diff options
4 files changed, 141 insertions, 0 deletions
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp index f973949..7ec463b 100644 --- a/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp +++ b/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp @@ -183,6 +183,11 @@ WebAssemblyTargetLowering::WebAssemblyTargetLowering( for (auto T : {MVT::i32, MVT::i64}) setOperationAction(Op, T, Custom); + if (Subtarget->hasRelaxedSIMD()) { + setOperationAction( + {ISD::FMINNUM, ISD::FMINIMUMNUM, ISD::FMAXNUM, ISD::FMAXIMUMNUM}, + {MVT::v4f32, MVT::v2f64}, Legal); + } // SIMD-specific configuration if (Subtarget->hasSIMD128()) { diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td b/llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td index 7840620..f0ac26b 100644 --- a/llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td +++ b/llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td @@ -1742,6 +1742,23 @@ defm SIMD_RELAXED_FMIN : defm SIMD_RELAXED_FMAX : RelaxedBinary<F64x2, int_wasm_relaxed_max, "relaxed_max", 0x110>; +let Predicates = [HasRelaxedSIMD] in { + foreach vec = [F32x4, F64x2] in { + defvar relaxed_min = !cast<NI>("SIMD_RELAXED_FMIN_"#vec); + defvar relaxed_max = !cast<NI>("SIMD_RELAXED_FMAX_"#vec); + + // Transform standard fminimum/fmaximum to relaxed versions + def : Pat<(vec.vt (fminnum (vec.vt V128:$lhs), (vec.vt V128:$rhs))), + (relaxed_min V128:$lhs, V128:$rhs)>; + def : Pat<(vec.vt (fminimumnum (vec.vt V128:$lhs), (vec.vt V128:$rhs))), + (relaxed_min V128:$lhs, V128:$rhs)>; + def : Pat<(vec.vt (fmaxnum (vec.vt V128:$lhs), (vec.vt V128:$rhs))), + (relaxed_max V128:$lhs, V128:$rhs)>; + def : Pat<(vec.vt (fmaximumnum (vec.vt V128:$lhs), (vec.vt V128:$rhs))), + (relaxed_max V128:$lhs, V128:$rhs)>; + } +} + //===----------------------------------------------------------------------===// // Relaxed rounding q15 multiplication //===----------------------------------------------------------------------===// diff --git a/llvm/test/CodeGen/WebAssembly/simd-relaxed-fmax.ll b/llvm/test/CodeGen/WebAssembly/simd-relaxed-fmax.ll new file mode 100644 index 0000000..45f4ddd --- /dev/null +++ b/llvm/test/CodeGen/WebAssembly/simd-relaxed-fmax.ll @@ -0,0 +1,60 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 + +; RUN: llc < %s -mtriple=wasm32-unknown-unknown -mattr=+simd128,+relaxed-simd | FileCheck %s + +; Test that fmaxnum and fmaximumnum get transformed to relaxed_max + +target triple = "wasm32" + +define <4 x float> @test_maxnum_f32x4(<4 x float> %a, <4 x float> %b) { +; CHECK-LABEL: test_maxnum_f32x4: +; CHECK: .functype test_maxnum_f32x4 (v128, v128) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: f32x4.relaxed_max +; CHECK-NEXT: # fallthrough-return + %result = call <4 x float> @llvm.maxnum.v4f32(<4 x float> %a, <4 x float> %b) + ret <4 x float> %result +} + +define <4 x float> @test_maximumnum_f32x4(<4 x float> %a, <4 x float> %b) { +; CHECK-LABEL: test_maximumnum_f32x4: +; CHECK: .functype test_maximumnum_f32x4 (v128, v128) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: f32x4.relaxed_max +; CHECK-NEXT: # fallthrough-return + %result = call <4 x float> @llvm.maximumnum.v4f32(<4 x float> %a, <4 x float> %b) + ret <4 x float> %result +} + +define <2 x double> @test_maxnum_f64x2(<2 x double> %a, <2 x double> %b) { +; CHECK-LABEL: test_maxnum_f64x2: +; CHECK: .functype test_maxnum_f64x2 (v128, v128) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: f64x2.relaxed_max +; CHECK-NEXT: # fallthrough-return + %result = call <2 x double> @llvm.maxnum.v2f64(<2 x double> %a, <2 x double> %b) + ret <2 x double> %result +} + +define <2 x double> @test_minimumnum_f64x2(<2 x double> %a, <2 x double> %b) { +; CHECK-LABEL: test_minimumnum_f64x2: +; CHECK: .functype test_minimumnum_f64x2 (v128, v128) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: f64x2.relaxed_max +; CHECK-NEXT: # fallthrough-return + %result = call <2 x double> @llvm.maximumnum.v2f64(<2 x double> %a, <2 x double> %b) + ret <2 x double> %result +} + +declare <4 x float> @llvm.maxnum.v4f32(<4 x float>, <4 x float>) +declare <4 x float> @llvm.maximumnum.v4f32(<4 x float>, <4 x float>) +declare <2 x double> @llvm.maxnum.v2f64(<2 x double>, <2 x double>) +declare <2 x double> @llvm.maximumnum.v2f64(<2 x double>, <2 x double>) diff --git a/llvm/test/CodeGen/WebAssembly/simd-relaxed-fmin.ll b/llvm/test/CodeGen/WebAssembly/simd-relaxed-fmin.ll new file mode 100644 index 0000000..f3eec02 --- /dev/null +++ b/llvm/test/CodeGen/WebAssembly/simd-relaxed-fmin.ll @@ -0,0 +1,59 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc < %s -mtriple=wasm32-unknown-unknown -mattr=+simd128,+relaxed-simd | FileCheck %s + +; Test that fminnum and fminimumnum get transformed to relaxed_min + +target triple = "wasm32" + +define <4 x float> @test_minnum_f32x4(<4 x float> %a, <4 x float> %b) { +; CHECK-LABEL: test_minnum_f32x4: +; CHECK: .functype test_minnum_f32x4 (v128, v128) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: f32x4.relaxed_min +; CHECK-NEXT: # fallthrough-return + %result = call <4 x float> @llvm.minnum.v4f32(<4 x float> %a, <4 x float> %b) + ret <4 x float> %result +} + +define <4 x float> @test_minimumnum_f32x4(<4 x float> %a, <4 x float> %b) { +; CHECK-LABEL: test_minimumnum_f32x4: +; CHECK: .functype test_minimumnum_f32x4 (v128, v128) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: f32x4.relaxed_min +; CHECK-NEXT: # fallthrough-return + %result = call <4 x float> @llvm.minimumnum.v4f32(<4 x float> %a, <4 x float> %b) + ret <4 x float> %result +} + +define <2 x double> @test_minnum_f64x2(<2 x double> %a, <2 x double> %b) { +; CHECK-LABEL: test_minnum_f64x2: +; CHECK: .functype test_minnum_f64x2 (v128, v128) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: f64x2.relaxed_min +; CHECK-NEXT: # fallthrough-return + %result = call <2 x double> @llvm.minnum.v2f64(<2 x double> %a, <2 x double> %b) + ret <2 x double> %result +} + +define <2 x double> @test_minimumnum_f64x2(<2 x double> %a, <2 x double> %b) { +; CHECK-LABEL: test_minimumnum_f64x2: +; CHECK: .functype test_minimumnum_f64x2 (v128, v128) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: f64x2.relaxed_min +; CHECK-NEXT: # fallthrough-return + %result = call <2 x double> @llvm.minimumnum.v2f64(<2 x double> %a, <2 x double> %b) + ret <2 x double> %result +} + +declare <4 x float> @llvm.minnum.v4f32(<4 x float>, <4 x float>) +declare <4 x float> @llvm.fminimumnum.v4f32(<4 x float>, <4 x float>) +declare <2 x double> @llvm.minnum.v2f64(<2 x double>, <2 x double>) +declare <2 x double> @llvm.fminimumnum.v2f64(<2 x double>, <2 x double>) |