Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 1 | /* |
| 2 | |
Christophe Lyon | c94d4c1 | 2013-03-29 16:32:37 +0100 | [diff] [blame] | 3 | Copyright (c) 2009, 2010, 2011, 2012 STMicroelectronics |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 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 | #ifndef INSN_NAME |
| 35 | #define INSN_NAME vqdmlal_lane |
| 36 | #define TEST_MSG "VQDMLAL_LANE" |
| 37 | #endif |
| 38 | |
| 39 | #define FNNAME1(NAME) void exec_ ## NAME (void) |
| 40 | #define FNNAME(NAME) FNNAME1(NAME) |
| 41 | |
| 42 | FNNAME (INSN_NAME) |
| 43 | { |
| 44 | /* vector_res = vqdmlxl_lane(vector, vector3, vector4, lane), |
| 45 | then store the result. */ |
Christophe Lyon | 4a6e5cc | 2014-06-03 22:47:52 +0200 | [diff] [blame] | 46 | #define TEST_VQDMLXL_LANE1(INSN, T1, T2, W, W2, N, V) \ |
Christophe Lyon | c1cc782 | 2015-01-20 16:04:24 +0100 | [diff] [blame] | 47 | Set_Neon_Cumulative_Sat(0, VECT_VAR(vector_res, T1, W, N)); \ |
Christophe Lyon | 4a6e5cc | 2014-06-03 22:47:52 +0200 | [diff] [blame] | 48 | VECT_VAR(vector_res, T1, W, N) = \ |
| 49 | INSN##_##T2##W2(VECT_VAR(vector, T1, W, N), \ |
| 50 | VECT_VAR(vector3, T1, W2, N), \ |
| 51 | VECT_VAR(vector4, T1, W2, N), \ |
| 52 | V); \ |
| 53 | vst1q_##T2##W(VECT_VAR(result, T1, W, N), \ |
| 54 | VECT_VAR(vector_res, T1, W, N)); \ |
| 55 | dump_neon_cumulative_sat(TEST_MSG, xSTR(INSN##_##T2##W2), \ |
| 56 | xSTR(T1), W, N) |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 57 | |
| 58 | #define TEST_VQDMLXL_LANE(INSN, T1, T2, W, W2, N, V) \ |
| 59 | TEST_VQDMLXL_LANE1(INSN, T1, T2, W, W2, N, V) |
| 60 | |
| 61 | /* With ARM RVCT, we need to declare variables before any executable |
| 62 | statement */ |
| 63 | DECL_VARIABLE(vector, int, 32, 4); |
| 64 | DECL_VARIABLE(vector3, int, 16, 4); |
| 65 | DECL_VARIABLE(vector4, int, 16, 4); |
| 66 | DECL_VARIABLE(vector_res, int, 32, 4); |
| 67 | |
| 68 | DECL_VARIABLE(vector, int, 64, 2); |
| 69 | DECL_VARIABLE(vector3, int, 32, 2); |
| 70 | DECL_VARIABLE(vector4, int, 32, 2); |
| 71 | DECL_VARIABLE(vector_res, int, 64, 2); |
| 72 | |
| 73 | clean_results (); |
| 74 | |
Christophe Lyon | f205367 | 2014-12-16 10:26:00 +0100 | [diff] [blame] | 75 | VLOAD(vector, buffer, q, int, s, 32, 4); |
| 76 | VLOAD(vector, buffer, q, int, s, 64, 2); |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 77 | |
Christophe Lyon | f205367 | 2014-12-16 10:26:00 +0100 | [diff] [blame] | 78 | VDUP(vector3, , int, s, 16, 4, 0x55); |
| 79 | VDUP(vector4, , int, s, 16, 4, 0xBB); |
| 80 | VDUP(vector3, , int, s, 32, 2, 0x55); |
| 81 | VDUP(vector4, , int, s, 32, 2, 0xBB); |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 82 | |
Christophe Lyon | 4a6e5cc | 2014-06-03 22:47:52 +0200 | [diff] [blame] | 83 | fprintf(ref_file, "\n%s cumulative saturation output:\n", TEST_MSG); |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 84 | TEST_VQDMLXL_LANE(INSN_NAME, int, s, 32, 16, 4, 0); |
| 85 | TEST_VQDMLXL_LANE(INSN_NAME, int, s, 64, 32, 2, 0); |
| 86 | dump_results_hex (TEST_MSG); |
| 87 | |
Christophe Lyon | f205367 | 2014-12-16 10:26:00 +0100 | [diff] [blame] | 88 | VDUP(vector3, , int, s, 16, 4, 0); |
| 89 | VDUP(vector3, , int, s, 32, 2, 0); |
Christophe Lyon | 4a6e5cc | 2014-06-03 22:47:52 +0200 | [diff] [blame] | 90 | fprintf(ref_file, "\n%s cumulative saturation output:\n", |
| 91 | TEST_MSG " (mul with input=0)"); |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 92 | TEST_VQDMLXL_LANE(INSN_NAME, int, s, 32, 16, 4, 0); |
| 93 | TEST_VQDMLXL_LANE(INSN_NAME, int, s, 64, 32, 2, 0); |
| 94 | dump_results_hex2 (TEST_MSG, " (mul with input=0)"); |
| 95 | |
Christophe Lyon | f205367 | 2014-12-16 10:26:00 +0100 | [diff] [blame] | 96 | VDUP(vector3, , int, s, 16, 4, 0x8000); |
| 97 | VDUP(vector3, , int, s, 32, 2, 0x80000000); |
| 98 | VDUP(vector4, , int, s, 16, 4, 0x8000); |
| 99 | VDUP(vector4, , int, s, 32, 2, 0x80000000); |
Christophe Lyon | 4a6e5cc | 2014-06-03 22:47:52 +0200 | [diff] [blame] | 100 | fprintf(ref_file, "\n%s cumulative saturation output:\n", |
| 101 | TEST_MSG " (check mul cumulative saturation)"); |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 102 | TEST_VQDMLXL_LANE(INSN_NAME, int, s, 32, 16, 4, 0); |
| 103 | TEST_VQDMLXL_LANE(INSN_NAME, int, s, 64, 32, 2, 0); |
Christophe Lyon | 4a6e5cc | 2014-06-03 22:47:52 +0200 | [diff] [blame] | 104 | dump_results_hex2 (TEST_MSG, " (check mul cumulative saturation)"); |
Christophe Lyon | 073831a | 2011-01-24 17:37:40 +0100 | [diff] [blame] | 105 | } |