Home | History | Annotate | Download | only in cuda

Lines Matching full:queryidx

82         __device__ void loadQueryToSmem(int queryIdx, const PtrStepSz<T>& query, U* s_query)
88 s_query[threadIdx.y * MAX_DESC_LEN + loadX] = loadX < query.cols ? query.ptr(::min(queryIdx, query.rows - 1))[loadX] : 0;
93 __device__ void loopUnrolledCached(int queryIdx, const PtrStepSz<T>& query,volatile int imgIdx, const PtrStepSz<T>& train, const Mask& mask,
129 if (queryIdx < query.rows && trainIdx < train.rows && distVal < bestDistance && mask(queryIdx, trainIdx))
143 const int queryIdx = blockIdx.x * BLOCK_SIZE + threadIdx.y;
148 loadQueryToSmem<BLOCK_SIZE, MAX_DESC_LEN>(queryIdx, query, s_query);
153 loopUnrolledCached<BLOCK_SIZE, MAX_DESC_LEN, Dist>(queryIdx, query, 0, train, mask, s_query, s_train, myBestDistance, myBestTrainIdx, myBestTrainIdx);
162 if (queryIdx < query.rows && threadIdx.x == 0)
164 bestTrainIdx[queryIdx] = myBestTrainIdx;
165 bestDistance[queryIdx] = myBestDistance;
192 const int queryIdx = blockIdx.x * BLOCK_SIZE + threadIdx.y;
197 loadQueryToSmem<BLOCK_SIZE, MAX_DESC_LEN>(queryIdx, query, s_query);
209 loopUnrolledCached<BLOCK_SIZE, MAX_DESC_LEN, Dist>(queryIdx, query, imgIdx, train, m, s_query, s_train, myBestDistance, myBestTrainIdx, myBestImgIdx);
220 if (queryIdx < query.rows && threadIdx.x == 0)
222 bestTrainIdx[queryIdx] = myBestTrainIdx;
223 bestImgIdx[queryIdx] = myBestImgIdx;
224 bestDistance[queryIdx] = myBestDistance;
249 __device__ void loopUnrolled(int queryIdx, const PtrStepSz<T>& query,volatile int imgIdx, const PtrStepSz<T>& train, const Mask& mask,
269 ForceGlob<T>::Load(query.ptr(::min(queryIdx, query.rows - 1)), loadX, val);
289 if (queryIdx < query.rows && trainIdx < train.rows && distVal < bestDistance && mask(queryIdx, trainIdx))
303 const int queryIdx = blockIdx.x * BLOCK_SIZE + threadIdx.y;
311 loopUnrolled<BLOCK_SIZE, MAX_DESC_LEN, Dist>(queryIdx, query, 0, train, mask, s_query, s_train, myBestDistance, myBestTrainIdx, myBestTrainIdx);
320 if (queryIdx < query.rows && threadIdx.x == 0)
322 bestTrainIdx[queryIdx] = myBestTrainIdx;
323 bestDistance[queryIdx] = myBestDistance;
350 const int queryIdx = blockIdx.x * BLOCK_SIZE + threadIdx.y;
365 loopUnrolled<BLOCK_SIZE, MAX_DESC_LEN, Dist>(queryIdx, query, imgIdx, train, m, s_query, s_train, myBestDistance, myBestTrainIdx, myBestImgIdx);
376 if (queryIdx < query.rows && threadIdx.x == 0)
378 bestTrainIdx[queryIdx] = myBestTrainIdx;
379 bestImgIdx[queryIdx] = myBestImgIdx;
380 bestDistance[queryIdx] = myBestDistance;
405 __device__ void loop(int queryIdx, const PtrStepSz<T>& query, volatile int imgIdx, const PtrStepSz<T>& train, const Mask& mask,
424 ForceGlob<T>::Load(query.ptr(::min(queryIdx, query.rows - 1)), loadX, val);
444 if (queryIdx < query.rows && trainIdx < train.rows && distVal < bestDistance && mask(queryIdx, trainIdx))
458 const int queryIdx = blockIdx.x * BLOCK_SIZE + threadIdx.y;
466 loop<BLOCK_SIZE, Dist>(queryIdx, query, 0, train, mask, s_query, s_train, myBestDistance, myBestTrainIdx, myBestTrainIdx);
475 if (queryIdx < query.rows && threadIdx.x == 0)
477 bestTrainIdx[queryIdx] = myBestTrainIdx;
478 bestDistance[queryIdx] = myBestDistance;
505 const int queryIdx = blockIdx.x * BLOCK_SIZE + threadIdx.y;
519 loop<BLOCK_SIZE, Dist>(queryIdx, query, imgIdx, train, m, s_query, s_train, myBestDistance, myBestTrainIdx, myBestImgIdx);
530 if (queryIdx < query.rows && threadIdx.x == 0)
532 bestTrainIdx[queryIdx] = myBestTrainIdx;
533 bestImgIdx[queryIdx] = myBestImgIdx;
534 bestDistance[queryIdx] = myBestDistance;