Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 1 | /* |
| 2 | |
| 3 | Copyright (c) 2009, 2010, 2011 STMicroelectronics |
| 4 | Written by Christophe Lyon |
| 5 | |
| 6 | Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | of this software and associated documentation files (the "Software"), to deal |
| 8 | in the Software without restriction, including without limitation the rights |
| 9 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell |
| 10 | copies of the Software, and to permit persons to whom the Software is |
| 11 | furnished to do so, subject to the following conditions: |
| 12 | |
| 13 | The above copyright notice and this permission notice shall be included in |
| 14 | all copies or substantial portions of the Software. |
| 15 | |
| 16 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN |
| 22 | THE SOFTWARE. |
| 23 | |
| 24 | */ |
| 25 | |
Christophe Lyon | 1775be0 | 2014-07-10 13:46:54 +0200 | [diff] [blame] | 26 | #if defined(__arm__) || defined(__aarch64__) |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 27 | #include <arm_neon.h> |
| 28 | #else |
Christophe Lyon | 0dab5f7 | 2011-07-19 17:14:09 +0200 | [diff] [blame] | 29 | #include "stm-arm-neon.h" |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 30 | #endif |
| 31 | |
| 32 | #include "stm-arm-neon-ref.h" |
| 33 | |
| 34 | #define TEST_MSG "VMUL_LANE" |
| 35 | void exec_vmul_lane (void) |
| 36 | { |
| 37 | #define DECL_VMUL(VAR) \ |
| 38 | DECL_VARIABLE(VAR, int, 16, 4); \ |
| 39 | DECL_VARIABLE(VAR, int, 32, 2); \ |
| 40 | DECL_VARIABLE(VAR, uint, 16, 4); \ |
| 41 | DECL_VARIABLE(VAR, uint, 32, 2); \ |
| 42 | DECL_VARIABLE(VAR, float, 32, 2); \ |
| 43 | DECL_VARIABLE(VAR, int, 16, 8); \ |
| 44 | DECL_VARIABLE(VAR, int, 32, 4); \ |
| 45 | DECL_VARIABLE(VAR, uint, 16, 8); \ |
| 46 | DECL_VARIABLE(VAR, uint, 32, 4); \ |
| 47 | DECL_VARIABLE(VAR, float, 32, 4) |
| 48 | |
| 49 | /* vector_res = vmul_lane(vector,vector2,lane), then store the result. */ |
| 50 | #define TEST_VMUL_LANE(Q, T1, T2, W, N, N2, L) \ |
| 51 | VECT_VAR(vector_res, T1, W, N) = \ |
| 52 | vmul##Q##_lane_##T2##W(VECT_VAR(vector, T1, W, N), \ |
| 53 | VECT_VAR(vector2, T1, W, N2), \ |
| 54 | L); \ |
| 55 | vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N), \ |
| 56 | VECT_VAR(vector_res, T1, W, N)) |
| 57 | |
| 58 | /* With ARM RVCT, we need to declare variables before any executable |
| 59 | statement */ |
| 60 | DECL_VMUL(vector); |
| 61 | DECL_VMUL(vector_res); |
| 62 | |
| 63 | DECL_VARIABLE(vector2, int, 16, 4); |
| 64 | DECL_VARIABLE(vector2, int, 32, 2); |
| 65 | DECL_VARIABLE(vector2, uint, 16, 4); |
| 66 | DECL_VARIABLE(vector2, uint, 32, 2); |
| 67 | DECL_VARIABLE(vector2, float, 32, 2); |
| 68 | |
| 69 | clean_results (); |
| 70 | |
| 71 | /* Initialize vector from pre-initialized values */ |
Christophe Lyon | f205367 | 2014-12-16 10:26:00 +0100 | [diff] [blame] | 72 | VLOAD(vector, buffer, , int, s, 16, 4); |
| 73 | VLOAD(vector, buffer, , int, s, 32, 2); |
| 74 | VLOAD(vector, buffer, , uint, u, 16, 4); |
| 75 | VLOAD(vector, buffer, , uint, u, 32, 2); |
| 76 | VLOAD(vector, buffer, , float, f, 32, 2); |
| 77 | VLOAD(vector, buffer, q, int, s, 16, 8); |
| 78 | VLOAD(vector, buffer, q, int, s, 32, 4); |
| 79 | VLOAD(vector, buffer, q, uint, u, 16, 8); |
| 80 | VLOAD(vector, buffer, q, uint, u, 32, 4); |
| 81 | VLOAD(vector, buffer, q, float, f, 32, 4); |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 82 | |
| 83 | /* Initialize vector2 */ |
Christophe Lyon | f205367 | 2014-12-16 10:26:00 +0100 | [diff] [blame] | 84 | VDUP(vector2, , int, s, 16, 4, 0x4); |
| 85 | VDUP(vector2, , int, s, 32, 2, 0x22); |
| 86 | VDUP(vector2, , uint, u, 16, 4, 0x444); |
| 87 | VDUP(vector2, , uint, u, 32, 2, 0x532); |
| 88 | VDUP(vector2, , float, f, 32, 2, 22.8f); |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 89 | |
| 90 | /* Choose lane arbitrarily */ |
| 91 | TEST_VMUL_LANE(, int, s, 16, 4, 4, 2); |
| 92 | TEST_VMUL_LANE(, int, s, 32, 2, 2, 1); |
| 93 | TEST_VMUL_LANE(, uint, u, 16, 4, 4, 2); |
| 94 | TEST_VMUL_LANE(, uint, u, 32, 2, 2, 1); |
| 95 | TEST_VMUL_LANE(, float, f, 32, 2, 2, 1); |
| 96 | TEST_VMUL_LANE(q, int, s, 16, 8, 4, 2); |
| 97 | TEST_VMUL_LANE(q, int, s, 32, 4, 2, 0); |
| 98 | TEST_VMUL_LANE(q, uint, u, 16, 8, 4, 2); |
| 99 | TEST_VMUL_LANE(q, uint, u, 32, 4, 2, 1); |
| 100 | TEST_VMUL_LANE(q, float, f, 32, 4, 2, 0); |
| 101 | |
| 102 | /* FIXME: only a subset of the result buffers are used, but we |
| 103 | output all of them */ |
| 104 | dump_results_hex (TEST_MSG); |
| 105 | } |