1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mcpu=skylake | FileCheck %s --check-prefix=X86
3 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mcpu=skx | FileCheck %s --check-prefix=X86
4 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skylake | FileCheck %s --check-prefix=X64
5 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skx | FileCheck %s --check-prefix=X64
7 define void @fetch_r16g16_snorm_unorm8(ptr, ptr, i32, i32, ptr) nounwind {
8 ; X86-LABEL: fetch_r16g16_snorm_unorm8:
9 ; X86: # %bb.0: # %entry
10 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
11 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
12 ; X86-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
13 ; X86-NEXT: vpxor %xmm1, %xmm1, %xmm1
14 ; X86-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
15 ; X86-NEXT: vpsrlw $7, %xmm0, %xmm0
16 ; X86-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2],zero,xmm0[u,u,u,u,u,u,u,u,u,u,u,u,u]
17 ; X86-NEXT: vmovd %xmm0, %ecx
18 ; X86-NEXT: orl $-16777216, %ecx # imm = 0xFF000000
19 ; X86-NEXT: movl %ecx, (%eax)
22 ; X64-LABEL: fetch_r16g16_snorm_unorm8:
23 ; X64: # %bb.0: # %entry
24 ; X64-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
25 ; X64-NEXT: vpxor %xmm1, %xmm1, %xmm1
26 ; X64-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
27 ; X64-NEXT: vpsrlw $7, %xmm0, %xmm0
28 ; X64-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2],zero,xmm0[u,u,u,u,u,u,u,u,u,u,u,u,u]
29 ; X64-NEXT: vmovd %xmm0, %eax
30 ; X64-NEXT: orl $-16777216, %eax # imm = 0xFF000000
31 ; X64-NEXT: movl %eax, (%rdi)
34 %5 = load <2 x i16>, ptr %1, align 2
35 %6 = shufflevector <2 x i16> %5, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
36 %7 = icmp sgt <4 x i16> %6, zeroinitializer
37 %8 = select <4 x i1> %7, <4 x i16> %6, <4 x i16> zeroinitializer
38 %9 = lshr <4 x i16> %8, <i16 7, i16 7, i16 7, i16 7>
39 %10 = shufflevector <4 x i16> %9, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
40 %11 = shufflevector <4 x i16> %9, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
41 %12 = bitcast <2 x i16> %10 to <4 x i8>
42 %13 = bitcast <2 x i16> %11 to <4 x i8>
43 %14 = shufflevector <4 x i8> %12, <4 x i8> %13, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
44 %15 = bitcast <4 x i8> %14 to i32
45 %16 = and i32 %15, 65535
46 %17 = or i32 %16, -16777216
47 store i32 %17, ptr %0, align 4