saturate_u $push[[R:[0-9]+]]=, $0, $1{{$}} +; SIMD128-NEXT: return $pop[[R]]{{$}} +declare <8 x i16> @llvm.wasm.sub.saturate.unsigned.v8i16(<8 x i16>, <8 x i16>) +define <8 x i16> @sub_sat_u_v8i16(<8 x i16> %x, <8 x i16> %y) { + %a = call <8 x i16> @llvm.wasm.sub.saturate.unsigned.v8i16( + <8 x i16> %x, <8 x i16> %y + ) + ret <8 x i16> %a +} + ; CHECK-LABEL: any_v8i16: ; SIMD128-NEXT: .param v128{{$}} ; SIMD128-NEXT: .result i32{{$}} Index: llvm/trunk/test/MC/WebAssembly/simd-encodings.s =================================================================== --- llvm/trunk/test/MC/WebAssembly/simd-encodings.s +++ llvm/trunk/test/MC/WebAssembly/simd-encodings.s @@ -145,6 +145,30 @@ # CHECK: i64x2.neg # encoding: [0xfd,0x27] i64x2.neg + # CHECK: i8x16.add_saturate_s # encoding: [0xfd,0x28] + i8x16.add_saturate_s + + # CHECK: i8x16.add_saturate_u # encoding: [0xfd,0x29] + i8x16.add_saturate_u + + # CHECK: i16x8.add_saturate_s # encoding: [0xfd,0x2a] + i16x8.add_saturate_s + + # CHECK: i16x8.add_saturate_u # encoding: [0xfd,0x2b] + i16x8.add_saturate_u + + # CHECK: i8x16.sub_saturate_s # encoding: [0xfd,0x2c] + i8x16.sub_saturate_s + + # CHECK: i8x16.sub_saturate_u # encoding: [0xfd,0x2d] + i8x16.sub_saturate_u + + # CHECK: i16x8.sub_saturate_s # encoding: [0xfd,0x2e] + i16x8.sub_saturate_s + + # CHECK: i16x8.sub_saturate_u # encoding: [0xfd,0x2f] + i16x8.sub_saturate_u + # CHECK: i8x16.shl # encoding: [0xfd,0x30] i8x16.shl