diff --git a/llvm/test/CodeGen/WebAssembly/simd-offset.ll b/llvm/test/CodeGen/WebAssembly/simd-offset.ll --- a/llvm/test/CodeGen/WebAssembly/simd-offset.ll +++ b/llvm/test/CodeGen/WebAssembly/simd-offset.ll @@ -1,5 +1,5 @@ -; RUN: llc < %s -asm-verbose=false -verify-machineinstrs -disable-wasm-fallthrough-return-opt -wasm-keep-registers -wasm-disable-explicit-locals -mattr=+simd128 | FileCheck %s --check-prefixes CHECK,SIMD128 -; RUN: llc < %s -asm-verbose=false -verify-machineinstrs -disable-wasm-fallthrough-return-opt -wasm-keep-registers -wasm-disable-explicit-locals | FileCheck %s --check-prefixes CHECK,NO-SIMD128 +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc < %s -verify-machineinstrs -mattr=+simd128 | FileCheck %s ; Test SIMD loads and stores @@ -9,34 +9,37 @@ ; ============================================================================== ; 16 x i8 ; ============================================================================== -; CHECK-LABEL: load_v16i8: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v16i8 (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_v16i8(<16 x i8>* %p) { +; CHECK-LABEL: load_v16i8: +; CHECK: .functype load_v16i8 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %v = load <16 x i8>, <16 x i8>* %p ret <16 x i8> %v } -; CHECK-LABEL: load_splat_v16i8: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v16i8 (i32) -> (v128){{$}} -; SIMD128-NEXT: v8x16.load_splat $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_splat_v16i8(i8* %p) { +; CHECK-LABEL: load_splat_v16i8: +; CHECK: .functype load_splat_v16i8 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v8x16.load_splat 0 +; CHECK-NEXT: # fallthrough-return %e = load i8, i8* %p %v1 = insertelement <16 x i8> undef, i8 %e, i32 0 %v2 = shufflevector <16 x i8> %v1, <16 x i8> undef, <16 x i32> zeroinitializer ret <16 x i8> %v2 } -; CHECK-LABEL: load_v16i8_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v16i8_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_v16i8_with_folded_offset(<16 x i8>* %p) { +; CHECK-LABEL: load_v16i8_with_folded_offset: +; CHECK: .functype load_v16i8_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <16 x i8>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <16 x i8>* @@ -44,12 +47,13 @@ ret <16 x i8> %v } -; CHECK-LABEL: load_splat_v16i8_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v16i8_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v8x16.load_splat $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_splat_v16i8_with_folded_offset(i8* %p) { +; CHECK-LABEL: load_splat_v16i8_with_folded_offset: +; CHECK: .functype load_splat_v16i8_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v8x16.load_splat 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint i8* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to i8* @@ -59,23 +63,25 @@ ret <16 x i8> %v2 } -; CHECK-LABEL: load_v16i8_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v16i8_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_v16i8_with_folded_gep_offset(<16 x i8>* %p) { +; CHECK-LABEL: load_v16i8_with_folded_gep_offset: +; CHECK: .functype load_v16i8_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <16 x i8>, <16 x i8>* %p, i32 1 %v = load <16 x i8>, <16 x i8>* %s ret <16 x i8> %v } -; CHECK-LABEL: load_splat_v16i8_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v16i8_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v8x16.load_splat $push[[R:[0-9]+]]=, 1($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_splat_v16i8_with_folded_gep_offset(i8* %p) { +; CHECK-LABEL: load_splat_v16i8_with_folded_gep_offset: +; CHECK: .functype load_splat_v16i8_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v8x16.load_splat 1 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds i8, i8* %p, i32 1 %e = load i8, i8* %s %v1 = insertelement <16 x i8> undef, i8 %e, i32 0 @@ -83,27 +89,29 @@ ret <16 x i8> %v2 } -; CHECK-LABEL: load_v16i8_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v16i8_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_v16i8_with_unfolded_gep_negative_offset(<16 x i8>* %p) { +; CHECK-LABEL: load_v16i8_with_unfolded_gep_negative_offset: +; CHECK: .functype load_v16i8_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <16 x i8>, <16 x i8>* %p, i32 -1 %v = load <16 x i8>, <16 x i8>* %s ret <16 x i8> %v } -; CHECK-LABEL: load_splat_v16i8_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v16i8_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -1{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v8x16.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_splat_v16i8_with_unfolded_gep_negative_offset(i8* %p) { +; CHECK-LABEL: load_splat_v16i8_with_unfolded_gep_negative_offset: +; CHECK: .functype load_splat_v16i8_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -1 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v8x16.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds i8, i8* %p, i32 -1 %e = load i8, i8* %s %v1 = insertelement <16 x i8> undef, i8 %e, i32 0 @@ -111,14 +119,15 @@ ret <16 x i8> %v2 } -; CHECK-LABEL: load_v16i8_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v16i8_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_v16i8_with_unfolded_offset(<16 x i8>* %p) { +; CHECK-LABEL: load_v16i8_with_unfolded_offset: +; CHECK: .functype load_v16i8_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <16 x i8>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <16 x i8>* @@ -126,14 +135,15 @@ ret <16 x i8> %v } -; CHECK-LABEL: load_splat_v16i8_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v16i8_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v8x16.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_splat_v16i8_with_unfolded_offset(i8* %p) { +; CHECK-LABEL: load_splat_v16i8_with_unfolded_offset: +; CHECK: .functype load_splat_v16i8_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v8x16.load_splat 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint i8* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to i8* @@ -143,27 +153,29 @@ ret <16 x i8> %v2 } -; CHECK-LABEL: load_v16i8_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v16i8_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_v16i8_with_unfolded_gep_offset(<16 x i8>* %p) { +; CHECK-LABEL: load_v16i8_with_unfolded_gep_offset: +; CHECK: .functype load_v16i8_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <16 x i8>, <16 x i8>* %p, i32 1 %v = load <16 x i8>, <16 x i8>* %s ret <16 x i8> %v } -; CHECK-LABEL: load_splat_v16i8_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v16i8_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 1{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v8x16.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_splat_v16i8_with_unfolded_gep_offset(i8* %p) { +; CHECK-LABEL: load_splat_v16i8_with_unfolded_gep_offset: +; CHECK: .functype load_splat_v16i8_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 1 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v8x16.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr i8, i8* %p, i32 1 %e = load i8, i8* %s %v1 = insertelement <16 x i8> undef, i8 %e, i32 0 @@ -171,25 +183,25 @@ ret <16 x i8> %v2 } -; CHECK-LABEL: load_v16i8_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v16i8_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_v16i8_from_numeric_address() { +; CHECK-LABEL: load_v16i8_from_numeric_address: +; CHECK: .functype load_v16i8_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <16 x i8>* %v = load <16 x i8>, <16 x i8>* %s ret <16 x i8> %v } -; CHECK-LABEL: load_splat_v16i8_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v16i8_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v8x16.load_splat $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <16 x i8> @load_splat_v16i8_from_numeric_address() { +; CHECK-LABEL: load_splat_v16i8_from_numeric_address: +; CHECK: .functype load_splat_v16i8_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v8x16.load_splat 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to i8* %e = load i8, i8* %s %v1 = insertelement <16 x i8> undef, i8 %e, i32 0 @@ -197,46 +209,52 @@ ret <16 x i8> %v2 } -; CHECK-LABEL: load_v16i8_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v16i8_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, gv_v16i8($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_v16i8 = global <16 x i8> define <16 x i8> @load_v16i8_from_global_address() { +; CHECK-LABEL: load_v16i8_from_global_address: +; CHECK: .functype load_v16i8_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load gv_v16i8 +; CHECK-NEXT: # fallthrough-return %v = load <16 x i8>, <16 x i8>* @gv_v16i8 ret <16 x i8> %v } -; CHECK-LABEL: load_splat_v16i8_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v16i8_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v8x16.load_splat $push[[R:[0-9]+]]=, gv_i8($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_i8 = global i8 42 define <16 x i8> @load_splat_v16i8_from_global_address() { +; CHECK-LABEL: load_splat_v16i8_from_global_address: +; CHECK: .functype load_splat_v16i8_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v8x16.load_splat gv_i8 +; CHECK-NEXT: # fallthrough-return %e = load i8, i8* @gv_i8 %v1 = insertelement <16 x i8> undef, i8 %e, i32 0 %v2 = shufflevector <16 x i8> %v1, <16 x i8> undef, <16 x i32> zeroinitializer ret <16 x i8> %v2 } -; CHECK-LABEL: store_v16i8: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v16i8 (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 0($1), $0{{$}} define void @store_v16i8(<16 x i8> %v, <16 x i8>* %p) { +; CHECK-LABEL: store_v16i8: +; CHECK: .functype store_v16i8 (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return store <16 x i8> %v , <16 x i8>* %p ret void } -; CHECK-LABEL: store_v16i8_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v16i8_with_folded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v16i8_with_folded_offset(<16 x i8> %v, <16 x i8>* %p) { +; CHECK-LABEL: store_v16i8_with_folded_offset: +; CHECK: .functype store_v16i8_with_folded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <16 x i8>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <16 x i8>* @@ -244,69 +262,85 @@ ret void } -; CHECK-LABEL: store_v16i8_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v16i8_with_folded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v16i8_with_folded_gep_offset(<16 x i8> %v, <16 x i8>* %p) { +; CHECK-LABEL: store_v16i8_with_folded_gep_offset: +; CHECK: .functype store_v16i8_with_folded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <16 x i8>, <16 x i8>* %p, i32 1 store <16 x i8> %v , <16 x i8>* %s ret void } -; CHECK-LABEL: store_v16i8_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v16i8_with_unfolded_gep_negative_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v16i8_with_unfolded_gep_negative_offset(<16 x i8> %v, <16 x i8>* %p) { +; CHECK-LABEL: store_v16i8_with_unfolded_gep_negative_offset: +; CHECK: .functype store_v16i8_with_unfolded_gep_negative_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <16 x i8>, <16 x i8>* %p, i32 -1 store <16 x i8> %v , <16 x i8>* %s ret void } -; CHECK-LABEL: store_v16i8_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v16i8_with_unfolded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v16i8_with_unfolded_offset(<16 x i8> %v, <16 x i8>* %p) { +; CHECK-LABEL: store_v16i8_with_unfolded_offset: +; CHECK: .functype store_v16i8_with_unfolded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <16 x i8>, <16 x i8>* %p, i32 -1 store <16 x i8> %v , <16 x i8>* %s ret void } -; CHECK-LABEL: store_v16i8_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v16i8_with_unfolded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v16i8_with_unfolded_gep_offset(<16 x i8> %v, <16 x i8>* %p) { +; CHECK-LABEL: store_v16i8_with_unfolded_gep_offset: +; CHECK: .functype store_v16i8_with_unfolded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <16 x i8>, <16 x i8>* %p, i32 1 store <16 x i8> %v , <16 x i8>* %s ret void } -; CHECK-LABEL: store_v16i8_to_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v16i8_to_numeric_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[R:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store 32($pop[[R]]), $0{{$}} define void @store_v16i8_to_numeric_address(<16 x i8> %v) { +; CHECK-LABEL: store_v16i8_to_numeric_address: +; CHECK: .functype store_v16i8_to_numeric_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <16 x i8>* store <16 x i8> %v , <16 x i8>* %s ret void } -; CHECK-LABEL: store_v16i8_to_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v16i8_to_global_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[R:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store gv_v16i8($pop[[R]]), $0{{$}} define void @store_v16i8_to_global_address(<16 x i8> %v) { +; CHECK-LABEL: store_v16i8_to_global_address: +; CHECK: .functype store_v16i8_to_global_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store gv_v16i8 +; CHECK-NEXT: # fallthrough-return store <16 x i8> %v , <16 x i8>* @gv_v16i8 ret void } @@ -314,66 +348,72 @@ ; ============================================================================== ; 8 x i16 ; ============================================================================== -; CHECK-LABEL: load_v8i16: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v8i16 (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_v8i16(<8 x i16>* %p) { +; CHECK-LABEL: load_v8i16: +; CHECK: .functype load_v8i16 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %v = load <8 x i16>, <8 x i16>* %p ret <8 x i16> %v } -; CHECK-LABEL: load_splat_v8i16: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v8i16 (i32) -> (v128){{$}} -; SIMD128-NEXT: v16x8.load_splat $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_splat_v8i16(i16* %p) { +; CHECK-LABEL: load_splat_v8i16: +; CHECK: .functype load_splat_v8i16 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v16x8.load_splat 0 +; CHECK-NEXT: # fallthrough-return %e = load i16, i16* %p %v1 = insertelement <8 x i16> undef, i16 %e, i32 0 %v2 = shufflevector <8 x i16> %v1, <8 x i16> undef, <8 x i32> zeroinitializer ret <8 x i16> %v2 } -; CHECK-LABEL: load_sext_v8i16: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v8i16 (i32) -> (v128){{$}} -; SIMD128-NEXT: i16x8.load8x8_s $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_sext_v8i16(<8 x i8>* %p) { +; CHECK-LABEL: load_sext_v8i16: +; CHECK: .functype load_sext_v8i16 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i16x8.load8x8_s 0 +; CHECK-NEXT: # fallthrough-return %v = load <8 x i8>, <8 x i8>* %p %v2 = sext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_zext_v8i16: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v8i16 (i32) -> (v128){{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_zext_v8i16(<8 x i8>* %p) { +; CHECK-LABEL: load_zext_v8i16: +; CHECK: .functype load_zext_v8i16 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i16x8.load8x8_u 0 +; CHECK-NEXT: # fallthrough-return %v = load <8 x i8>, <8 x i8>* %p %v2 = zext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_ext_v8i16: -; NO-SIMD128-NOT: load8x8 -; SIMD128-NEXT: .functype load_ext_v8i16 (i32) -> (v128){{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i8> @load_ext_v8i16(<8 x i8>* %p) { +; CHECK-LABEL: load_ext_v8i16: +; CHECK: .functype load_ext_v8i16 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i16x8.load8x8_u 0 +; CHECK-NEXT: # fallthrough-return %v = load <8 x i8>, <8 x i8>* %p ret <8 x i8> %v } -; CHECK-LABEL: load_v8i16_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v8i16_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_v8i16_with_folded_offset(<8 x i16>* %p) { +; CHECK-LABEL: load_v8i16_with_folded_offset: +; CHECK: .functype load_v8i16_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <8 x i16>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <8 x i16>* @@ -381,12 +421,13 @@ ret <8 x i16> %v } -; CHECK-LABEL: load_splat_v8i16_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v8i16_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v16x8.load_splat $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_splat_v8i16_with_folded_offset(i16* %p) { +; CHECK-LABEL: load_splat_v8i16_with_folded_offset: +; CHECK: .functype load_splat_v8i16_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v16x8.load_splat 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint i16* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to i16* @@ -396,12 +437,13 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_sext_v8i16_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v8i16_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i16x8.load8x8_s $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_sext_v8i16_with_folded_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_sext_v8i16_with_folded_offset: +; CHECK: .functype load_sext_v8i16_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i16x8.load8x8_s 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <8 x i8>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <8 x i8>* @@ -410,12 +452,13 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_zext_v8i16_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v8i16_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_zext_v8i16_with_folded_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_zext_v8i16_with_folded_offset: +; CHECK: .functype load_zext_v8i16_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i16x8.load8x8_u 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <8 x i8>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <8 x i8>* @@ -424,12 +467,13 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_ext_v8i16_with_folded_offset: -; NO-SIMD128-NOT: load8x8 -; SIMD128-NEXT: .functype load_ext_v8i16_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i8> @load_ext_v8i16_with_folded_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_ext_v8i16_with_folded_offset: +; CHECK: .functype load_ext_v8i16_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i16x8.load8x8_u 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <8 x i8>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <8 x i8>* @@ -437,23 +481,25 @@ ret <8 x i8> %v } -; CHECK-LABEL: load_v8i16_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v8i16_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_v8i16_with_folded_gep_offset(<8 x i16>* %p) { +; CHECK-LABEL: load_v8i16_with_folded_gep_offset: +; CHECK: .functype load_v8i16_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i16>, <8 x i16>* %p, i32 1 %v = load <8 x i16>, <8 x i16>* %s ret <8 x i16> %v } -; CHECK-LABEL: load_splat_v8i16_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v8i16_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v16x8.load_splat $push[[R:[0-9]+]]=, 2($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_splat_v8i16_with_folded_gep_offset(i16* %p) { +; CHECK-LABEL: load_splat_v8i16_with_folded_gep_offset: +; CHECK: .functype load_splat_v8i16_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v16x8.load_splat 2 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds i16, i16* %p, i32 1 %e = load i16, i16* %s %v1 = insertelement <8 x i16> undef, i16 %e, i32 0 @@ -461,62 +507,67 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_sext_v8i16_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v8i16_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i16x8.load8x8_s $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_sext_v8i16_with_folded_gep_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_sext_v8i16_with_folded_gep_offset: +; CHECK: .functype load_sext_v8i16_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i16x8.load8x8_s 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i8>, <8 x i8>* %p, i32 1 %v = load <8 x i8>, <8 x i8>* %s %v2 = sext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_zext_v8i16_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v8i16_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_zext_v8i16_with_folded_gep_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_zext_v8i16_with_folded_gep_offset: +; CHECK: .functype load_zext_v8i16_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i16x8.load8x8_u 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i8>, <8 x i8>* %p, i32 1 %v = load <8 x i8>, <8 x i8>* %s %v2 = zext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_ext_v8i16_with_folded_gep_offset: -; NO-SIMD128-NOT: load8x8 -; SIMD128-NEXT: .functype load_ext_v8i16_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i8> @load_ext_v8i16_with_folded_gep_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_ext_v8i16_with_folded_gep_offset: +; CHECK: .functype load_ext_v8i16_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i16x8.load8x8_u 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i8>, <8 x i8>* %p, i32 1 %v = load <8 x i8>, <8 x i8>* %s ret <8 x i8> %v } -; CHECK-LABEL: load_v8i16_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_v8i16_with_unfolded_gep_negative_offset(<8 x i16>* %p) { +; CHECK-LABEL: load_v8i16_with_unfolded_gep_negative_offset: +; CHECK: .functype load_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i16>, <8 x i16>* %p, i32 -1 %v = load <8 x i16>, <8 x i16>* %s ret <8 x i16> %v } -; CHECK-LABEL: load_splat_v8i16_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -2{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v16x8.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_splat_v8i16_with_unfolded_gep_negative_offset(i16* %p) { +; CHECK-LABEL: load_splat_v8i16_with_unfolded_gep_negative_offset: +; CHECK: .functype load_splat_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -2 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v16x8.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds i16, i16* %p, i32 -1 %e = load i16, i16* %s %v1 = insertelement <8 x i16> undef, i16 %e, i32 0 @@ -524,55 +575,59 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_sext_v8i16_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i16x8.load8x8_s $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_sext_v8i16_with_unfolded_gep_negative_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_sext_v8i16_with_unfolded_gep_negative_offset: +; CHECK: .functype load_sext_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i16x8.load8x8_s 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i8>, <8 x i8>* %p, i32 -1 %v = load <8 x i8>, <8 x i8>* %s %v2 = sext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_zext_v8i16_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_zext_v8i16_with_unfolded_gep_negative_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_zext_v8i16_with_unfolded_gep_negative_offset: +; CHECK: .functype load_zext_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i16x8.load8x8_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i8>, <8 x i8>* %p, i32 -1 %v = load <8 x i8>, <8 x i8>* %s %v2 = zext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_ext_v8i16_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: load8x8 -; SIMD128-NEXT: .functype load_ext_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i8> @load_ext_v8i16_with_unfolded_gep_negative_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_ext_v8i16_with_unfolded_gep_negative_offset: +; CHECK: .functype load_ext_v8i16_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i16x8.load8x8_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i8>, <8 x i8>* %p, i32 -1 %v = load <8 x i8>, <8 x i8>* %s ret <8 x i8> %v } -; CHECK-LABEL: load_v8i16_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v8i16_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[L0:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[L0]]{{$}} define <8 x i16> @load_v8i16_with_unfolded_offset(<8 x i16>* %p) { +; CHECK-LABEL: load_v8i16_with_unfolded_offset: +; CHECK: .functype load_v8i16_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <8 x i16>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <8 x i16>* @@ -580,14 +635,15 @@ ret <8 x i16> %v } -; CHECK-LABEL: load_splat_v8i16_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v8i16_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v16x8.load_splat $push[[L0:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[L0]]{{$}} define <8 x i16> @load_splat_v8i16_with_unfolded_offset(i16* %p) { +; CHECK-LABEL: load_splat_v8i16_with_unfolded_offset: +; CHECK: .functype load_splat_v8i16_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v16x8.load_splat 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint i16* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to i16* @@ -597,14 +653,15 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_sext_v8i16_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v8i16_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i16x8.load8x8_s $push[[L0:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[L0]]{{$}} define <8 x i16> @load_sext_v8i16_with_unfolded_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_sext_v8i16_with_unfolded_offset: +; CHECK: .functype load_sext_v8i16_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i16x8.load8x8_s 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <8 x i8>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <8 x i8>* @@ -613,14 +670,15 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_zext_v8i16_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v8i16_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[L0:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[L0]]{{$}} define <8 x i16> @load_zext_v8i16_with_unfolded_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_zext_v8i16_with_unfolded_offset: +; CHECK: .functype load_zext_v8i16_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i16x8.load8x8_u 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <8 x i8>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <8 x i8>* @@ -629,14 +687,15 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_ext_v8i16_with_unfolded_offset: -; NO-SIMD128-NOT: load8x8 -; SIMD128-NEXT: .functype load_ext_v8i16_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[L0:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[L0]]{{$}} define <8 x i8> @load_ext_v8i16_with_unfolded_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_ext_v8i16_with_unfolded_offset: +; CHECK: .functype load_ext_v8i16_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i16x8.load8x8_u 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <8 x i8>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <8 x i8>* @@ -644,27 +703,29 @@ ret <8 x i8> %v } -; CHECK-LABEL: load_v8i16_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v8i16_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_v8i16_with_unfolded_gep_offset(<8 x i16>* %p) { +; CHECK-LABEL: load_v8i16_with_unfolded_gep_offset: +; CHECK: .functype load_v8i16_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <8 x i16>, <8 x i16>* %p, i32 1 %v = load <8 x i16>, <8 x i16>* %s ret <8 x i16> %v } -; CHECK-LABEL: load_splat_v8i16_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v8i16_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 2{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v16x8.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_splat_v8i16_with_unfolded_gep_offset(i16* %p) { +; CHECK-LABEL: load_splat_v8i16_with_unfolded_gep_offset: +; CHECK: .functype load_splat_v8i16_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 2 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v16x8.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr i16, i16* %p, i32 1 %e = load i16, i16* %s %v1 = insertelement <8 x i16> undef, i16 %e, i32 0 @@ -672,66 +733,69 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_sext_v8i16_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v8i16_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i16x8.load8x8_s $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_sext_v8i16_with_unfolded_gep_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_sext_v8i16_with_unfolded_gep_offset: +; CHECK: .functype load_sext_v8i16_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i16x8.load8x8_s 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <8 x i8>, <8 x i8>* %p, i32 1 %v = load <8 x i8>, <8 x i8>* %s %v2 = sext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_zext_v8i16_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v8i16_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_zext_v8i16_with_unfolded_gep_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_zext_v8i16_with_unfolded_gep_offset: +; CHECK: .functype load_zext_v8i16_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i16x8.load8x8_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <8 x i8>, <8 x i8>* %p, i32 1 %v = load <8 x i8>, <8 x i8>* %s %v2 = zext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_ext_v8i16_with_unfolded_gep_offset: -; NO-SIMD128-NOT: load8x8 -; SIMD128-NEXT: .functype load_ext_v8i16_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i8> @load_ext_v8i16_with_unfolded_gep_offset(<8 x i8>* %p) { +; CHECK-LABEL: load_ext_v8i16_with_unfolded_gep_offset: +; CHECK: .functype load_ext_v8i16_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i16x8.load8x8_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <8 x i8>, <8 x i8>* %p, i32 1 %v = load <8 x i8>, <8 x i8>* %s ret <8 x i8> %v } -; CHECK-LABEL: load_v8i16_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v8i16_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_v8i16_from_numeric_address() { +; CHECK-LABEL: load_v8i16_from_numeric_address: +; CHECK: .functype load_v8i16_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <8 x i16>* %v = load <8 x i16>, <8 x i16>* %s ret <8 x i16> %v } -; CHECK-LABEL: load_splat_v8i16_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v8i16_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v16x8.load_splat $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_splat_v8i16_from_numeric_address() { +; CHECK-LABEL: load_splat_v8i16_from_numeric_address: +; CHECK: .functype load_splat_v8i16_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v16x8.load_splat 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to i16* %e = load i16, i16* %s %v1 = insertelement <8 x i16> undef, i16 %e, i32 0 @@ -739,121 +803,127 @@ ret <8 x i16> %v2 } -; CHECK-LABEL: load_sext_v8i16_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v8i16_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i16x8.load8x8_s $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_sext_v8i16_from_numeric_address() { +; CHECK-LABEL: load_sext_v8i16_from_numeric_address: +; CHECK: .functype load_sext_v8i16_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i16x8.load8x8_s 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <8 x i8>* %v = load <8 x i8>, <8 x i8>* %s %v2 = sext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_zext_v8i16_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v8i16_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_zext_v8i16_from_numeric_address() { +; CHECK-LABEL: load_zext_v8i16_from_numeric_address: +; CHECK: .functype load_zext_v8i16_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i16x8.load8x8_u 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <8 x i8>* %v = load <8 x i8>, <8 x i8>* %s %v2 = zext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_ext_v8i16_from_numeric_address: -; NO-SIMD128-NOT: load8x8 -; SIMD128-NEXT: .functype load_ext_v8i16_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i8> @load_ext_v8i16_from_numeric_address() { +; CHECK-LABEL: load_ext_v8i16_from_numeric_address: +; CHECK: .functype load_ext_v8i16_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i16x8.load8x8_u 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <8 x i8>* %v = load <8 x i8>, <8 x i8>* %s ret <8 x i8> %v } -; CHECK-LABEL: load_v8i16_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v8i16_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, gv_v8i16($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_v8i16 = global <8 x i16> define <8 x i16> @load_v8i16_from_global_address() { +; CHECK-LABEL: load_v8i16_from_global_address: +; CHECK: .functype load_v8i16_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load gv_v8i16 +; CHECK-NEXT: # fallthrough-return %v = load <8 x i16>, <8 x i16>* @gv_v8i16 ret <8 x i16> %v } -; CHECK-LABEL: load_splat_v8i16_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v8i16_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v16x8.load_splat $push[[R:[0-9]+]]=, gv_i16($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_i16 = global i16 42 define <8 x i16> @load_splat_v8i16_from_global_address() { +; CHECK-LABEL: load_splat_v8i16_from_global_address: +; CHECK: .functype load_splat_v8i16_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v16x8.load_splat gv_i16 +; CHECK-NEXT: # fallthrough-return %e = load i16, i16* @gv_i16 %v1 = insertelement <8 x i16> undef, i16 %e, i32 0 %v2 = shufflevector <8 x i16> %v1, <8 x i16> undef, <8 x i32> zeroinitializer ret <8 x i16> %v2 } -; CHECK-LABEL: load_sext_v8i16_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v8i16_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i16x8.load8x8_s $push[[R:[0-9]+]]=, gv_v8i8($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_v8i8 = global <8 x i8> define <8 x i16> @load_sext_v8i16_from_global_address() { +; CHECK-LABEL: load_sext_v8i16_from_global_address: +; CHECK: .functype load_sext_v8i16_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i16x8.load8x8_s gv_v8i8 +; CHECK-NEXT: # fallthrough-return %v = load <8 x i8>, <8 x i8>* @gv_v8i8 %v2 = sext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_zext_v8i16_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v8i16_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, gv_v8i8($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i16> @load_zext_v8i16_from_global_address() { +; CHECK-LABEL: load_zext_v8i16_from_global_address: +; CHECK: .functype load_zext_v8i16_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i16x8.load8x8_u gv_v8i8 +; CHECK-NEXT: # fallthrough-return %v = load <8 x i8>, <8 x i8>* @gv_v8i8 %v2 = zext <8 x i8> %v to <8 x i16> ret <8 x i16> %v2 } -; CHECK-LABEL: load_ext_v8i16_from_global_address: -; NO-SIMD128-NOT: load8x8 -; SIMD128-NEXT: .functype load_ext_v8i16_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i16x8.load8x8_u $push[[R:[0-9]+]]=, gv_v8i8($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <8 x i8> @load_ext_v8i16_from_global_address() { +; CHECK-LABEL: load_ext_v8i16_from_global_address: +; CHECK: .functype load_ext_v8i16_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i16x8.load8x8_u gv_v8i8 +; CHECK-NEXT: # fallthrough-return %v = load <8 x i8>, <8 x i8>* @gv_v8i8 ret <8 x i8> %v } -; CHECK-LABEL: store_v8i16: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v8i16 (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 0($1), $0{{$}} define void @store_v8i16(<8 x i16> %v, <8 x i16>* %p) { +; CHECK-LABEL: store_v8i16: +; CHECK: .functype store_v8i16 (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return store <8 x i16> %v , <8 x i16>* %p ret void } -; CHECK-LABEL: store_v8i16_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v8i16_with_folded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v8i16_with_folded_offset(<8 x i16> %v, <8 x i16>* %p) { +; CHECK-LABEL: store_v8i16_with_folded_offset: +; CHECK: .functype store_v8i16_with_folded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <8 x i16>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <8 x i16>* @@ -861,69 +931,85 @@ ret void } -; CHECK-LABEL: store_v8i16_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v8i16_with_folded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v8i16_with_folded_gep_offset(<8 x i16> %v, <8 x i16>* %p) { +; CHECK-LABEL: store_v8i16_with_folded_gep_offset: +; CHECK: .functype store_v8i16_with_folded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i16>, <8 x i16>* %p, i32 1 store <8 x i16> %v , <8 x i16>* %s ret void } -; CHECK-LABEL: store_v8i16_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v8i16_with_unfolded_gep_negative_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v8i16_with_unfolded_gep_negative_offset(<8 x i16> %v, <8 x i16>* %p) { +; CHECK-LABEL: store_v8i16_with_unfolded_gep_negative_offset: +; CHECK: .functype store_v8i16_with_unfolded_gep_negative_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i16>, <8 x i16>* %p, i32 -1 store <8 x i16> %v , <8 x i16>* %s ret void } -; CHECK-LABEL: store_v8i16_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v8i16_with_unfolded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v8i16_with_unfolded_offset(<8 x i16> %v, <8 x i16>* %p) { +; CHECK-LABEL: store_v8i16_with_unfolded_offset: +; CHECK: .functype store_v8i16_with_unfolded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <8 x i16>, <8 x i16>* %p, i32 -1 store <8 x i16> %v , <8 x i16>* %s ret void } -; CHECK-LABEL: store_v8i16_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v8i16_with_unfolded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v8i16_with_unfolded_gep_offset(<8 x i16> %v, <8 x i16>* %p) { +; CHECK-LABEL: store_v8i16_with_unfolded_gep_offset: +; CHECK: .functype store_v8i16_with_unfolded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <8 x i16>, <8 x i16>* %p, i32 1 store <8 x i16> %v , <8 x i16>* %s ret void } -; CHECK-LABEL: store_v8i16_to_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v8i16_to_numeric_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store 32($pop[[L0]]), $0{{$}} define void @store_v8i16_to_numeric_address(<8 x i16> %v) { +; CHECK-LABEL: store_v8i16_to_numeric_address: +; CHECK: .functype store_v8i16_to_numeric_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <8 x i16>* store <8 x i16> %v , <8 x i16>* %s ret void } -; CHECK-LABEL: store_v8i16_to_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v8i16_to_global_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[R:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store gv_v8i16($pop[[R]]), $0{{$}} define void @store_v8i16_to_global_address(<8 x i16> %v) { +; CHECK-LABEL: store_v8i16_to_global_address: +; CHECK: .functype store_v8i16_to_global_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store gv_v8i16 +; CHECK-NEXT: # fallthrough-return store <8 x i16> %v , <8 x i16>* @gv_v8i16 ret void } @@ -931,65 +1017,72 @@ ; ============================================================================== ; 4 x i32 ; ============================================================================== -; CHECK-LABEL: load_v4i32: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4i32 (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_v4i32(<4 x i32>* %p) { +; CHECK-LABEL: load_v4i32: +; CHECK: .functype load_v4i32 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %v = load <4 x i32>, <4 x i32>* %p ret <4 x i32> %v } -; CHECK-LABEL: load_splat_v4i32: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4i32 (i32) -> (v128){{$}} -; SIMD128-NEXT: v32x4.load_splat define <4 x i32> @load_splat_v4i32(i32* %addr) { +; CHECK-LABEL: load_splat_v4i32: +; CHECK: .functype load_splat_v4i32 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v32x4.load_splat 0 +; CHECK-NEXT: # fallthrough-return %e = load i32, i32* %addr, align 4 %v1 = insertelement <4 x i32> undef, i32 %e, i32 0 %v2 = shufflevector <4 x i32> %v1, <4 x i32> undef, <4 x i32> zeroinitializer ret <4 x i32> %v2 } -; CHECK-LABEL: load_sext_v4i32: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v4i32 (i32) -> (v128){{$}} -; SIMD128-NEXT: i32x4.load16x4_s $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_sext_v4i32(<4 x i16>* %p) { +; CHECK-LABEL: load_sext_v4i32: +; CHECK: .functype load_sext_v4i32 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32x4.load16x4_s 0 +; CHECK-NEXT: # fallthrough-return %v = load <4 x i16>, <4 x i16>* %p %v2 = sext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_zext_v4i32: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v4i32 (i32) -> (v128){{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_zext_v4i32(<4 x i16>* %p) { +; CHECK-LABEL: load_zext_v4i32: +; CHECK: .functype load_zext_v4i32 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32x4.load16x4_u 0 +; CHECK-NEXT: # fallthrough-return %v = load <4 x i16>, <4 x i16>* %p %v2 = zext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_ext_v4i32: -; NO-SIMD128-NOT: load16x4 -; SIMD128-NEXT: .functype load_ext_v4i32 (i32) -> (v128){{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i16> @load_ext_v4i32(<4 x i16>* %p) { +; CHECK-LABEL: load_ext_v4i32: +; CHECK: .functype load_ext_v4i32 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32x4.load16x4_u 0 +; CHECK-NEXT: # fallthrough-return %v = load <4 x i16>, <4 x i16>* %p ret <4 x i16> %v } -; CHECK-LABEL: load_v4i32_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4i32_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_v4i32_with_folded_offset(<4 x i32>* %p) { +; CHECK-LABEL: load_v4i32_with_folded_offset: +; CHECK: .functype load_v4i32_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x i32>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <4 x i32>* @@ -997,12 +1090,13 @@ ret <4 x i32> %v } -; CHECK-LABEL: load_splat_v4i32_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4i32_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_splat_v4i32_with_folded_offset(i32* %p) { +; CHECK-LABEL: load_splat_v4i32_with_folded_offset: +; CHECK: .functype load_splat_v4i32_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v32x4.load_splat 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint i32* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to i32* @@ -1012,12 +1106,13 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_sext_v4i32_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v4i32_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32x4.load16x4_s $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_sext_v4i32_with_folded_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_sext_v4i32_with_folded_offset: +; CHECK: .functype load_sext_v4i32_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32x4.load16x4_s 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x i16>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <4 x i16>* @@ -1026,12 +1121,13 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_zext_v4i32_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v4i32_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_zext_v4i32_with_folded_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_zext_v4i32_with_folded_offset: +; CHECK: .functype load_zext_v4i32_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32x4.load16x4_u 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x i16>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <4 x i16>* @@ -1040,12 +1136,13 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_ext_v4i32_with_folded_offset: -; NO-SIMD128-NOT: load16x4 -; SIMD128-NEXT: .functype load_ext_v4i32_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i16> @load_ext_v4i32_with_folded_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_ext_v4i32_with_folded_offset: +; CHECK: .functype load_ext_v4i32_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32x4.load16x4_u 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x i16>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <4 x i16>* @@ -1053,23 +1150,25 @@ ret <4 x i16> %v } -; CHECK-LABEL: load_v4i32_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4i32_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_v4i32_with_folded_gep_offset(<4 x i32>* %p) { +; CHECK-LABEL: load_v4i32_with_folded_gep_offset: +; CHECK: .functype load_v4i32_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i32>, <4 x i32>* %p, i32 1 %v = load <4 x i32>, <4 x i32>* %s ret <4 x i32> %v } -; CHECK-LABEL: load_splat_v4i32_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4i32_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 4($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_splat_v4i32_with_folded_gep_offset(i32* %p) { +; CHECK-LABEL: load_splat_v4i32_with_folded_gep_offset: +; CHECK: .functype load_splat_v4i32_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v32x4.load_splat 4 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds i32, i32* %p, i32 1 %e = load i32, i32* %s %v1 = insertelement <4 x i32> undef, i32 %e, i32 0 @@ -1077,62 +1176,67 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_sext_v4i32_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v4i32_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32x4.load16x4_s $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_sext_v4i32_with_folded_gep_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_sext_v4i32_with_folded_gep_offset: +; CHECK: .functype load_sext_v4i32_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32x4.load16x4_s 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i16>, <4 x i16>* %p, i32 1 %v = load <4 x i16>, <4 x i16>* %s %v2 = sext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_zext_v4i32_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v4i32_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_zext_v4i32_with_folded_gep_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_zext_v4i32_with_folded_gep_offset: +; CHECK: .functype load_zext_v4i32_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32x4.load16x4_u 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i16>, <4 x i16>* %p, i32 1 %v = load <4 x i16>, <4 x i16>* %s %v2 = zext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_ext_v4i32_with_folded_gep_offset: -; NO-SIMD128-NOT: load16x4 -; SIMD128-NEXT: .functype load_ext_v4i32_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i16> @load_ext_v4i32_with_folded_gep_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_ext_v4i32_with_folded_gep_offset: +; CHECK: .functype load_ext_v4i32_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32x4.load16x4_u 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i16>, <4 x i16>* %p, i32 1 %v = load <4 x i16>, <4 x i16>* %s ret <4 x i16> %v } -; CHECK-LABEL: load_v4i32_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_v4i32_with_unfolded_gep_negative_offset(<4 x i32>* %p) { +; CHECK-LABEL: load_v4i32_with_unfolded_gep_negative_offset: +; CHECK: .functype load_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i32>, <4 x i32>* %p, i32 -1 %v = load <4 x i32>, <4 x i32>* %s ret <4 x i32> %v } -; CHECK-LABEL: load_splat_v4i32_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -4{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_splat_v4i32_with_unfolded_gep_negative_offset(i32* %p) { +; CHECK-LABEL: load_splat_v4i32_with_unfolded_gep_negative_offset: +; CHECK: .functype load_splat_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -4 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v32x4.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds i32, i32* %p, i32 -1 %e = load i32, i32* %s %v1 = insertelement <4 x i32> undef, i32 %e, i32 0 @@ -1140,55 +1244,59 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_sext_v4i32_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i32x4.load16x4_s $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_sext_v4i32_with_unfolded_gep_negative_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_sext_v4i32_with_unfolded_gep_negative_offset: +; CHECK: .functype load_sext_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i32x4.load16x4_s 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i16>, <4 x i16>* %p, i32 -1 %v = load <4 x i16>, <4 x i16>* %s %v2 = sext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_zext_v4i32_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_zext_v4i32_with_unfolded_gep_negative_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_zext_v4i32_with_unfolded_gep_negative_offset: +; CHECK: .functype load_zext_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i32x4.load16x4_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i16>, <4 x i16>* %p, i32 -1 %v = load <4 x i16>, <4 x i16>* %s %v2 = zext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_ext_v4i32_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: load16x4 -; SIMD128-NEXT: .functype load_ext_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i16> @load_ext_v4i32_with_unfolded_gep_negative_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_ext_v4i32_with_unfolded_gep_negative_offset: +; CHECK: .functype load_ext_v4i32_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i32x4.load16x4_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i16>, <4 x i16>* %p, i32 -1 %v = load <4 x i16>, <4 x i16>* %s ret <4 x i16> %v } -; CHECK-LABEL: load_v4i32_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4i32_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_v4i32_with_unfolded_offset(<4 x i32>* %p) { +; CHECK-LABEL: load_v4i32_with_unfolded_offset: +; CHECK: .functype load_v4i32_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x i32>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <4 x i32>* @@ -1196,14 +1304,15 @@ ret <4 x i32> %v } -; CHECK-LABEL: load_splat_v4i32_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4i32_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_splat_v4i32_with_unfolded_offset(i32* %p) { +; CHECK-LABEL: load_splat_v4i32_with_unfolded_offset: +; CHECK: .functype load_splat_v4i32_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v32x4.load_splat 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint i32* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to i32* @@ -1213,14 +1322,15 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_sext_v4i32_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v4i32_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i32x4.load16x4_s $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_sext_v4i32_with_unfolded_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_sext_v4i32_with_unfolded_offset: +; CHECK: .functype load_sext_v4i32_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i32x4.load16x4_s 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x i16>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <4 x i16>* @@ -1229,14 +1339,15 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_zext_v4i32_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v4i32_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_zext_v4i32_with_unfolded_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_zext_v4i32_with_unfolded_offset: +; CHECK: .functype load_zext_v4i32_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i32x4.load16x4_u 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x i16>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <4 x i16>* @@ -1245,14 +1356,15 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_ext_v4i32_with_unfolded_offset: -; NO-SIMD128-NOT: load16x4 -; SIMD128-NEXT: .functype load_ext_v4i32_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i16> @load_ext_v4i32_with_unfolded_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_ext_v4i32_with_unfolded_offset: +; CHECK: .functype load_ext_v4i32_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i32x4.load16x4_u 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x i16>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <4 x i16>* @@ -1260,27 +1372,29 @@ ret <4 x i16> %v } -; CHECK-LABEL: load_v4i32_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4i32_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_v4i32_with_unfolded_gep_offset(<4 x i32>* %p) { +; CHECK-LABEL: load_v4i32_with_unfolded_gep_offset: +; CHECK: .functype load_v4i32_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <4 x i32>, <4 x i32>* %p, i32 1 %v = load <4 x i32>, <4 x i32>* %s ret <4 x i32> %v } -; CHECK-LABEL: load_splat_v4i32_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4i32_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 4{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_splat_v4i32_with_unfolded_gep_offset(i32* %p) { +; CHECK-LABEL: load_splat_v4i32_with_unfolded_gep_offset: +; CHECK: .functype load_splat_v4i32_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 4 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v32x4.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr i32, i32* %p, i32 1 %e = load i32, i32* %s %v1 = insertelement <4 x i32> undef, i32 %e, i32 0 @@ -1288,66 +1402,69 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_sext_v4i32_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v4i32_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i32x4.load16x4_s $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_sext_v4i32_with_unfolded_gep_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_sext_v4i32_with_unfolded_gep_offset: +; CHECK: .functype load_sext_v4i32_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i32x4.load16x4_s 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <4 x i16>, <4 x i16>* %p, i32 1 %v = load <4 x i16>, <4 x i16>* %s %v2 = sext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_zext_v4i32_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v4i32_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_zext_v4i32_with_unfolded_gep_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_zext_v4i32_with_unfolded_gep_offset: +; CHECK: .functype load_zext_v4i32_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i32x4.load16x4_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <4 x i16>, <4 x i16>* %p, i32 1 %v = load <4 x i16>, <4 x i16>* %s %v2 = zext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_ext_v4i32_with_unfolded_gep_offset: -; NO-SIMD128-NOT: load16x4 -; SIMD128-NEXT: .functype load_ext_v4i32_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i16> @load_ext_v4i32_with_unfolded_gep_offset(<4 x i16>* %p) { +; CHECK-LABEL: load_ext_v4i32_with_unfolded_gep_offset: +; CHECK: .functype load_ext_v4i32_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i32x4.load16x4_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <4 x i16>, <4 x i16>* %p, i32 1 %v = load <4 x i16>, <4 x i16>* %s ret <4 x i16> %v } -; CHECK-LABEL: load_v4i32_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4i32_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_v4i32_from_numeric_address() { +; CHECK-LABEL: load_v4i32_from_numeric_address: +; CHECK: .functype load_v4i32_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <4 x i32>* %v = load <4 x i32>, <4 x i32>* %s ret <4 x i32> %v } -; CHECK-LABEL: load_splat_v4i32_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4i32_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_splat_v4i32_from_numeric_address() { +; CHECK-LABEL: load_splat_v4i32_from_numeric_address: +; CHECK: .functype load_splat_v4i32_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v32x4.load_splat 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to i32* %e = load i32, i32* %s %v1 = insertelement <4 x i32> undef, i32 %e, i32 0 @@ -1355,120 +1472,126 @@ ret <4 x i32> %v2 } -; CHECK-LABEL: load_sext_v4i32_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v4i32_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i32x4.load16x4_s $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_sext_v4i32_from_numeric_address() { +; CHECK-LABEL: load_sext_v4i32_from_numeric_address: +; CHECK: .functype load_sext_v4i32_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i32x4.load16x4_s 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <4 x i16>* %v = load <4 x i16>, <4 x i16>* %s %v2 = sext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_zext_v4i32_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v4i32_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_zext_v4i32_from_numeric_address() { +; CHECK-LABEL: load_zext_v4i32_from_numeric_address: +; CHECK: .functype load_zext_v4i32_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i32x4.load16x4_u 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <4 x i16>* %v = load <4 x i16>, <4 x i16>* %s %v2 = zext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_ext_v4i32_from_numeric_address: -; NO-SIMD128-NOT: load16x4 -; SIMD128-NEXT: .functype load_ext_v4i32_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i16> @load_ext_v4i32_from_numeric_address() { +; CHECK-LABEL: load_ext_v4i32_from_numeric_address: +; CHECK: .functype load_ext_v4i32_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i32x4.load16x4_u 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <4 x i16>* %v = load <4 x i16>, <4 x i16>* %s ret <4 x i16> %v } -; CHECK-LABEL: load_v4i32_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4i32_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, gv_v4i32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_v4i32 = global <4 x i32> define <4 x i32> @load_v4i32_from_global_address() { +; CHECK-LABEL: load_v4i32_from_global_address: +; CHECK: .functype load_v4i32_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load gv_v4i32 +; CHECK-NEXT: # fallthrough-return %v = load <4 x i32>, <4 x i32>* @gv_v4i32 ret <4 x i32> %v } -; CHECK-LABEL: load_splat_v4i32_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4i32_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, gv_i32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_i32 = global i32 42 define <4 x i32> @load_splat_v4i32_from_global_address() { +; CHECK-LABEL: load_splat_v4i32_from_global_address: +; CHECK: .functype load_splat_v4i32_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v32x4.load_splat gv_i32 +; CHECK-NEXT: # fallthrough-return %e = load i32, i32* @gv_i32 %v1 = insertelement <4 x i32> undef, i32 %e, i32 0 %v2 = shufflevector <4 x i32> %v1, <4 x i32> undef, <4 x i32> zeroinitializer ret <4 x i32> %v2 } -; CHECK-LABEL: load_sext_v4i32_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v4i32_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i32x4.load16x4_s $push[[R:[0-9]+]]=, gv_v4i16($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_v4i16 = global <4 x i16> define <4 x i32> @load_sext_v4i32_from_global_address() { +; CHECK-LABEL: load_sext_v4i32_from_global_address: +; CHECK: .functype load_sext_v4i32_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i32x4.load16x4_s gv_v4i16 +; CHECK-NEXT: # fallthrough-return %v = load <4 x i16>, <4 x i16>* @gv_v4i16 %v2 = sext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_zext_v4i32_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v4i32_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, gv_v4i16($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i32> @load_zext_v4i32_from_global_address() { +; CHECK-LABEL: load_zext_v4i32_from_global_address: +; CHECK: .functype load_zext_v4i32_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i32x4.load16x4_u gv_v4i16 +; CHECK-NEXT: # fallthrough-return %v = load <4 x i16>, <4 x i16>* @gv_v4i16 %v2 = zext <4 x i16> %v to <4 x i32> ret <4 x i32> %v2 } -; CHECK-LABEL: load_ext_v4i32_from_global_address: -; NO-SIMD128-NOT: load16x4 -; SIMD128-NEXT: .functype load_ext_v4i32_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i32x4.load16x4_u $push[[R:[0-9]+]]=, gv_v4i16($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x i16> @load_ext_v4i32_from_global_address() { +; CHECK-LABEL: load_ext_v4i32_from_global_address: +; CHECK: .functype load_ext_v4i32_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i32x4.load16x4_u gv_v4i16 +; CHECK-NEXT: # fallthrough-return %v = load <4 x i16>, <4 x i16>* @gv_v4i16 ret <4 x i16> %v } -; CHECK-LABEL: store_v4i32: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4i32 (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 0($1), $0{{$}} define void @store_v4i32(<4 x i32> %v, <4 x i32>* %p) { +; CHECK-LABEL: store_v4i32: +; CHECK: .functype store_v4i32 (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return store <4 x i32> %v , <4 x i32>* %p ret void } -; CHECK-LABEL: store_v4i32_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4i32_with_folded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v4i32_with_folded_offset(<4 x i32> %v, <4 x i32>* %p) { +; CHECK-LABEL: store_v4i32_with_folded_offset: +; CHECK: .functype store_v4i32_with_folded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x i32>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <4 x i32>* @@ -1476,69 +1599,85 @@ ret void } -; CHECK-LABEL: store_v4i32_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4i32_with_folded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v4i32_with_folded_gep_offset(<4 x i32> %v, <4 x i32>* %p) { +; CHECK-LABEL: store_v4i32_with_folded_gep_offset: +; CHECK: .functype store_v4i32_with_folded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i32>, <4 x i32>* %p, i32 1 store <4 x i32> %v , <4 x i32>* %s ret void } -; CHECK-LABEL: store_v4i32_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4i32_with_unfolded_gep_negative_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v4i32_with_unfolded_gep_negative_offset(<4 x i32> %v, <4 x i32>* %p) { +; CHECK-LABEL: store_v4i32_with_unfolded_gep_negative_offset: +; CHECK: .functype store_v4i32_with_unfolded_gep_negative_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i32>, <4 x i32>* %p, i32 -1 store <4 x i32> %v , <4 x i32>* %s ret void } -; CHECK-LABEL: store_v4i32_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4i32_with_unfolded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v4i32_with_unfolded_offset(<4 x i32> %v, <4 x i32>* %p) { +; CHECK-LABEL: store_v4i32_with_unfolded_offset: +; CHECK: .functype store_v4i32_with_unfolded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x i32>, <4 x i32>* %p, i32 -1 store <4 x i32> %v , <4 x i32>* %s ret void } -; CHECK-LABEL: store_v4i32_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4i32_with_unfolded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v4i32_with_unfolded_gep_offset(<4 x i32> %v, <4 x i32>* %p) { +; CHECK-LABEL: store_v4i32_with_unfolded_gep_offset: +; CHECK: .functype store_v4i32_with_unfolded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <4 x i32>, <4 x i32>* %p, i32 1 store <4 x i32> %v , <4 x i32>* %s ret void } -; CHECK-LABEL: store_v4i32_to_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4i32_to_numeric_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store 32($pop[[L0]]), $0{{$}} define void @store_v4i32_to_numeric_address(<4 x i32> %v) { +; CHECK-LABEL: store_v4i32_to_numeric_address: +; CHECK: .functype store_v4i32_to_numeric_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <4 x i32>* store <4 x i32> %v , <4 x i32>* %s ret void } -; CHECK-LABEL: store_v4i32_to_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4i32_to_global_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[R:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store gv_v4i32($pop[[R]]), $0{{$}} define void @store_v4i32_to_global_address(<4 x i32> %v) { +; CHECK-LABEL: store_v4i32_to_global_address: +; CHECK: .functype store_v4i32_to_global_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store gv_v4i32 +; CHECK-NEXT: # fallthrough-return store <4 x i32> %v , <4 x i32>* @gv_v4i32 ret void } @@ -1546,66 +1685,72 @@ ; ============================================================================== ; 2 x i64 ; ============================================================================== -; CHECK-LABEL: load_v2i64: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2i64 (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_v2i64(<2 x i64>* %p) { +; CHECK-LABEL: load_v2i64: +; CHECK: .functype load_v2i64 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %v = load <2 x i64>, <2 x i64>* %p ret <2 x i64> %v } -; CHECK-LABEL: load_splat_v2i64: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2i64 (i32) -> (v128){{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_splat_v2i64(i64* %p) { +; CHECK-LABEL: load_splat_v2i64: +; CHECK: .functype load_splat_v2i64 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v64x2.load_splat 0 +; CHECK-NEXT: # fallthrough-return %e = load i64, i64* %p %v1 = insertelement <2 x i64> undef, i64 %e, i32 0 %v2 = shufflevector <2 x i64> %v1, <2 x i64> undef, <2 x i32> zeroinitializer ret <2 x i64> %v2 } -; CHECK-LABEL: load_sext_v2i64: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v2i64 (i32) -> (v128){{$}} -; SIMD128-NEXT: i64x2.load32x2_s $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_sext_v2i64(<2 x i32>* %p) { +; CHECK-LABEL: load_sext_v2i64: +; CHECK: .functype load_sext_v2i64 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i64x2.load32x2_s 0 +; CHECK-NEXT: # fallthrough-return %v = load <2 x i32>, <2 x i32>* %p %v2 = sext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_zext_v2i64: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v2i64 (i32) -> (v128){{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_zext_v2i64(<2 x i32>* %p) { +; CHECK-LABEL: load_zext_v2i64: +; CHECK: .functype load_zext_v2i64 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i64x2.load32x2_u 0 +; CHECK-NEXT: # fallthrough-return %v = load <2 x i32>, <2 x i32>* %p %v2 = zext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_ext_v2i64: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_ext_v2i64 (i32) -> (v128){{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i32> @load_ext_v2i64(<2 x i32>* %p) { +; CHECK-LABEL: load_ext_v2i64: +; CHECK: .functype load_ext_v2i64 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i64x2.load32x2_u 0 +; CHECK-NEXT: # fallthrough-return %v = load <2 x i32>, <2 x i32>* %p ret <2 x i32> %v } -; CHECK-LABEL: load_v2i64_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2i64_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_v2i64_with_folded_offset(<2 x i64>* %p) { +; CHECK-LABEL: load_v2i64_with_folded_offset: +; CHECK: .functype load_v2i64_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x i64>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <2 x i64>* @@ -1613,12 +1758,13 @@ ret <2 x i64> %v } -; CHECK-LABEL: load_splat_v2i64_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2i64_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_splat_v2i64_with_folded_offset(i64* %p) { +; CHECK-LABEL: load_splat_v2i64_with_folded_offset: +; CHECK: .functype load_splat_v2i64_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v64x2.load_splat 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint i64* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to i64* @@ -1628,12 +1774,13 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_sext_v2i64_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v2i64_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i64x2.load32x2_s $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_sext_v2i64_with_folded_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_sext_v2i64_with_folded_offset: +; CHECK: .functype load_sext_v2i64_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i64x2.load32x2_s 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x i32>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <2 x i32>* @@ -1642,12 +1789,13 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_zext_v2i64_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v2i64_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_zext_v2i64_with_folded_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_zext_v2i64_with_folded_offset: +; CHECK: .functype load_zext_v2i64_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i64x2.load32x2_u 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x i32>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <2 x i32>* @@ -1656,12 +1804,13 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_ext_v2i64_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_ext_v2i64_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i32> @load_ext_v2i64_with_folded_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_ext_v2i64_with_folded_offset: +; CHECK: .functype load_ext_v2i64_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i64x2.load32x2_u 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x i32>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <2 x i32>* @@ -1669,23 +1818,25 @@ ret <2 x i32> %v } -; CHECK-LABEL: load_v2i64_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2i64_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_v2i64_with_folded_gep_offset(<2 x i64>* %p) { +; CHECK-LABEL: load_v2i64_with_folded_gep_offset: +; CHECK: .functype load_v2i64_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i64>, <2 x i64>* %p, i32 1 %v = load <2 x i64>, <2 x i64>* %s ret <2 x i64> %v } -; CHECK-LABEL: load_splat_v2i64_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2i64_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_splat_v2i64_with_folded_gep_offset(i64* %p) { +; CHECK-LABEL: load_splat_v2i64_with_folded_gep_offset: +; CHECK: .functype load_splat_v2i64_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v64x2.load_splat 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds i64, i64* %p, i32 1 %e = load i64, i64* %s %v1 = insertelement <2 x i64> undef, i64 %e, i32 0 @@ -1693,62 +1844,67 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_sext_v2i64_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v2i64_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i64x2.load32x2_s $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_sext_v2i64_with_folded_gep_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_sext_v2i64_with_folded_gep_offset: +; CHECK: .functype load_sext_v2i64_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i64x2.load32x2_s 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i32>, <2 x i32>* %p, i32 1 %v = load <2 x i32>, <2 x i32>* %s %v2 = sext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_zext_v2i64_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v2i64_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_zext_v2i64_with_folded_gep_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_zext_v2i64_with_folded_gep_offset: +; CHECK: .functype load_zext_v2i64_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i64x2.load32x2_u 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i32>, <2 x i32>* %p, i32 1 %v = load <2 x i32>, <2 x i32>* %s %v2 = zext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_ext_v2i64_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_ext_v2i64_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i32> @load_ext_v2i64_with_folded_gep_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_ext_v2i64_with_folded_gep_offset: +; CHECK: .functype load_ext_v2i64_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i64x2.load32x2_u 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i32>, <2 x i32>* %p, i32 1 %v = load <2 x i32>, <2 x i32>* %s ret <2 x i32> %v } -; CHECK-LABEL: load_v2i64_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_v2i64_with_unfolded_gep_negative_offset(<2 x i64>* %p) { +; CHECK-LABEL: load_v2i64_with_unfolded_gep_negative_offset: +; CHECK: .functype load_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i64>, <2 x i64>* %p, i32 -1 %v = load <2 x i64>, <2 x i64>* %s ret <2 x i64> %v } -; CHECK-LABEL: load_splat_v2i64_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_splat_v2i64_with_unfolded_gep_negative_offset(i64* %p) { +; CHECK-LABEL: load_splat_v2i64_with_unfolded_gep_negative_offset: +; CHECK: .functype load_splat_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v64x2.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds i64, i64* %p, i32 -1 %e = load i64, i64* %s %v1 = insertelement <2 x i64> undef, i64 %e, i32 0 @@ -1756,55 +1912,59 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_sext_v2i64_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i64x2.load32x2_s $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_sext_v2i64_with_unfolded_gep_negative_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_sext_v2i64_with_unfolded_gep_negative_offset: +; CHECK: .functype load_sext_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i64x2.load32x2_s 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i32>, <2 x i32>* %p, i32 -1 %v = load <2 x i32>, <2 x i32>* %s %v2 = sext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_zext_v2i64_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_zext_v2i64_with_unfolded_gep_negative_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_zext_v2i64_with_unfolded_gep_negative_offset: +; CHECK: .functype load_zext_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i64x2.load32x2_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i32>, <2 x i32>* %p, i32 -1 %v = load <2 x i32>, <2 x i32>* %s %v2 = zext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_ext_v2i64_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_ext_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i32> @load_ext_v2i64_with_unfolded_gep_negative_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_ext_v2i64_with_unfolded_gep_negative_offset: +; CHECK: .functype load_ext_v2i64_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i64x2.load32x2_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i32>, <2 x i32>* %p, i32 -1 %v = load <2 x i32>, <2 x i32>* %s ret <2 x i32> %v } -; CHECK-LABEL: load_v2i64_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2i64_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_v2i64_with_unfolded_offset(<2 x i64>* %p) { +; CHECK-LABEL: load_v2i64_with_unfolded_offset: +; CHECK: .functype load_v2i64_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x i64>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <2 x i64>* @@ -1812,14 +1972,15 @@ ret <2 x i64> %v } -; CHECK-LABEL: load_splat_v2i64_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2i64_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_splat_v2i64_with_unfolded_offset(i64* %p) { +; CHECK-LABEL: load_splat_v2i64_with_unfolded_offset: +; CHECK: .functype load_splat_v2i64_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v64x2.load_splat 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint i64* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to i64* @@ -1829,14 +1990,15 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_sext_v2i64_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v2i64_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i64x2.load32x2_s $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_sext_v2i64_with_unfolded_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_sext_v2i64_with_unfolded_offset: +; CHECK: .functype load_sext_v2i64_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i64x2.load32x2_s 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x i32>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <2 x i32>* @@ -1845,14 +2007,15 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_zext_v2i64_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v2i64_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_zext_v2i64_with_unfolded_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_zext_v2i64_with_unfolded_offset: +; CHECK: .functype load_zext_v2i64_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i64x2.load32x2_u 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x i32>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <2 x i32>* @@ -1861,14 +2024,15 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_ext_v2i64_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_ext_v2i64_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i32> @load_ext_v2i64_with_unfolded_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_ext_v2i64_with_unfolded_offset: +; CHECK: .functype load_ext_v2i64_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i64x2.load32x2_u 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x i32>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <2 x i32>* @@ -1876,27 +2040,29 @@ ret <2 x i32> %v } -; CHECK-LABEL: load_v2i64_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2i64_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_v2i64_with_unfolded_gep_offset(<2 x i64>* %p) { +; CHECK-LABEL: load_v2i64_with_unfolded_gep_offset: +; CHECK: .functype load_v2i64_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <2 x i64>, <2 x i64>* %p, i32 1 %v = load <2 x i64>, <2 x i64>* %s ret <2 x i64> %v } -; CHECK-LABEL: load_splat_v2i64_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2i64_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_splat_v2i64_with_unfolded_gep_offset(i64* %p) { +; CHECK-LABEL: load_splat_v2i64_with_unfolded_gep_offset: +; CHECK: .functype load_splat_v2i64_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v64x2.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr i64, i64* %p, i32 1 %e = load i64, i64* %s %v1 = insertelement <2 x i64> undef, i64 %e, i32 0 @@ -1904,66 +2070,69 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_sext_v2i64_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v2i64_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i64x2.load32x2_s $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_sext_v2i64_with_unfolded_gep_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_sext_v2i64_with_unfolded_gep_offset: +; CHECK: .functype load_sext_v2i64_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i64x2.load32x2_s 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <2 x i32>, <2 x i32>* %p, i32 1 %v = load <2 x i32>, <2 x i32>* %s %v2 = sext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_zext_v2i64_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v2i64_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_zext_v2i64_with_unfolded_gep_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_zext_v2i64_with_unfolded_gep_offset: +; CHECK: .functype load_zext_v2i64_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i64x2.load32x2_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <2 x i32>, <2 x i32>* %p, i32 1 %v = load <2 x i32>, <2 x i32>* %s %v2 = zext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_ext_v2i64_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_ext_v2i64_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i32> @load_ext_v2i64_with_unfolded_gep_offset(<2 x i32>* %p) { +; CHECK-LABEL: load_ext_v2i64_with_unfolded_gep_offset: +; CHECK: .functype load_ext_v2i64_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: i64x2.load32x2_u 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <2 x i32>, <2 x i32>* %p, i32 1 %v = load <2 x i32>, <2 x i32>* %s ret <2 x i32> %v } -; CHECK-LABEL: load_v2i64_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2i64_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_v2i64_from_numeric_address() { +; CHECK-LABEL: load_v2i64_from_numeric_address: +; CHECK: .functype load_v2i64_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <2 x i64>* %v = load <2 x i64>, <2 x i64>* %s ret <2 x i64> %v } -; CHECK-LABEL: load_splat_v2i64_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2i64_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_splat_v2i64_from_numeric_address() { +; CHECK-LABEL: load_splat_v2i64_from_numeric_address: +; CHECK: .functype load_splat_v2i64_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v64x2.load_splat 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to i64* %e = load i64, i64* %s %v1 = insertelement <2 x i64> undef, i64 %e, i32 0 @@ -1971,120 +2140,126 @@ ret <2 x i64> %v2 } -; CHECK-LABEL: load_sext_v2i64_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v2i64_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i64x2.load32x2_s $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_sext_v2i64_from_numeric_address() { +; CHECK-LABEL: load_sext_v2i64_from_numeric_address: +; CHECK: .functype load_sext_v2i64_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i64x2.load32x2_s 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <2 x i32>* %v = load <2 x i32>, <2 x i32>* %s %v2 = sext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_zext_v2i64_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v2i64_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_zext_v2i64_from_numeric_address() { +; CHECK-LABEL: load_zext_v2i64_from_numeric_address: +; CHECK: .functype load_zext_v2i64_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i64x2.load32x2_u 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <2 x i32>* %v = load <2 x i32>, <2 x i32>* %s %v2 = zext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_ext_v2i64_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_ext_v2i64_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i32> @load_ext_v2i64_from_numeric_address() { +; CHECK-LABEL: load_ext_v2i64_from_numeric_address: +; CHECK: .functype load_ext_v2i64_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i64x2.load32x2_u 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <2 x i32>* %v = load <2 x i32>, <2 x i32>* %s ret <2 x i32> %v } -; CHECK-LABEL: load_v2i64_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2i64_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, gv_v2i64($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_v2i64 = global <2 x i64> define <2 x i64> @load_v2i64_from_global_address() { +; CHECK-LABEL: load_v2i64_from_global_address: +; CHECK: .functype load_v2i64_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load gv_v2i64 +; CHECK-NEXT: # fallthrough-return %v = load <2 x i64>, <2 x i64>* @gv_v2i64 ret <2 x i64> %v } -; CHECK-LABEL: load_splat_v2i64_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2i64_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, gv_i64($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_i64 = global i64 42 define <2 x i64> @load_splat_v2i64_from_global_address() { +; CHECK-LABEL: load_splat_v2i64_from_global_address: +; CHECK: .functype load_splat_v2i64_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v64x2.load_splat gv_i64 +; CHECK-NEXT: # fallthrough-return %e = load i64, i64* @gv_i64 %v1 = insertelement <2 x i64> undef, i64 %e, i32 0 %v2 = shufflevector <2 x i64> %v1, <2 x i64> undef, <2 x i32> zeroinitializer ret <2 x i64> %v2 } -; CHECK-LABEL: load_sext_v2i64_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_sext_v2i64_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i64x2.load32x2_s $push[[R:[0-9]+]]=, gv_v2i32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_v2i32 = global <2 x i32> define <2 x i64> @load_sext_v2i64_from_global_address() { +; CHECK-LABEL: load_sext_v2i64_from_global_address: +; CHECK: .functype load_sext_v2i64_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i64x2.load32x2_s gv_v2i32 +; CHECK-NEXT: # fallthrough-return %v = load <2 x i32>, <2 x i32>* @gv_v2i32 %v2 = sext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_zext_v2i64_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_zext_v2i64_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, gv_v2i32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i64> @load_zext_v2i64_from_global_address() { +; CHECK-LABEL: load_zext_v2i64_from_global_address: +; CHECK: .functype load_zext_v2i64_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i64x2.load32x2_u gv_v2i32 +; CHECK-NEXT: # fallthrough-return %v = load <2 x i32>, <2 x i32>* @gv_v2i32 %v2 = zext <2 x i32> %v to <2 x i64> ret <2 x i64> %v2 } -; CHECK-LABEL: load_ext_v2i64_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_ext_v2i64_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: i64x2.load32x2_u $push[[R:[0-9]+]]=, gv_v2i32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x i32> @load_ext_v2i64_from_global_address() { +; CHECK-LABEL: load_ext_v2i64_from_global_address: +; CHECK: .functype load_ext_v2i64_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: i64x2.load32x2_u gv_v2i32 +; CHECK-NEXT: # fallthrough-return %v = load <2 x i32>, <2 x i32>* @gv_v2i32 ret <2 x i32> %v } -; CHECK-LABEL: store_v2i64: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2i64 (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 0($1), $0{{$}} define void @store_v2i64(<2 x i64> %v, <2 x i64>* %p) { +; CHECK-LABEL: store_v2i64: +; CHECK: .functype store_v2i64 (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return store <2 x i64> %v , <2 x i64>* %p ret void } -; CHECK-LABEL: store_v2i64_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2i64_with_folded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v2i64_with_folded_offset(<2 x i64> %v, <2 x i64>* %p) { +; CHECK-LABEL: store_v2i64_with_folded_offset: +; CHECK: .functype store_v2i64_with_folded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x i64>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <2 x i64>* @@ -2092,69 +2267,85 @@ ret void } -; CHECK-LABEL: store_v2i64_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2i64_with_folded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v2i64_with_folded_gep_offset(<2 x i64> %v, <2 x i64>* %p) { +; CHECK-LABEL: store_v2i64_with_folded_gep_offset: +; CHECK: .functype store_v2i64_with_folded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i64>, <2 x i64>* %p, i32 1 store <2 x i64> %v , <2 x i64>* %s ret void } -; CHECK-LABEL: store_v2i64_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2i64_with_unfolded_gep_negative_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v2i64_with_unfolded_gep_negative_offset(<2 x i64> %v, <2 x i64>* %p) { +; CHECK-LABEL: store_v2i64_with_unfolded_gep_negative_offset: +; CHECK: .functype store_v2i64_with_unfolded_gep_negative_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i64>, <2 x i64>* %p, i32 -1 store <2 x i64> %v , <2 x i64>* %s ret void } -; CHECK-LABEL: store_v2i64_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2i64_with_unfolded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v2i64_with_unfolded_offset(<2 x i64> %v, <2 x i64>* %p) { +; CHECK-LABEL: store_v2i64_with_unfolded_offset: +; CHECK: .functype store_v2i64_with_unfolded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x i64>, <2 x i64>* %p, i32 -1 store <2 x i64> %v , <2 x i64>* %s ret void } -; CHECK-LABEL: store_v2i64_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2i64_with_unfolded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v2i64_with_unfolded_gep_offset(<2 x i64> %v, <2 x i64>* %p) { +; CHECK-LABEL: store_v2i64_with_unfolded_gep_offset: +; CHECK: .functype store_v2i64_with_unfolded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <2 x i64>, <2 x i64>* %p, i32 1 store <2 x i64> %v , <2 x i64>* %s ret void } -; CHECK-LABEL: store_v2i64_to_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2i64_to_numeric_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store 32($pop[[L0]]), $0{{$}} define void @store_v2i64_to_numeric_address(<2 x i64> %v) { +; CHECK-LABEL: store_v2i64_to_numeric_address: +; CHECK: .functype store_v2i64_to_numeric_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <2 x i64>* store <2 x i64> %v , <2 x i64>* %s ret void } -; CHECK-LABEL: store_v2i64_to_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2i64_to_global_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[R:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store gv_v2i64($pop[[R]]), $0{{$}} define void @store_v2i64_to_global_address(<2 x i64> %v) { +; CHECK-LABEL: store_v2i64_to_global_address: +; CHECK: .functype store_v2i64_to_global_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store gv_v2i64 +; CHECK-NEXT: # fallthrough-return store <2 x i64> %v , <2 x i64>* @gv_v2i64 ret void } @@ -2162,34 +2353,37 @@ ; ============================================================================== ; 4 x float ; ============================================================================== -; CHECK-LABEL: load_v4f32: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4f32 (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_v4f32(<4 x float>* %p) { +; CHECK-LABEL: load_v4f32: +; CHECK: .functype load_v4f32 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %v = load <4 x float>, <4 x float>* %p ret <4 x float> %v } -; CHECK-LABEL: load_splat_v4f32: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4f32 (i32) -> (v128){{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_splat_v4f32(float* %p) { +; CHECK-LABEL: load_splat_v4f32: +; CHECK: .functype load_splat_v4f32 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v32x4.load_splat 0 +; CHECK-NEXT: # fallthrough-return %e = load float, float* %p %v1 = insertelement <4 x float> undef, float %e, i32 0 %v2 = shufflevector <4 x float> %v1, <4 x float> undef, <4 x i32> zeroinitializer ret <4 x float> %v2 } -; CHECK-LABEL: load_v4f32_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4f32_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_v4f32_with_folded_offset(<4 x float>* %p) { +; CHECK-LABEL: load_v4f32_with_folded_offset: +; CHECK: .functype load_v4f32_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x float>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <4 x float>* @@ -2197,12 +2391,13 @@ ret <4 x float> %v } -; CHECK-LABEL: load_splat_v4f32_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4f32_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_splat_v4f32_with_folded_offset(float* %p) { +; CHECK-LABEL: load_splat_v4f32_with_folded_offset: +; CHECK: .functype load_splat_v4f32_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v32x4.load_splat 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint float* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to float* @@ -2212,23 +2407,25 @@ ret <4 x float> %v2 } -; CHECK-LABEL: load_v4f32_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4f32_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_v4f32_with_folded_gep_offset(<4 x float>* %p) { +; CHECK-LABEL: load_v4f32_with_folded_gep_offset: +; CHECK: .functype load_v4f32_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x float>, <4 x float>* %p, i32 1 %v = load <4 x float>, <4 x float>* %s ret <4 x float> %v } -; CHECK-LABEL: load_splat_v4f32_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4f32_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 4($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_splat_v4f32_with_folded_gep_offset(float* %p) { +; CHECK-LABEL: load_splat_v4f32_with_folded_gep_offset: +; CHECK: .functype load_splat_v4f32_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v32x4.load_splat 4 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds float, float* %p, i32 1 %e = load float, float* %s %v1 = insertelement <4 x float> undef, float %e, i32 0 @@ -2236,27 +2433,29 @@ ret <4 x float> %v2 } -; CHECK-LABEL: load_v4f32_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4f32_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_v4f32_with_unfolded_gep_negative_offset(<4 x float>* %p) { +; CHECK-LABEL: load_v4f32_with_unfolded_gep_negative_offset: +; CHECK: .functype load_v4f32_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x float>, <4 x float>* %p, i32 -1 %v = load <4 x float>, <4 x float>* %s ret <4 x float> %v } -; CHECK-LABEL: load_splat_v4f32_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4f32_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -4{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_splat_v4f32_with_unfolded_gep_negative_offset(float* %p) { +; CHECK-LABEL: load_splat_v4f32_with_unfolded_gep_negative_offset: +; CHECK: .functype load_splat_v4f32_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -4 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v32x4.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds float, float* %p, i32 -1 %e = load float, float* %s %v1 = insertelement <4 x float> undef, float %e, i32 0 @@ -2264,14 +2463,15 @@ ret <4 x float> %v2 } -; CHECK-LABEL: load_v4f32_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4f32_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_v4f32_with_unfolded_offset(<4 x float>* %p) { +; CHECK-LABEL: load_v4f32_with_unfolded_offset: +; CHECK: .functype load_v4f32_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x float>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <4 x float>* @@ -2279,14 +2479,15 @@ ret <4 x float> %v } -; CHECK-LABEL: load_splat_v4f32_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4f32_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_splat_v4f32_with_unfolded_offset(float* %p) { +; CHECK-LABEL: load_splat_v4f32_with_unfolded_offset: +; CHECK: .functype load_splat_v4f32_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v32x4.load_splat 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint float* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to float* @@ -2296,27 +2497,29 @@ ret <4 x float> %v2 } -; CHECK-LABEL: load_v4f32_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4f32_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_v4f32_with_unfolded_gep_offset(<4 x float>* %p) { +; CHECK-LABEL: load_v4f32_with_unfolded_gep_offset: +; CHECK: .functype load_v4f32_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <4 x float>, <4 x float>* %p, i32 1 %v = load <4 x float>, <4 x float>* %s ret <4 x float> %v } -; CHECK-LABEL: load_splat_v4f32_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4f32_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 4{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_splat_v4f32_with_unfolded_gep_offset(float* %p) { +; CHECK-LABEL: load_splat_v4f32_with_unfolded_gep_offset: +; CHECK: .functype load_splat_v4f32_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 4 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v32x4.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr float, float* %p, i32 1 %e = load float, float* %s %v1 = insertelement <4 x float> undef, float %e, i32 0 @@ -2324,25 +2527,25 @@ ret <4 x float> %v2 } -; CHECK-LABEL: load_v4f32_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4f32_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_v4f32_from_numeric_address() { +; CHECK-LABEL: load_v4f32_from_numeric_address: +; CHECK: .functype load_v4f32_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <4 x float>* %v = load <4 x float>, <4 x float>* %s ret <4 x float> %v } -; CHECK-LABEL: load_splat_v4f32_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4f32_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <4 x float> @load_splat_v4f32_from_numeric_address() { +; CHECK-LABEL: load_splat_v4f32_from_numeric_address: +; CHECK: .functype load_splat_v4f32_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v32x4.load_splat 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to float* %e = load float, float* %s %v1 = insertelement <4 x float> undef, float %e, i32 0 @@ -2350,46 +2553,52 @@ ret <4 x float> %v2 } -; CHECK-LABEL: load_v4f32_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v4f32_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, gv_v4f32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_v4f32 = global <4 x float> define <4 x float> @load_v4f32_from_global_address() { +; CHECK-LABEL: load_v4f32_from_global_address: +; CHECK: .functype load_v4f32_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load gv_v4f32 +; CHECK-NEXT: # fallthrough-return %v = load <4 x float>, <4 x float>* @gv_v4f32 ret <4 x float> %v } -; CHECK-LABEL: load_splat_v4f32_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v4f32_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v32x4.load_splat $push[[R:[0-9]+]]=, gv_f32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_f32 = global float 42. define <4 x float> @load_splat_v4f32_from_global_address() { +; CHECK-LABEL: load_splat_v4f32_from_global_address: +; CHECK: .functype load_splat_v4f32_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v32x4.load_splat gv_f32 +; CHECK-NEXT: # fallthrough-return %e = load float, float* @gv_f32 %v1 = insertelement <4 x float> undef, float %e, i32 0 %v2 = shufflevector <4 x float> %v1, <4 x float> undef, <4 x i32> zeroinitializer ret <4 x float> %v2 } -; CHECK-LABEL: store_v4f32: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4f32 (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 0($1), $0{{$}} define void @store_v4f32(<4 x float> %v, <4 x float>* %p) { +; CHECK-LABEL: store_v4f32: +; CHECK: .functype store_v4f32 (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return store <4 x float> %v , <4 x float>* %p ret void } -; CHECK-LABEL: store_v4f32_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4f32_with_folded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v4f32_with_folded_offset(<4 x float> %v, <4 x float>* %p) { +; CHECK-LABEL: store_v4f32_with_folded_offset: +; CHECK: .functype store_v4f32_with_folded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <4 x float>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <4 x float>* @@ -2397,69 +2606,85 @@ ret void } -; CHECK-LABEL: store_v4f32_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4f32_with_folded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v4f32_with_folded_gep_offset(<4 x float> %v, <4 x float>* %p) { +; CHECK-LABEL: store_v4f32_with_folded_gep_offset: +; CHECK: .functype store_v4f32_with_folded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x float>, <4 x float>* %p, i32 1 store <4 x float> %v , <4 x float>* %s ret void } -; CHECK-LABEL: store_v4f32_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4f32_with_unfolded_gep_negative_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v4f32_with_unfolded_gep_negative_offset(<4 x float> %v, <4 x float>* %p) { +; CHECK-LABEL: store_v4f32_with_unfolded_gep_negative_offset: +; CHECK: .functype store_v4f32_with_unfolded_gep_negative_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x float>, <4 x float>* %p, i32 -1 store <4 x float> %v , <4 x float>* %s ret void } -; CHECK-LABEL: store_v4f32_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4f32_with_unfolded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v4f32_with_unfolded_offset(<4 x float> %v, <4 x float>* %p) { +; CHECK-LABEL: store_v4f32_with_unfolded_offset: +; CHECK: .functype store_v4f32_with_unfolded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <4 x float>, <4 x float>* %p, i32 -1 store <4 x float> %v , <4 x float>* %s ret void } -; CHECK-LABEL: store_v4f32_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4f32_with_unfolded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v4f32_with_unfolded_gep_offset(<4 x float> %v, <4 x float>* %p) { +; CHECK-LABEL: store_v4f32_with_unfolded_gep_offset: +; CHECK: .functype store_v4f32_with_unfolded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <4 x float>, <4 x float>* %p, i32 1 store <4 x float> %v , <4 x float>* %s ret void } -; CHECK-LABEL: store_v4f32_to_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4f32_to_numeric_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store 32($pop[[L0]]), $0{{$}} define void @store_v4f32_to_numeric_address(<4 x float> %v) { +; CHECK-LABEL: store_v4f32_to_numeric_address: +; CHECK: .functype store_v4f32_to_numeric_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <4 x float>* store <4 x float> %v , <4 x float>* %s ret void } -; CHECK-LABEL: store_v4f32_to_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v4f32_to_global_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[R:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store gv_v4f32($pop[[R]]), $0{{$}} define void @store_v4f32_to_global_address(<4 x float> %v) { +; CHECK-LABEL: store_v4f32_to_global_address: +; CHECK: .functype store_v4f32_to_global_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store gv_v4f32 +; CHECK-NEXT: # fallthrough-return store <4 x float> %v , <4 x float>* @gv_v4f32 ret void } @@ -2467,34 +2692,37 @@ ; ============================================================================== ; 2 x double ; ============================================================================== -; CHECK-LABEL: load_v2f64: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2f64 (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_v2f64(<2 x double>* %p) { +; CHECK-LABEL: load_v2f64: +; CHECK: .functype load_v2f64 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %v = load <2 x double>, <2 x double>* %p ret <2 x double> %v } -; CHECK-LABEL: load_splat_v2f64: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2f64 (i32) -> (v128){{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 0($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_splat_v2f64(double* %p) { +; CHECK-LABEL: load_splat_v2f64: +; CHECK: .functype load_splat_v2f64 (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v64x2.load_splat 0 +; CHECK-NEXT: # fallthrough-return %e = load double, double* %p %v1 = insertelement <2 x double> undef, double %e, i32 0 %v2 = shufflevector <2 x double> %v1, <2 x double> undef, <2 x i32> zeroinitializer ret <2 x double> %v2 } -; CHECK-LABEL: load_v2f64_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2f64_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_v2f64_with_folded_offset(<2 x double>* %p) { +; CHECK-LABEL: load_v2f64_with_folded_offset: +; CHECK: .functype load_v2f64_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x double>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <2 x double>* @@ -2502,12 +2730,13 @@ ret <2 x double> %v } -; CHECK-LABEL: load_splat_v2f64_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2f64_with_folded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_splat_v2f64_with_folded_offset(double* %p) { +; CHECK-LABEL: load_splat_v2f64_with_folded_offset: +; CHECK: .functype load_splat_v2f64_with_folded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v64x2.load_splat 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint double* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to double* @@ -2517,23 +2746,25 @@ ret <2 x double> %v2 } -; CHECK-LABEL: load_v2f64_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2f64_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 16($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_v2f64_with_folded_gep_offset(<2 x double>* %p) { +; CHECK-LABEL: load_v2f64_with_folded_gep_offset: +; CHECK: .functype load_v2f64_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.load 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x double>, <2 x double>* %p, i32 1 %v = load <2 x double>, <2 x double>* %s ret <2 x double> %v } -; CHECK-LABEL: load_splat_v2f64_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2f64_with_folded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 8($0){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_splat_v2f64_with_folded_gep_offset(double* %p) { +; CHECK-LABEL: load_splat_v2f64_with_folded_gep_offset: +; CHECK: .functype load_splat_v2f64_with_folded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v64x2.load_splat 8 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds double, double* %p, i32 1 %e = load double, double* %s %v1 = insertelement <2 x double> undef, double %e, i32 0 @@ -2541,27 +2772,29 @@ ret <2 x double> %v2 } -; CHECK-LABEL: load_v2f64_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2f64_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_v2f64_with_unfolded_gep_negative_offset(<2 x double>* %p) { +; CHECK-LABEL: load_v2f64_with_unfolded_gep_negative_offset: +; CHECK: .functype load_v2f64_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x double>, <2 x double>* %p, i32 -1 %v = load <2 x double>, <2 x double>* %s ret <2 x double> %v } -; CHECK-LABEL: load_splat_v2f64_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2f64_with_unfolded_gep_negative_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_splat_v2f64_with_unfolded_gep_negative_offset(double* %p) { +; CHECK-LABEL: load_splat_v2f64_with_unfolded_gep_negative_offset: +; CHECK: .functype load_splat_v2f64_with_unfolded_gep_negative_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const -8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v64x2.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds double, double* %p, i32 -1 %e = load double, double* %s %v1 = insertelement <2 x double> undef, double %e, i32 0 @@ -2569,14 +2802,15 @@ ret <2 x double> %v2 } -; CHECK-LABEL: load_v2f64_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2f64_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_v2f64_with_unfolded_offset(<2 x double>* %p) { +; CHECK-LABEL: load_v2f64_with_unfolded_offset: +; CHECK: .functype load_v2f64_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x double>* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to <2 x double>* @@ -2584,14 +2818,15 @@ ret <2 x double> %v } -; CHECK-LABEL: load_splat_v2f64_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2f64_with_unfolded_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_splat_v2f64_with_unfolded_offset(double* %p) { +; CHECK-LABEL: load_splat_v2f64_with_unfolded_offset: +; CHECK: .functype load_splat_v2f64_with_unfolded_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v64x2.load_splat 0 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint double* %p to i32 %r = add nsw i32 %q, 16 %s = inttoptr i32 %r to double* @@ -2601,27 +2836,29 @@ ret <2 x double> %v2 } -; CHECK-LABEL: load_v2f64_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2f64_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_v2f64_with_unfolded_gep_offset(<2 x double>* %p) { +; CHECK-LABEL: load_v2f64_with_unfolded_gep_offset: +; CHECK: .functype load_v2f64_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v128.load 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <2 x double>, <2 x double>* %p, i32 1 %v = load <2 x double>, <2 x double>* %s ret <2 x double> %v } -; CHECK-LABEL: load_splat_v2f64_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2f64_with_unfolded_gep_offset (i32) -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 8{{$}} -; SIMD128-NEXT: i32.add $push[[L1:[0-9]+]]=, $0, $pop[[L0]]{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 0($pop[[L1]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_splat_v2f64_with_unfolded_gep_offset(double* %p) { +; CHECK-LABEL: load_splat_v2f64_with_unfolded_gep_offset: +; CHECK: .functype load_splat_v2f64_with_unfolded_gep_offset (i32) -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: i32.const 8 +; CHECK-NEXT: i32.add +; CHECK-NEXT: v64x2.load_splat 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr double, double* %p, i32 1 %e = load double, double* %s %v1 = insertelement <2 x double> undef, double %e, i32 0 @@ -2629,25 +2866,25 @@ ret <2 x double> %v2 } -; CHECK-LABEL: load_v2f64_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2f64_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_v2f64_from_numeric_address() { +; CHECK-LABEL: load_v2f64_from_numeric_address: +; CHECK: .functype load_v2f64_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <2 x double>* %v = load <2 x double>, <2 x double>* %s ret <2 x double> %v } -; CHECK-LABEL: load_splat_v2f64_from_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2f64_from_numeric_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, 32($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} define <2 x double> @load_splat_v2f64_from_numeric_address() { +; CHECK-LABEL: load_splat_v2f64_from_numeric_address: +; CHECK: .functype load_splat_v2f64_from_numeric_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v64x2.load_splat 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to double* %e = load double, double* %s %v1 = insertelement <2 x double> undef, double %e, i32 0 @@ -2655,46 +2892,52 @@ ret <2 x double> %v2 } -; CHECK-LABEL: load_v2f64_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_v2f64_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.load $push[[R:[0-9]+]]=, gv_v2f64($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_v2f64 = global <2 x double> define <2 x double> @load_v2f64_from_global_address() { +; CHECK-LABEL: load_v2f64_from_global_address: +; CHECK: .functype load_v2f64_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v128.load gv_v2f64 +; CHECK-NEXT: # fallthrough-return %v = load <2 x double>, <2 x double>* @gv_v2f64 ret <2 x double> %v } -; CHECK-LABEL: load_splat_v2f64_from_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype load_splat_v2f64_from_global_address () -> (v128){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v64x2.load_splat $push[[R:[0-9]+]]=, gv_f64($pop[[L0]]){{$}} -; SIMD128-NEXT: return $pop[[R]]{{$}} @gv_f64 = global double 42. define <2 x double> @load_splat_v2f64_from_global_address() { +; CHECK-LABEL: load_splat_v2f64_from_global_address: +; CHECK: .functype load_splat_v2f64_from_global_address () -> (v128) +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: v64x2.load_splat gv_f64 +; CHECK-NEXT: # fallthrough-return %e = load double, double* @gv_f64 %v1 = insertelement <2 x double> undef, double %e, i32 0 %v2 = shufflevector <2 x double> %v1, <2 x double> undef, <2 x i32> zeroinitializer ret <2 x double> %v2 } -; CHECK-LABEL: store_v2f64: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2f64 (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 0($1), $0{{$}} define void @store_v2f64(<2 x double> %v, <2 x double>* %p) { +; CHECK-LABEL: store_v2f64: +; CHECK: .functype store_v2f64 (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return store <2 x double> %v , <2 x double>* %p ret void } -; CHECK-LABEL: store_v2f64_with_folded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2f64_with_folded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v2f64_with_folded_offset(<2 x double> %v, <2 x double>* %p) { +; CHECK-LABEL: store_v2f64_with_folded_offset: +; CHECK: .functype store_v2f64_with_folded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %q = ptrtoint <2 x double>* %p to i32 %r = add nuw i32 %q, 16 %s = inttoptr i32 %r to <2 x double>* @@ -2702,69 +2945,85 @@ ret void } -; CHECK-LABEL: store_v2f64_with_folded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2f64_with_folded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: v128.store 16($1), $0{{$}} define void @store_v2f64_with_folded_gep_offset(<2 x double> %v, <2 x double>* %p) { +; CHECK-LABEL: store_v2f64_with_folded_gep_offset: +; CHECK: .functype store_v2f64_with_folded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 16 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x double>, <2 x double>* %p, i32 1 store <2 x double> %v , <2 x double>* %s ret void } -; CHECK-LABEL: store_v2f64_with_unfolded_gep_negative_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2f64_with_unfolded_gep_negative_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v2f64_with_unfolded_gep_negative_offset(<2 x double> %v, <2 x double>* %p) { +; CHECK-LABEL: store_v2f64_with_unfolded_gep_negative_offset: +; CHECK: .functype store_v2f64_with_unfolded_gep_negative_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x double>, <2 x double>* %p, i32 -1 store <2 x double> %v , <2 x double>* %s ret void } -; CHECK-LABEL: store_v2f64_with_unfolded_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2f64_with_unfolded_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, -16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v2f64_with_unfolded_offset(<2 x double> %v, <2 x double>* %p) { +; CHECK-LABEL: store_v2f64_with_unfolded_offset: +; CHECK: .functype store_v2f64_with_unfolded_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const -16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr inbounds <2 x double>, <2 x double>* %p, i32 -1 store <2 x double> %v , <2 x double>* %s ret void } -; CHECK-LABEL: store_v2f64_with_unfolded_gep_offset: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2f64_with_unfolded_gep_offset (v128, i32) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 16{{$}} -; SIMD128-NEXT: i32.add $push[[R:[0-9]+]]=, $1, $pop[[L0]]{{$}} -; SIMD128-NEXT: v128.store 0($pop[[R]]), $0{{$}} define void @store_v2f64_with_unfolded_gep_offset(<2 x double> %v, <2 x double>* %p) { +; CHECK-LABEL: store_v2f64_with_unfolded_gep_offset: +; CHECK: .functype store_v2f64_with_unfolded_gep_offset (v128, i32) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: local.get 1 +; CHECK-NEXT: i32.const 16 +; CHECK-NEXT: i32.add +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 0 +; CHECK-NEXT: # fallthrough-return %s = getelementptr <2 x double>, <2 x double>* %p, i32 1 store <2 x double> %v , <2 x double>* %s ret void } -; CHECK-LABEL: store_v2f64_to_numeric_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2f64_to_numeric_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[L0:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store 32($pop[[L0]]), $0{{$}} define void @store_v2f64_to_numeric_address(<2 x double> %v) { +; CHECK-LABEL: store_v2f64_to_numeric_address: +; CHECK: .functype store_v2f64_to_numeric_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store 32 +; CHECK-NEXT: # fallthrough-return %s = inttoptr i32 32 to <2 x double>* store <2 x double> %v , <2 x double>* %s ret void } -; CHECK-LABEL: store_v2f64_to_global_address: -; NO-SIMD128-NOT: v128 -; SIMD128-NEXT: .functype store_v2f64_to_global_address (v128) -> (){{$}} -; SIMD128-NEXT: i32.const $push[[R:[0-9]+]]=, 0{{$}} -; SIMD128-NEXT: v128.store gv_v2f64($pop[[R]]), $0{{$}} define void @store_v2f64_to_global_address(<2 x double> %v) { +; CHECK-LABEL: store_v2f64_to_global_address: +; CHECK: .functype store_v2f64_to_global_address (v128) -> () +; CHECK-NEXT: # %bb.0: +; CHECK-NEXT: i32.const 0 +; CHECK-NEXT: local.get 0 +; CHECK-NEXT: v128.store gv_v2f64 +; CHECK-NEXT: # fallthrough-return store <2 x double> %v , <2 x double>* @gv_v2f64 ret void }