1; RUN: llc < %s -mcpu=corei7-avx -mattr=+avx | FileCheck %s
2; PR11862
3target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
4target triple = "x86_64-pc-win32"
5
6; This function has live ymm registers across a win64 call.
7; The ymm6-15 registers are still call-clobbered even if xmm6-15 are callee-saved.
8; Verify that callee-saved registers are not being used.
9
10; CHECK: f___vyf
11; CHECK: pushq %rbp
12; CHECK: vmovmsk
13; CHECK: vmovaps %ymm{{.*}}(%r
14; CHECK: vmovaps %ymm{{.*}}(%r
15; CHECK: call
16; Two reloads. It's OK if these get folded.
17; CHECK: vmovaps {{.*\(%r.*}}, %ymm
18; CHECK: vmovaps {{.*\(%r.*}}, %ymm
19; CHECK: blend
20define <8 x float> @f___vyf(<8 x float> %x, <8 x i32> %__mask) nounwind readnone {
21allocas:
22  %bincmp = fcmp oeq <8 x float> %x, zeroinitializer
23  %val_to_boolvec32 = sext <8 x i1> %bincmp to <8 x i32>
24  %"~test" = xor <8 x i32> %val_to_boolvec32, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>
25  %"internal_mask&function_mask25" = and <8 x i32> %"~test", %__mask
26  %floatmask.i46 = bitcast <8 x i32> %"internal_mask&function_mask25" to <8 x float>
27  %v.i47 = call i32 @llvm.x86.avx.movmsk.ps.256(<8 x float> %floatmask.i46) nounwind readnone
28  %any_mm_cmp27 = icmp eq i32 %v.i47, 0
29  br i1 %any_mm_cmp27, label %safe_if_after_false, label %safe_if_run_false
30
31safe_if_run_false:                                ; preds = %allocas
32  %binop = fadd <8 x float> %x, <float -1.000000e+00, float -1.000000e+00, float -1.000000e+00, float -1.000000e+00, float -1.000000e+00, float -1.000000e+00, float -1.000000e+00, float -1.000000e+00>
33  %calltmp = call <8 x float> @f___vyf(<8 x float> %binop, <8 x i32> %"internal_mask&function_mask25")
34  %binop33 = fadd <8 x float> %calltmp, %x
35  %mask_as_float.i48 = bitcast <8 x i32> %"~test" to <8 x float>
36  %blend.i52 = call <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float> %x, <8 x float> %binop33, <8 x float> %mask_as_float.i48) nounwind
37  br label %safe_if_after_false
38
39safe_if_after_false:                              ; preds = %safe_if_run_false, %allocas
40  %0 = phi <8 x float> [ %x, %allocas ], [ %blend.i52, %safe_if_run_false ]
41  ret <8 x float> %0
42}
43
44declare i32 @llvm.x86.avx.movmsk.ps.256(<8 x float>) nounwind readnone
45declare <8 x float> @llvm.x86.avx.maskload.ps.256(i8*, <8 x float>) nounwind readonly
46declare void @llvm.x86.avx.maskstore.ps.256(i8*, <8 x float>, <8 x float>) nounwind
47declare <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float>, <8 x float>, <8 x float>) nounwind readnone
48