1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512fp16 --fp-contract=fast --enable-unsafe-fp-math | FileCheck %s
4 define dso_local <32 x half> @test1(<32 x half> %lhs.coerce.conj, <32 x half> %rhs.coerce) local_unnamed_addr #0 {
6 ; CHECK: # %bb.0: # %entry
7 ; CHECK-NEXT: vfcmulcph %zmm0, %zmm1, %zmm2
8 ; CHECK-NEXT: vmovaps %zmm2, %zmm0
11 %0 = bitcast <32 x half> %lhs.coerce.conj to <16 x i32>
12 %xor.i.i = xor <16 x i32> %0, <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648>
13 %1 = bitcast <16 x i32> %xor.i.i to <16 x float>
14 %2 = bitcast <32 x half> %rhs.coerce to <16 x float>
15 %3 = tail call fast <16 x float> @llvm.x86.avx512fp16.mask.vfmul.cph.512(<16 x float> %1, <16 x float> %2, <16 x float> zeroinitializer, i16 -1, i32 4) #2
16 %4 = bitcast <16 x float> %3 to <32 x half>
20 ; Function Attrs: nounwind readnone uwtable
21 define dso_local <16 x half> @test2(<16 x half> %lhs.coerce.conj, <16 x half> %rhs.coerce) local_unnamed_addr #0 {
23 ; CHECK: # %bb.0: # %entry
24 ; CHECK-NEXT: vfcmulcph %ymm0, %ymm1, %ymm2
25 ; CHECK-NEXT: vmovaps %ymm2, %ymm0
28 %0 = bitcast <16 x half> %lhs.coerce.conj to <8 x i32>
29 %xor.i.i = xor <8 x i32> %0, <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648>
30 %1 = bitcast <8 x i32> %xor.i.i to <8 x float>
31 %2 = bitcast <16 x half> %rhs.coerce to <8 x float>
32 %3 = tail call fast <8 x float> @llvm.x86.avx512fp16.mask.vfmul.cph.256(<8 x float> %1, <8 x float> %2, <8 x float> zeroinitializer, i8 -1) #2
33 %4 = bitcast <8 x float> %3 to <16 x half>
37 define dso_local <8 x half> @test3(<8 x half> %lhs.coerce.conj, <8 x half> %rhs.coerce) local_unnamed_addr #0 {
39 ; CHECK: # %bb.0: # %entry
40 ; CHECK-NEXT: vfcmulcph %xmm0, %xmm1, %xmm2
41 ; CHECK-NEXT: vmovaps %xmm2, %xmm0
44 %0 = bitcast <8 x half> %lhs.coerce.conj to <4 x i32>
45 %xor.i.i = xor <4 x i32> %0, <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648>
46 %1 = bitcast <4 x i32> %xor.i.i to <4 x float>
47 %2 = bitcast <8 x half> %rhs.coerce to <4 x float>
48 %3 = tail call fast <4 x float> @llvm.x86.avx512fp16.mask.vfmul.cph.128(<4 x float> %1, <4 x float> %2, <4 x float> zeroinitializer, i8 -1) #2
49 %4 = bitcast <4 x float> %3 to <8 x half>
53 define dso_local <8 x half> @test4(<8 x half> %lhs.coerce.conj, <8 x half> %rhs.coerce) local_unnamed_addr #0 {
55 ; CHECK: # %bb.0: # %entry
56 ; CHECK-NEXT: vfmulcph %xmm0, %xmm1, %xmm2
57 ; CHECK-NEXT: vmovaps %xmm2, %xmm0
60 %0 = bitcast <8 x half> %lhs.coerce.conj to <4 x i32>
61 %xor.i.i = xor <4 x i32> %0, <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648>
62 %1 = bitcast <4 x i32> %xor.i.i to <4 x float>
63 %2 = bitcast <8 x half> %rhs.coerce to <4 x float>
64 %3 = tail call fast <4 x float> @llvm.x86.avx512fp16.mask.vfcmul.cph.128(<4 x float> %1, <4 x float> %2, <4 x float> zeroinitializer, i8 -1) #2
65 %4 = bitcast <4 x float> %3 to <8 x half>
69 define dso_local <32 x half> @test5(<32 x half> noundef %a, <32 x half> noundef %b) local_unnamed_addr #0 {
71 ; CHECK: # %bb.0: # %entry
72 ; CHECK-NEXT: vpbroadcastw {{.*#+}} zmm2 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
73 ; CHECK-NEXT: vpxorq %zmm2, %zmm1, %zmm2
74 ; CHECK-NEXT: vfmulcph %zmm2, %zmm0, %zmm1
75 ; CHECK-NEXT: vmovaps %zmm1, %zmm0
78 %fneg = fneg <32 x half> %b
79 %0 = bitcast <32 x half> %a to <16 x float>
80 %1 = bitcast <32 x half> %fneg to <16 x float>
81 %2 = tail call <16 x float> @llvm.x86.avx512fp16.mask.vfmul.cph.512(<16 x float> %0, <16 x float> %1, <16 x float> zeroinitializer, i16 -1, i32 4)
82 %3 = bitcast <16 x float> %2 to <32 x half>
86 declare <16 x float> @llvm.x86.avx512fp16.mask.vfmul.cph.512(<16 x float>, <16 x float>, <16 x float>, i16, i32 immarg)
87 declare <8 x float> @llvm.x86.avx512fp16.mask.vfmul.cph.256(<8 x float>, <8 x float>, <8 x float>, i8)
88 declare <4 x float> @llvm.x86.avx512fp16.mask.vfmul.cph.128(<4 x float>, <4 x float>, <4 x float>, i8)
89 declare <4 x float> @llvm.x86.avx512fp16.mask.vfcmul.cph.128(<4 x float>, <4 x float>, <4 x float>, i8)