| 1 |
#include <arm_neon.h> |
| 2 |
#include "arm-neon-ref.h" |
| 3 |
#include "compute-ref-data.h" |
| 4 |
|
| 5 |
/* Expected results. */ |
| 6 |
VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, |
| 7 |
0xf6, 0xf6, 0xf6, 0xf6 }; |
| 8 |
VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2 }; |
| 9 |
VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff0 }; |
| 10 |
VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffffd }; |
| 11 |
VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf3, 0xf3, 0xf3, 0xf3, |
| 12 |
0xf7, 0xf7, 0xf7, 0xf7 }; |
| 13 |
VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2 }; |
| 14 |
VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff0 }; |
| 15 |
VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffff1 }; |
| 16 |
VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf3, 0xf3, 0xf3, 0xf3, |
| 17 |
0xf7, 0xf7, 0xf7, 0xf7 }; |
| 18 |
VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2 }; |
| 19 |
#if MFLOAT8_SUPPORTED |
| 20 |
VECT_VAR_DECL(expected,hmfloat,8,8) [] = { 0xfa, 0xfa, 0xfa, 0xfa, |
| 21 |
0xfe, 0xfe, 0xfe, 0xfe }; |
| 22 |
#endif |
| 23 |
#if defined (FP16_SUPPORTED) |
| 24 |
VECT_VAR_DECL (expected, hfloat, 16, 4) [] = { 0xcc09, 0xcb89, |
| 25 |
0xcb09, 0xca89 }; |
| 26 |
#endif |
| 27 |
VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800004, 0xc1700004 }; |
| 28 |
VECT_VAR_DECL(expected,int,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, |
| 29 |
0xf6, 0xf6, 0xf6, 0xf6, |
| 30 |
0xf2, 0xf2, 0xf2, 0xf2, |
| 31 |
0xf6, 0xf6, 0xf6, 0xf6 }; |
| 32 |
VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2, |
| 33 |
0xfff4, 0xfff4, 0xfff6, 0xfff6 }; |
| 34 |
VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff0, |
| 35 |
0xfffffff2, 0xfffffff2 }; |
| 36 |
VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffffd, |
| 37 |
0xfffffffffffffffd }; |
| 38 |
VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf3, 0xf3, 0xf3, 0xf3, |
| 39 |
0xf7, 0xf7, 0xf7, 0xf7, |
| 40 |
0xf3, 0xf3, 0xf3, 0xf3, |
| 41 |
0xf7, 0xf7, 0xf7, 0xf7 }; |
| 42 |
VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2, |
| 43 |
0xfff4, 0xfff4, 0xfff6, 0xfff6 }; |
| 44 |
VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff0, |
| 45 |
0xfffffff2, 0xfffffff2 }; |
| 46 |
VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffff1, |
| 47 |
0xfffffff1 }; |
| 48 |
VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf3, 0xf3, 0xf3, 0xf3, |
| 49 |
0xf7, 0xf7, 0xf7, 0xf7, |
| 50 |
0xf3, 0xf3, 0xf3, 0xf3, |
| 51 |
0xf7, 0xf7, 0xf7, 0xf7 }; |
| 52 |
VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2, |
| 53 |
0xfff4, 0xfff4, 0xfff6, 0xfff6 }; |
| 54 |
#if MFLOAT8_SUPPORTED |
| 55 |
VECT_VAR_DECL(expected,hmfloat,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, |
| 56 |
0xf5, 0xf5, 0xf5, 0xf5, |
| 57 |
0xf1, 0xf1, 0xf1, 0xf1, |
| 58 |
0xf5, 0xf5, 0xf5, 0xf5 }; |
| 59 |
#endif |
| 60 |
#if defined (FP16_SUPPORTED) |
| 61 |
VECT_VAR_DECL (expected, hfloat, 16, 8) [] = { 0xcc09, 0xcb89, |
| 62 |
0xcb09, 0xca89, |
| 63 |
0xca09, 0xc989, |
| 64 |
0xc909, 0xc889 }; |
| 65 |
#endif |
| 66 |
VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800001, 0xc1700001, |
| 67 |
0xc1600001, 0xc1500001 }; |
| 68 |
|
| 69 |
#define TEST_MSG "VBSL/VBSLQ" |
| 70 |
void exec_vbsl (void) |
| 71 |
{ |
| 72 |
/* Basic test: y=vbsl(unsigned_vec,x1,x2), then store the result. */ |
| 73 |
#define TEST_VBSL(T3, Q, T1, T2, W, N) \ |
| 74 |
VECT_VAR(vector_res, T1, W, N) = \ |
| 75 |
vbsl##Q##_##T2##W(VECT_VAR(vector_first, T3, W, N), \ |
| 76 |
VECT_VAR(vector, T1, W, N), \ |
| 77 |
VECT_VAR(vector2, T1, W, N)); \ |
| 78 |
vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N), VECT_VAR(vector_res, T1, W, N)) |
| 79 |
|
| 80 |
DECL_VARIABLE_ALL_VARIANTS(vector); |
| 81 |
DECL_VARIABLE_ALL_VARIANTS(vector2); |
| 82 |
DECL_VARIABLE_ALL_VARIANTS(vector_res); |
| 83 |
|
| 84 |
DECL_VARIABLE_UNSIGNED_VARIANTS(vector_first); |
| 85 |
|
| 86 |
clean_results (); |
| 87 |
|
| 88 |
TEST_MACRO_ALL_VARIANTS_2_5(VLOAD, vector, buffer); |
| 89 |
#if MFLOAT8_SUPPORTED |
| 90 |
VLOAD(vector, buffer, , mfloat, mf, 8, 8); |
| 91 |
VLOAD(vector, buffer, q, mfloat, mf, 8, 16); |
| 92 |
#endif |
| 93 |
#if defined (FP16_SUPPORTED) |
| 94 |
VLOAD(vector, buffer, , float, f, 16, 4); |
| 95 |
VLOAD(vector, buffer, q, float, f, 16, 8); |
| 96 |
#endif |
| 97 |
VLOAD(vector, buffer, , float, f, 32, 2); |
| 98 |
VLOAD(vector, buffer, q, float, f, 32, 4); |
| 99 |
|
| 100 |
/* Choose init value arbitrarily, will be used for vector |
| 101 |
comparison. As we want different values for each type variant, we |
| 102 |
can't use generic initialization macros. */ |
| 103 |
VDUP(vector2, , int, s, 8, 8, -10); |
| 104 |
VDUP(vector2, , int, s, 16, 4, -14); |
| 105 |
VDUP(vector2, , int, s, 32, 2, -30); |
| 106 |
VDUP(vector2, , int, s, 64, 1, -33); |
| 107 |
VDUP(vector2, , uint, u, 8, 8, 0xF3); |
| 108 |
VDUP(vector2, , uint, u, 16, 4, 0xFFF2); |
| 109 |
VDUP(vector2, , uint, u, 32, 2, 0xFFFFFFF0); |
| 110 |
VDUP(vector2, , uint, u, 64, 1, 0xFFFFFFF3); |
| 111 |
MFLOAT8_ONLY(VDUP(vector2, , mfloat, mf, 8, 8, MFLOAT8(0xca));) |
| 112 |
#if defined (FP16_SUPPORTED) |
| 113 |
VDUP(vector2, , float, f, 16, 4, -2.4f); /* -2.4f is 0xC0CD. */ |
| 114 |
#endif |
| 115 |
VDUP(vector2, , float, f, 32, 2, -30.3f); |
| 116 |
VDUP(vector2, , poly, p, 8, 8, 0xF3); |
| 117 |
VDUP(vector2, , poly, p, 16, 4, 0xFFF2); |
| 118 |
|
| 119 |
VDUP(vector2, q, int, s, 8, 16, -10); |
| 120 |
VDUP(vector2, q, int, s, 16, 8, -14); |
| 121 |
VDUP(vector2, q, int, s, 32, 4, -30); |
| 122 |
VDUP(vector2, q, int, s, 64, 2, -33); |
| 123 |
VDUP(vector2, q, uint, u, 8, 16, 0xF3); |
| 124 |
VDUP(vector2, q, uint, u, 16, 8, 0xFFF2); |
| 125 |
VDUP(vector2, q, uint, u, 32, 4, 0xFFFFFFF0); |
| 126 |
VDUP(vector2, q, uint, u, 64, 2, 0xFFFFFFF3); |
| 127 |
VDUP(vector2, q, poly, p, 8, 16, 0xF3); |
| 128 |
VDUP(vector2, q, poly, p, 16, 8, 0xFFF2); |
| 129 |
MFLOAT8_ONLY(VDUP(vector2, q, mfloat, mf, 8, 16, MFLOAT8(0x55));) |
| 130 |
#if defined (FP16_SUPPORTED) |
| 131 |
VDUP(vector2, q, float, f, 16, 8, -2.4f); |
| 132 |
#endif |
| 133 |
VDUP(vector2, q, float, f, 32, 4, -30.4f); |
| 134 |
|
| 135 |
VDUP(vector_first, , uint, u, 8, 8, 0xF4); |
| 136 |
VDUP(vector_first, , uint, u, 16, 4, 0xFFF6); |
| 137 |
VDUP(vector_first, , uint, u, 32, 2, 0xFFFFFFF2); |
| 138 |
VDUP(vector_first, , uint, u, 64, 1, 0xFFFFFFF2); |
| 139 |
VDUP(vector_first, q, uint, u, 8, 16, 0xF4); |
| 140 |
VDUP(vector_first, q, uint, u, 16, 8, 0xFFF6); |
| 141 |
VDUP(vector_first, q, uint, u, 32, 4, 0xFFFFFFF2); |
| 142 |
VDUP(vector_first, q, uint, u, 64, 2, 0xFFFFFFF2); |
| 143 |
|
| 144 |
/* Execute the tests. */ |
| 145 |
TEST_MACRO_ALL_VARIANTS_1_5(TEST_VBSL, uint); |
| 146 |
TEST_VBSL(uint, , poly, p, 8, 8); |
| 147 |
TEST_VBSL(uint, , poly, p, 16, 4); |
| 148 |
TEST_VBSL(uint, q, poly, p, 8, 16); |
| 149 |
TEST_VBSL(uint, q, poly, p, 16, 8); |
| 150 |
#if MFLOAT8_SUPPORTED |
| 151 |
TEST_VBSL(uint, , mfloat, mf, 8, 8); |
| 152 |
TEST_VBSL(uint, q, mfloat, mf, 8, 16); |
| 153 |
#endif |
| 154 |
#if defined (FP16_SUPPORTED) |
| 155 |
TEST_VBSL(uint, , float, f, 16, 4); |
| 156 |
TEST_VBSL(uint, q, float, f, 16, 8); |
| 157 |
#endif |
| 158 |
TEST_VBSL(uint, , float, f, 32, 2); |
| 159 |
TEST_VBSL(uint, q, float, f, 32, 4); |
| 160 |
|
| 161 |
#if defined (FP16_SUPPORTED) |
| 162 |
CHECK_RESULTS (TEST_MSG, ""); |
| 163 |
#else |
| 164 |
CHECK_RESULTS_NO_FP16 (TEST_MSG, ""); |
| 165 |
#endif |
| 166 |
} |
| 167 |
|
| 168 |
int main (void) |
| 169 |
{ |
| 170 |
exec_vbsl (); |
| 171 |
return 0; |
| 172 |
} |