1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
|
// RUN: %clang_cc1 %s -O0 -ffreestanding -triple=x86_64-apple-darwin -target-cpu skylake-avx512 -emit-llvm -o - -Wall -Werror | FileCheck %s
// This test checks validity of att\gcc style inline assmebly for avx512 k and Yk constraints.
// Also checks mask register allows flexible type (size <= 64 bit)
#include <x86intrin.h>
__m512i mask_Yk_i8(char msk, __m512i x, __m512i y){
// 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> %{{.*}})
__m512i dst;
asm ("vpaddq\t%3, %2, %0 %{%1%}"
: "=x" (dst) //output
: "Yk" (msk), "x" (x), "x" (y)); //inputs
return dst;
}
__m512i mask_Yk_i16(short msk, __m512i x, __m512i y){
// 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> %{{.*}})
__m512i dst;
asm ("vpaddd\t%3, %2, %0 %{%1%}"
: "=x" (dst) //output
: "Yk" (msk), "x" (x), "x" (y)); //inputs
return dst;
}
__m512i mask_Yk_i32(int msk, __m512i x, __m512i y){
// 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> %{{.*}})
__m512i dst;
asm ("vpaddw\t%3, %2, %0 %{%1%}"
: "=x" (dst) //output
: "Yk" (msk), "x" (x), "x" (y)); //inputs
return dst;
}
__m512i mask_Yk_i64(long long msk, __m512i x, __m512i y){
// 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> %{{.*}})
__m512i dst;
asm ("vpaddb\t%3, %2, %0 %{%1%}"
: "=x" (dst) //output
: "Yk" (msk), "x" (x), "x" (y)); //inputs
return dst;
}
char k_wise_op_i8(char msk_src1,char msk_src2){
//CHECK: <8 x i1> asm "kandb\09$2, $1, $0", "=k,k,k,~{dirflag},~{fpsr},~{flags}"(<8 x i1> %{{.*}}, <8 x i1> %{{.*}})
char msk_dst;
asm ("kandb\t%2, %1, %0"
: "=k" (msk_dst)
: "k" (msk_src1), "k" (msk_src2));
return msk_dst;
}
short k_wise_op_i16(short msk_src1, short msk_src2){
//CHECK: <16 x i1> asm "kandw\09$2, $1, $0", "=k,k,k,~{dirflag},~{fpsr},~{flags}"(<16 x i1> %{{.*}}, <16 x i1> %{{.*}})
short msk_dst;
asm ("kandw\t%2, %1, %0"
: "=k" (msk_dst)
: "k" (msk_src1), "k" (msk_src2));
return msk_dst;
}
int k_wise_op_i32(int msk_src1, int msk_src2){
//CHECK: <32 x i1> asm "kandd\09$2, $1, $0", "=k,k,k,~{dirflag},~{fpsr},~{flags}"(<32 x i1> %{{.*}}, <32 x i1> %{{.*}})
int msk_dst;
asm ("kandd\t%2, %1, %0"
: "=k" (msk_dst)
: "k" (msk_src1), "k" (msk_src2));
return msk_dst;
}
long long k_wise_op_i64(long long msk_src1, long long msk_src2){
//CHECK: <64 x i1> asm "kandq\09$2, $1, $0", "=k,k,k,~{dirflag},~{fpsr},~{flags}"(<64 x i1> %{{.*}}, <64 x i1> %{{.*}})
long long msk_dst;
asm ("kandq\t%2, %1, %0"
: "=k" (msk_dst)
: "k" (msk_src1), "k" (msk_src2));
return msk_dst;
}
|
