1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2 ; REQUIRES: asserts 3 ; RUN: llc < %s -mattr=+sse3,+sse4.1 -mcpu=penryn -stats 2>&1 | grep "9 machinelicm" 4 ; RUN: llc < %s -mattr=+sse3,+sse4.1 -mcpu=penryn | FileCheck %s 5 ; rdar://6627786 6 ; rdar://7792037 7 8 target triple = "x86_64-apple-darwin10.0" 9 %struct.Key = type { i64 } 10 %struct.__Rec = type opaque 11 %struct.__vv = type { } 12 13 define %struct.__vv* @t(%struct.Key* %desc, i64 %p) nounwind ssp { 14 ; CHECK-LABEL: t: 15 ; CHECK: ## %bb.0: ## %entry 16 ; CHECK-NEXT: pushq %r14 17 ; CHECK-NEXT: pushq %rbx 18 ; CHECK-NEXT: pushq %rax 19 ; CHECK-NEXT: movq %rsi, %r14 20 ; CHECK-NEXT: movq %rdi, %rbx 21 ; CHECK-NEXT: orq $2097152, %r14 ## imm = 0x200000 22 ; CHECK-NEXT: andl $15728640, %r14d ## imm = 0xF00000 23 ; CHECK-NEXT: jmp LBB0_1 24 ; CHECK-NEXT: .p2align 4, 0x90 25 ; CHECK-NEXT: LBB0_3: ## %bb.i 26 ; CHECK-NEXT: ## in Loop: Header=BB0_1 Depth=1 27 ; CHECK-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero 28 ; CHECK-NEXT: pinsrd $1, 4, %xmm0 29 ; CHECK-NEXT: pinsrd $2, 8, %xmm0 30 ; CHECK-NEXT: movdqa %xmm0, %xmm1 31 ; CHECK-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0],mem[1],xmm1[2],mem[3],xmm1[4],mem[5],xmm1[6],mem[7] 32 ; CHECK-NEXT: psrld $16, %xmm0 33 ; CHECK-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2],mem[3],xmm0[4],mem[5],xmm0[6],mem[7] 34 ; CHECK-NEXT: addps {{.*}}(%rip), %xmm0 35 ; CHECK-NEXT: addps %xmm1, %xmm0 36 ; CHECK-NEXT: movaps %xmm0, 0 37 ; CHECK-NEXT: LBB0_1: ## %bb4 38 ; CHECK-NEXT: ## =>This Inner Loop Header: Depth=1 39 ; CHECK-NEXT: xorl %eax, %eax 40 ; CHECK-NEXT: callq _xxGetOffsetForCode 41 ; CHECK-NEXT: xorl %esi, %esi 42 ; CHECK-NEXT: xorl %eax, %eax 43 ; CHECK-NEXT: movq %rbx, %rdi 44 ; CHECK-NEXT: callq _xxCalculateMidType 45 ; CHECK-NEXT: cmpl $1, %eax 46 ; CHECK-NEXT: jne LBB0_1 47 ; CHECK-NEXT: ## %bb.2: ## %bb26 48 ; CHECK-NEXT: ## in Loop: Header=BB0_1 Depth=1 49 ; CHECK-NEXT: cmpq $1048576, %r14 ## imm = 0x100000 50 ; CHECK-NEXT: jne LBB0_1 51 ; CHECK-NEXT: jmp LBB0_3 52 entry: 53 br label %bb4 54 55 bb4: ; preds = %bb.i, %bb26, %bb4, %entry 56 57 %0 = call i32 (...) @xxGetOffsetForCode(i32 undef) nounwind ; <i32> [#uses=0] 58 %ins = or i64 %p, 2097152 ; <i64> [#uses=1] 59 %1 = call i32 (...) @xxCalculateMidType(%struct.Key* %desc, i32 0) nounwind ; <i32> [#uses=1] 60 %cond = icmp eq i32 %1, 1 ; <i1> [#uses=1] 61 br i1 %cond, label %bb26, label %bb4 62 63 bb26: ; preds = %bb4 64 %2 = and i64 %ins, 15728640 ; <i64> [#uses=1] 65 %cond.i = icmp eq i64 %2, 1048576 ; <i1> [#uses=1] 66 br i1 %cond.i, label %bb.i, label %bb4 67 68 bb.i: ; preds = %bb26 69 %3 = load i32, i32* null, align 4 ; <i32> [#uses=1] 70 %4 = uitofp i32 %3 to float ; <float> [#uses=1] 71 %.sum13.i = add i64 0, 4 ; <i64> [#uses=1] 72 %5 = getelementptr i8, i8* null, i64 %.sum13.i ; <i8*> [#uses=1] 73 %6 = bitcast i8* %5 to i32* ; <i32*> [#uses=1] 74 %7 = load i32, i32* %6, align 4 ; <i32> [#uses=1] 75 %8 = uitofp i32 %7 to float ; <float> [#uses=1] 76 %.sum.i = add i64 0, 8 ; <i64> [#uses=1] 77 %9 = getelementptr i8, i8* null, i64 %.sum.i ; <i8*> [#uses=1] 78 %10 = bitcast i8* %9 to i32* ; <i32*> [#uses=1] 79 %11 = load i32, i32* %10, align 4 ; <i32> [#uses=1] 80 %12 = uitofp i32 %11 to float ; <float> [#uses=1] 81 %13 = insertelement <4 x float> undef, float %4, i32 0 ; <<4 x float>> [#uses=1] 82 %14 = insertelement <4 x float> %13, float %8, i32 1 ; <<4 x float>> [#uses=1] 83 %15 = insertelement <4 x float> %14, float %12, i32 2 ; <<4 x float>> [#uses=1] 84 store <4 x float> %15, <4 x float>* null, align 16 85 br label %bb4 86 } 87 88 declare i32 @xxGetOffsetForCode(...) 89 90 declare i32 @xxCalculateMidType(...) 91
