Showing the changes to test/encoder/EncUT_SVC_me.cpp (1 of 8 files). Show all
Unified · Split
| @@ -0,0 +1,157 @@ | |||
| 1 | #include <gtest/gtest.h> | ||
| 2 | #include <math.h> | ||
| 3 | #include <stdlib.h> | ||
| 4 | #include <time.h> | ||
| 5 | |||
| 6 | #include "cpu_core.h" | ||
| 7 | #include "cpu.h" | ||
| 8 | #include "macros.h" | ||
| 9 | #include "svc_motion_estimate.h" | ||
| 10 | |||
| 11 | using namespace WelsSVCEnc; | ||
| 12 | #define SVC_ME_TEST_NUM 10 | ||
| 13 | static void FillWithRandomData (uint8_t* p, int32_t Len) { | ||
| 14 | for (int32_t i = 0; i < Len; i++) { | ||
| 15 | p[i] = rand() % 256; | ||
| 16 | } | ||
| 17 | } | ||
| 18 | |||
| 19 | //preprocess related | ||
| 20 | int32_t SumOf8x8SingleBlock_ref (uint8_t* pRef, const int32_t kiRefStride) { | ||
| 21 | int32_t iSum = 0, i; | ||
| 22 | for (i = 0; i < 8; i++) { | ||
| 23 | iSum += pRef[0] + pRef[1] + pRef[2] + pRef[3]; | ||
| 24 | iSum += pRef[4] + pRef[5] + pRef[6] + pRef[7]; | ||
| 25 | pRef += kiRefStride; | ||
| 26 | } | ||
| 27 | return iSum; | ||
| 28 | } | ||
| 29 | int32_t SumOf16x16SingleBlock_ref (uint8_t* pRef, const int32_t kiRefStride) { | ||
| 30 | int32_t iSum = 0, i; | ||
| 31 | for (i = 0; i < 16; i++) { | ||
| 32 | iSum += pRef[0] + pRef[1] + pRef[2] + pRef[3]; | ||
| 33 | iSum += pRef[4] + pRef[5] + pRef[6] + pRef[7]; | ||
| 34 | iSum += pRef[8] + pRef[9] + pRef[10] + pRef[11]; | ||
| 35 | iSum += pRef[12] + pRef[13] + pRef[14] + pRef[15]; | ||
| 36 | pRef += kiRefStride; | ||
| 37 | } | ||
| 38 | return iSum; | ||
| 39 | } | ||
| 40 | |||
| 41 | void SumOf8x8BlockOfFrame_ref (uint8_t* pRefPicture, const int32_t kiWidth, const int32_t kiHeight, | ||
| 42 | const int32_t kiRefStride, | ||
| 43 | uint16_t* pFeatureOfBlock, uint32_t pTimesOfFeatureValue[]) { | ||
| 44 | int32_t x, y; | ||
| 45 | uint8_t* pRef; | ||
| 46 | uint16_t* pBuffer; | ||
| 47 | int32_t iSum; | ||
| 48 | for (y = 0; y < kiHeight; y++) { | ||
| 49 | pRef = pRefPicture + kiRefStride * y; | ||
| 50 | pBuffer = pFeatureOfBlock + kiWidth * y; | ||
| 51 | for (x = 0; x < kiWidth; x++) { | ||
| 52 | iSum = SumOf8x8SingleBlock_c (pRef + x, kiRefStride); | ||
| 53 | |||
| 54 | pBuffer[x] = iSum; | ||
| 55 | pTimesOfFeatureValue[iSum]++; | ||
| 56 | } | ||
| 57 | } | ||
| 58 | } | ||
| 59 | |||
| 60 | void SumOf16x16BlockOfFrame_ref (uint8_t* pRefPicture, const int32_t kiWidth, const int32_t kiHeight, | ||
| 61 | const int32_t kiRefStride, | ||
| 62 | uint16_t* pFeatureOfBlock, uint32_t pTimesOfFeatureValue[]) { | ||
| 63 | //TODO: this is similar to SumOf8x8BlockOfFrame_c expect the calling of single block func, refactor-able? | ||
| 64 | int32_t x, y; | ||
| 65 | uint8_t* pRef; | ||
| 66 | uint16_t* pBuffer; | ||
| 67 | int32_t iSum; | ||
| 68 | for (y = 0; y < kiHeight; y++) { | ||
| 69 | pRef = pRefPicture + kiRefStride * y; | ||
| 70 | pBuffer = pFeatureOfBlock + kiWidth * y; | ||
| 71 | for (x = 0; x < kiWidth; x++) { | ||
| 72 | iSum = SumOf16x16SingleBlock_c (pRef + x, kiRefStride); | ||
| 73 | |||
| 74 | pBuffer[x] = iSum; | ||
| 75 | pTimesOfFeatureValue[iSum]++; | ||
| 76 | } | ||
| 77 | } | ||
| 78 | } | ||
| 79 | |||
| 80 | #define GENERATE_SumOfSingleBlock(anchor, method) \ | ||
| 81 | TEST (SVC_ME_FunTest, method) {\ | ||
| 82 | ENFORCE_STACK_ALIGN_1D (uint8_t, uiRefBuf, 16*320, 16);\ | ||
| 83 | int32_t iRes[2];\ | ||
| 84 | for (int32_t k = 0; k < SVC_ME_TEST_NUM; k++) {\ | ||
| 85 | FillWithRandomData (uiRefBuf,16*320);\ | ||
| 86 | iRes[0] = anchor (uiRefBuf,320);\ | ||
| 87 | iRes[1] = method (uiRefBuf,320);\ | ||
| 88 | ASSERT_EQ (iRes[0], iRes[1]);\ | ||
| 89 | }\ | ||
| 90 | } | ||
| 91 | |||
| 92 | GENERATE_SumOfSingleBlock (SumOf8x8SingleBlock_ref, SumOf8x8SingleBlock_c) | ||
| 93 | GENERATE_SumOfSingleBlock (SumOf16x16SingleBlock_ref, SumOf16x16SingleBlock_c) | ||
| 94 | |||
| 95 | #ifdef HAVE_NEON | ||
| 96 | GENERATE_SumOfSingleBlock (SumOf8x8SingleBlock_ref, SumOf8x8SingleBlock_neon) | ||
| 97 | GENERATE_SumOfSingleBlock (SumOf16x16SingleBlock_ref, SumOf16x16SingleBlock_neon) | ||
| 98 | #endif | ||
| 99 | |||
| 100 | #ifdef HAVE_NEON_AARCH64 | ||
| 101 | GENERATE_SumOfSingleBlock (SumOf8x8SingleBlock_ref, SumOf8x8SingleBlock_AArch64_neon) | ||
| 102 | GENERATE_SumOfSingleBlock (SumOf16x16SingleBlock_ref, SumOf16x16SingleBlock_AArch64_neon) | ||
| 103 | #endif | ||
| 104 | |||
| 105 | |||
| 106 | #define ENFORCE_NEW_ALIGN_1D(_tp, _nm, _nbuff, _sz, _al) \ | ||
| 107 | _tp *_nbuff = new _tp[(_sz)+(_al)-1]; \ | ||
| 108 | _tp *_nm = _nbuff + ((_al)-1) - (((uintptr_t)(_nbuff + ((_al)-1)) & ((_al)-1))/sizeof(_tp)); | ||
| 109 | |||
| 110 | #define GENERATE_SumOfFrame(anchor, method, kiWidth, kiHeight) \ | ||
| 111 | TEST (SVC_ME_FunTest, method##_##kiWidth##x##kiHeight) {\ | ||
| 112 | ENFORCE_NEW_ALIGN_1D (uint8_t, pRefPicture, pRefPictureBuff, ((kiHeight+16)*((((kiWidth+15)>>4)<<4)+16)), 16) \ | ||
| 113 | ENFORCE_NEW_ALIGN_1D (uint16_t, pFeatureOfBlock1, pFeatureOfBlockBuff1, (kiWidth*kiHeight), 16) \ | ||
| 114 | ENFORCE_NEW_ALIGN_1D (uint16_t, pFeatureOfBlock2, pFeatureOfBlockBuff2, (kiWidth*kiHeight), 16) \ | ||
| 115 | uint32_t pTimesOfFeatureValue[2][65536]; \ | ||
| 116 | for (int32_t k = 0; k < SVC_ME_TEST_NUM; k++) {\ | ||
| 117 | FillWithRandomData (pRefPicture,(kiHeight+16)*((((kiWidth+15)>>4)<<4)+16));\ | ||
| 118 | memset(pTimesOfFeatureValue[0], 0, 65536*sizeof(uint32_t)); \ | ||
| 119 | memset(pTimesOfFeatureValue[1], 0, 65536*sizeof(uint32_t)); \ | ||
| 120 | anchor (pRefPicture,kiWidth,kiHeight,((((kiWidth+15)>>4)<<4)+16),pFeatureOfBlock1,pTimesOfFeatureValue[0]); \ | ||
| 121 | method (pRefPicture,kiWidth,kiHeight,((((kiWidth+15)>>4)<<4)+16),pFeatureOfBlock2,pTimesOfFeatureValue[1]); \ | ||
| 122 | for(int32_t j=0;j<kiWidth*kiHeight;j++){\ | ||
| 123 | ASSERT_EQ (pFeatureOfBlock1[j], pFeatureOfBlock2[j]);\ | ||
| 124 | }\ | ||
| 125 | for(int32_t j=0;j<65536;j++){\ | ||
| 126 | ASSERT_EQ (pTimesOfFeatureValue[0][j], pTimesOfFeatureValue[1][j]);\ | ||
| 127 | }\ | ||
| 128 | }\ | ||
| 129 | delete[] pRefPictureBuff; \ | ||
| 130 | delete[] pFeatureOfBlockBuff1; \ | ||
| 131 | delete[] pFeatureOfBlockBuff2; \ | ||
| 132 | } | ||
| 133 | |||
| 134 | GENERATE_SumOfFrame (SumOf8x8BlockOfFrame_ref, SumOf8x8BlockOfFrame_c, 1, 1) | ||
| 135 | GENERATE_SumOfFrame (SumOf16x16BlockOfFrame_ref, SumOf16x16BlockOfFrame_c, 1, 1) | ||
| 136 | GENERATE_SumOfFrame (SumOf8x8BlockOfFrame_ref, SumOf8x8BlockOfFrame_c, 1, 320) | ||
| 137 | GENERATE_SumOfFrame (SumOf16x16BlockOfFrame_ref, SumOf16x16BlockOfFrame_c, 1, 320) | ||
| 138 | GENERATE_SumOfFrame (SumOf8x8BlockOfFrame_ref, SumOf8x8BlockOfFrame_c, 640, 320) | ||
| 139 | GENERATE_SumOfFrame (SumOf16x16BlockOfFrame_ref, SumOf16x16BlockOfFrame_c, 640, 320) | ||
| 140 | |||
| 141 | #ifdef HAVE_NEON | ||
| 142 | GENERATE_SumOfFrame (SumOf8x8BlockOfFrame_ref, SumOf8x8BlockOfFrame_neon, 1, 1) | ||
| 143 | GENERATE_SumOfFrame (SumOf16x16BlockOfFrame_ref, SumOf16x16BlockOfFrame_neon, 1, 1) | ||
| 144 | GENERATE_SumOfFrame (SumOf8x8BlockOfFrame_ref, SumOf8x8BlockOfFrame_neon, 1, 320) | ||
| 145 | GENERATE_SumOfFrame (SumOf16x16BlockOfFrame_ref, SumOf16x16BlockOfFrame_neon, 1, 320) | ||
| 146 | GENERATE_SumOfFrame (SumOf8x8BlockOfFrame_ref, SumOf8x8BlockOfFrame_neon, 640, 320) | ||
| 147 | GENERATE_SumOfFrame (SumOf16x16BlockOfFrame_ref, SumOf16x16BlockOfFrame_neon, 640, 320) | ||
| 148 | #endif | ||
| 149 | |||
| 150 | #ifdef HAVE_NEON_AARCH64 | ||
| 151 | GENERATE_SumOfFrame (SumOf8x8BlockOfFrame_ref, SumOf8x8BlockOfFrame_AArch64_neon, 1, 1) | ||
| 152 | GENERATE_SumOfFrame (SumOf16x16BlockOfFrame_ref, SumOf16x16BlockOfFrame_AArch64_neon, 1, 1) | ||
| 153 | GENERATE_SumOfFrame (SumOf8x8BlockOfFrame_ref, SumOf8x8BlockOfFrame_AArch64_neon, 1, 320) | ||
| 154 | GENERATE_SumOfFrame (SumOf16x16BlockOfFrame_ref, SumOf16x16BlockOfFrame_AArch64_neon, 1, 320) | ||
| 155 | GENERATE_SumOfFrame (SumOf8x8BlockOfFrame_ref, SumOf8x8BlockOfFrame_AArch64_neon, 640, 320) | ||
| 156 | GENERATE_SumOfFrame (SumOf16x16BlockOfFrame_ref, SumOf16x16BlockOfFrame_AArch64_neon, 640, 320) | ||
| 157 | #endif | ||