1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+mmx,+sse2 | FileCheck %s 3 4%SA = type <{ %union.anon, i32, [4 x i8], i8*, i8*, i8*, i32, [4 x i8] }> 5%union.anon = type { <1 x i64> } 6 7; Check that extra movd (copy) instructions aren't generated. 8 9define i32 @test(%SA* %pSA, i16* %A, i32 %B, i32 %C, i32 %D, i8* %E) { 10; CHECK-LABEL: test: 11; CHECK: # %bb.0: # %entry 12; CHECK-NEXT: pshufw $238, (%rdi), %mm0 # mm0 = mem[2,3,2,3] 13; CHECK-NEXT: movd %mm0, %eax 14; CHECK-NEXT: testl %eax, %eax 15; CHECK-NEXT: je .LBB0_1 16; CHECK-NEXT: # %bb.2: # %if.B 17; CHECK-NEXT: pshufw $238, %mm0, %mm0 # mm0 = mm0[2,3,2,3] 18; CHECK-NEXT: movq %mm0, %rax 19; CHECK-NEXT: jmp .LBB0_3 20; CHECK-NEXT: .LBB0_1: # %if.A 21; CHECK-NEXT: movd %edx, %mm1 22; CHECK-NEXT: psllq %mm1, %mm0 23; CHECK-NEXT: movq %mm0, %rax 24; CHECK-NEXT: testq %rax, %rax 25; CHECK-NEXT: jne .LBB0_4 26; CHECK-NEXT: .LBB0_3: # %if.C 27; CHECK-NEXT: testl %eax, %eax 28; CHECK-NEXT: je .LBB0_1 29; CHECK-NEXT: .LBB0_4: # %merge 30; CHECK-NEXT: pshufw $238, %mm0, %mm0 # mm0 = mm0[2,3,2,3] 31; CHECK-NEXT: movd %mm0, %eax 32; CHECK-NEXT: retq 33entry: 34 %shl = shl i32 1, %B 35 %shl1 = shl i32 %C, %B 36 %shl2 = shl i32 1, %D 37 %v = getelementptr inbounds %SA, %SA* %pSA, i64 0, i32 0, i32 0 38 %v0 = load <1 x i64>, <1 x i64>* %v, align 8 39 %SA0 = getelementptr inbounds %SA, %SA* %pSA, i64 0, i32 1 40 %v1 = load i32, i32* %SA0, align 4 41 %SA1 = getelementptr inbounds %SA, %SA* %pSA, i64 0, i32 3 42 %v2 = load i8*, i8** %SA1, align 8 43 %SA2 = getelementptr inbounds %SA, %SA* %pSA, i64 0, i32 4 44 %v3 = load i8*, i8** %SA2, align 8 45 %v4 = bitcast <1 x i64> %v0 to <4 x i16> 46 %v5 = bitcast <4 x i16> %v4 to x86_mmx 47 %v6 = tail call x86_mmx @llvm.x86.sse.pshuf.w(x86_mmx %v5, i8 -18) 48 %v7 = bitcast x86_mmx %v6 to <4 x i16> 49 %v8 = bitcast <4 x i16> %v7 to <1 x i64> 50 %v9 = extractelement <1 x i64> %v8, i32 0 51 %v10 = bitcast i64 %v9 to <2 x i32> 52 %v11 = extractelement <2 x i32> %v10, i32 0 53 %cmp = icmp eq i32 %v11, 0 54 br i1 %cmp, label %if.A, label %if.B 55 56if.A: 57 %pa = phi <1 x i64> [ %v8, %entry ], [ %vx, %if.C ] 58 %v17 = extractelement <1 x i64> %pa, i32 0 59 %v18 = bitcast i64 %v17 to x86_mmx 60 %v19 = tail call x86_mmx @llvm.x86.mmx.pslli.q(x86_mmx %v18, i32 %B) #2 61 %v20 = bitcast x86_mmx %v19 to i64 62 %v21 = insertelement <1 x i64> undef, i64 %v20, i32 0 63 %cmp3 = icmp eq i64 %v20, 0 64 br i1 %cmp3, label %if.C, label %merge 65 66if.B: 67 %v34 = bitcast <1 x i64> %v8 to <4 x i16> 68 %v35 = bitcast <4 x i16> %v34 to x86_mmx 69 %v36 = tail call x86_mmx @llvm.x86.sse.pshuf.w(x86_mmx %v35, i8 -18) 70 %v37 = bitcast x86_mmx %v36 to <4 x i16> 71 %v38 = bitcast <4 x i16> %v37 to <1 x i64> 72 br label %if.C 73 74if.C: 75 %vx = phi <1 x i64> [ %v21, %if.A ], [ %v38, %if.B ] 76 %cvt = bitcast <1 x i64> %vx to <2 x i32> 77 %ex = extractelement <2 x i32> %cvt, i32 0 78 %cmp2 = icmp eq i32 %ex, 0 79 br i1 %cmp2, label %if.A, label %merge 80 81merge: 82 %vy = phi <1 x i64> [ %v21, %if.A ], [ %vx, %if.C ] 83 %v130 = bitcast <1 x i64> %vy to <4 x i16> 84 %v131 = bitcast <4 x i16> %v130 to x86_mmx 85 %v132 = tail call x86_mmx @llvm.x86.sse.pshuf.w(x86_mmx %v131, i8 -18) 86 %v133 = bitcast x86_mmx %v132 to <4 x i16> 87 %v134 = bitcast <4 x i16> %v133 to <1 x i64> 88 %v135 = extractelement <1 x i64> %v134, i32 0 89 %v136 = bitcast i64 %v135 to <2 x i32> 90 %v137 = extractelement <2 x i32> %v136, i32 0 91 ret i32 %v137 92} 93 94 95declare x86_mmx @llvm.x86.sse.pshuf.w(x86_mmx, i8) 96declare x86_mmx @llvm.x86.mmx.pslli.q(x86_mmx, i32) 97