1 // RUN: %clang_cc1 %s -O0 -ffreestanding -triple=x86_64-apple-darwin -target-cpu skylake-avx512 -emit-llvm -o - -Wall -Werror |opt -instnamer -S |FileCheck %s
2 // This test checks validity of att\gcc style inline assmebly for avx512 k and Yk constraints.
3 // Also checks mask register allows flexible type (size <= 64 bit)
4
5 #include <x86intrin.h>
6
mask_Yk_i8(char msk,__m512i x,__m512i y)7 __m512i mask_Yk_i8(char msk, __m512i x, __m512i y){
8 // CHECK: <8 x i64> asm "vpaddq\09$3, $2, $0 {$1}", "=x,^Yk,x,x,~{dirflag},~{fpsr},~{flags}"(i8 %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}})
9 __m512i dst;
10 asm ("vpaddq\t%3, %2, %0 %{%1%}"
11 : "=x" (dst) //output
12 : "Yk" (msk), "x" (x), "x" (y)); //inputs
13 return dst;
14 }
15
mask_Yk_i16(short msk,__m512i x,__m512i y)16 __m512i mask_Yk_i16(short msk, __m512i x, __m512i y){
17 // CHECK: <8 x i64> asm "vpaddd\09$3, $2, $0 {$1}", "=x,^Yk,x,x,~{dirflag},~{fpsr},~{flags}"(i16 %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}})
18 __m512i dst;
19 asm ("vpaddd\t%3, %2, %0 %{%1%}"
20 : "=x" (dst) //output
21 : "Yk" (msk), "x" (x), "x" (y)); //inputs
22 return dst;
23 }
24
mask_Yk_i32(int msk,__m512i x,__m512i y)25 __m512i mask_Yk_i32(int msk, __m512i x, __m512i y){
26 // CHECK: <8 x i64> asm "vpaddw\09$3, $2, $0 {$1}", "=x,^Yk,x,x,~{dirflag},~{fpsr},~{flags}"(i32 %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}})
27 __m512i dst;
28 asm ("vpaddw\t%3, %2, %0 %{%1%}"
29 : "=x" (dst) //output
30 : "Yk" (msk), "x" (x), "x" (y)); //inputs
31 return dst;
32 }
33
mask_Yk_i64(long long msk,__m512i x,__m512i y)34 __m512i mask_Yk_i64(long long msk, __m512i x, __m512i y){
35 // CHECK: <8 x i64> asm "vpaddb\09$3, $2, $0 {$1}", "=x,^Yk,x,x,~{dirflag},~{fpsr},~{flags}"(i64 %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}})
36 __m512i dst;
37 asm ("vpaddb\t%3, %2, %0 %{%1%}"
38 : "=x" (dst) //output
39 : "Yk" (msk), "x" (x), "x" (y)); //inputs
40 return dst;
41 }
42
k_wise_op_i8(char msk_src1,char msk_src2)43 char k_wise_op_i8(char msk_src1,char msk_src2){
44 //CHECK: i8 asm "kandb\09$2, $1, $0", "=k,k,k,~{dirflag},~{fpsr},~{flags}"(i8 %{{.*}}, i8 %{{.*}})
45 char msk_dst;
46 asm ("kandb\t%2, %1, %0"
47 : "=k" (msk_dst)
48 : "k" (msk_src1), "k" (msk_src2));
49 return msk_dst;
50 }
51
k_wise_op_i16(short msk_src1,short msk_src2)52 short k_wise_op_i16(short msk_src1, short msk_src2){
53 //CHECK: i16 asm "kandw\09$2, $1, $0", "=k,k,k,~{dirflag},~{fpsr},~{flags}"(i16 %{{.*}}, i16 %{{.*}})
54 short msk_dst;
55 asm ("kandw\t%2, %1, %0"
56 : "=k" (msk_dst)
57 : "k" (msk_src1), "k" (msk_src2));
58 return msk_dst;
59 }
60
k_wise_op_i32(int msk_src1,int msk_src2)61 int k_wise_op_i32(int msk_src1, int msk_src2){
62 //CHECK: i32 asm "kandd\09$2, $1, $0", "=k,k,k,~{dirflag},~{fpsr},~{flags}"(i32 %{{.*}}, i32 %{{.*}})
63 int msk_dst;
64 asm ("kandd\t%2, %1, %0"
65 : "=k" (msk_dst)
66 : "k" (msk_src1), "k" (msk_src2));
67 return msk_dst;
68 }
69
k_wise_op_i64(long long msk_src1,long long msk_src2)70 long long k_wise_op_i64(long long msk_src1, long long msk_src2){
71 //CHECK: i64 asm "kandq\09$2, $1, $0", "=k,k,k,~{dirflag},~{fpsr},~{flags}"(i64 %{{.*}}, i64 %{{.*}})
72 long long msk_dst;
73 asm ("kandq\t%2, %1, %0"
74 : "=k" (msk_dst)
75 : "k" (msk_src1), "k" (msk_src2));
76 return msk_dst;
77 }
78