Lines Matching refs:train
93 __device__ void loopUnrolledCached(int queryIdx, const PtrStepSz<T>& query,volatile int imgIdx, const PtrStepSz<T>& train, const Mask& mask,
97 for (int t = 0, endt = (train.rows + BLOCK_SIZE - 1) / BLOCK_SIZE; t < endt; ++t)
108 if (loadX < train.cols)
112 ForceGlob<T>::Load(train.ptr(::min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
129 if (queryIdx < query.rows && trainIdx < train.rows && distVal < bestDistance && mask(queryIdx, trainIdx))
139 __global__ void matchUnrolledCached(const PtrStepSz<T> query, const PtrStepSz<T> train, const Mask mask, int* bestTrainIdx, float* bestDistance)
153 loopUnrolledCached<BLOCK_SIZE, MAX_DESC_LEN, Dist>(queryIdx, query, 0, train, mask, s_query, s_train, myBestDistance, myBestTrainIdx, myBestTrainIdx);
170 void matchUnrolledCached(const PtrStepSz<T>& query, const PtrStepSz<T>& train, const Mask& mask,
179 matchUnrolledCached<BLOCK_SIZE, MAX_DESC_LEN, Dist><<<grid, block, smemSize, stream>>>(query, train, mask, trainIdx.data, distance.data);
207 const PtrStepSz<T> train = trains[imgIdx];
209 loopUnrolledCached<BLOCK_SIZE, MAX_DESC_LEN, Dist>(queryIdx, query, imgIdx, train, m, s_query, s_train, myBestDistance, myBestTrainIdx, myBestImgIdx);
249 __device__ void loopUnrolled(int queryIdx, const PtrStepSz<T>& query,volatile int imgIdx, const PtrStepSz<T>& train, const Mask& mask,
253 for (int t = 0, endt = (train.rows + BLOCK_SIZE - 1) / BLOCK_SIZE; t < endt; ++t)
272 ForceGlob<T>::Load(train.ptr(::min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
289 if (queryIdx < query.rows && trainIdx < train.rows && distVal < bestDistance && mask(queryIdx, trainIdx))
299 __global__ void matchUnrolled(const PtrStepSz<T> query, const PtrStepSz<T> train, const Mask mask, int* bestTrainIdx, float* bestDistance)
311 loopUnrolled<BLOCK_SIZE, MAX_DESC_LEN, Dist>(queryIdx, query, 0, train, mask, s_query, s_train, myBestDistance, myBestTrainIdx, myBestTrainIdx);
328 void matchUnrolled(const PtrStepSz<T>& query, const PtrStepSz<T>& train, const Mask& mask,
337 matchUnrolled<BLOCK_SIZE, MAX_DESC_LEN, Dist><<<grid, block, smemSize, stream>>>(query, train, mask, trainIdx.data, distance.data);
363 const PtrStepSz<T> train = trains[imgIdx];
365 loopUnrolled<BLOCK_SIZE, MAX_DESC_LEN, Dist>(queryIdx, query, imgIdx, train, m, s_query, s_train, myBestDistance, myBestTrainIdx, myBestImgIdx);
405 __device__ void loop(int queryIdx, const PtrStepSz<T>& query, volatile int imgIdx, const PtrStepSz<T>& train, const Mask& mask,
409 for (int t = 0, endt = (train.rows + BLOCK_SIZE - 1) / BLOCK_SIZE; t < endt; ++t)
427 ForceGlob<T>::Load(train.ptr(::min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
444 if (queryIdx < query.rows && trainIdx < train.rows && distVal < bestDistance && mask(queryIdx, trainIdx))
454 __global__ void match(const PtrStepSz<T> query, const PtrStepSz<T> train, const Mask mask, int* bestTrainIdx, float* bestDistance)
466 loop<BLOCK_SIZE, Dist>(queryIdx, query, 0, train, mask, s_query, s_train, myBestDistance, myBestTrainIdx, myBestTrainIdx);
483 void match(const PtrStepSz<T>& query, const PtrStepSz<T>& train, const Mask& mask,
492 match<BLOCK_SIZE, Dist><<<grid, block, smemSize, stream>>>(query, train, mask, trainIdx.data, distance.data);
517 const PtrStepSz<T> train = trains[imgIdx];
519 loop<BLOCK_SIZE, Dist>(queryIdx, query, imgIdx, train, m, s_query, s_train, myBestDistance, myBestTrainIdx, myBestImgIdx);
559 void matchDispatcher(const PtrStepSz<T>& query, const PtrStepSz<T>& train, const Mask& mask,
565 matchUnrolledCached<16, 64, Dist>(query, train, mask, trainIdx, distance, stream);
569 matchUnrolledCached<16, 128, Dist>(query, train, mask, trainIdx, distance, stream);
573 matchUnrolled<16, 256, Dist>(query, train, mask, trainIdx, distance, stream);
577 matchUnrolled<16, 512, Dist>(query, train, mask, trainIdx, distance, stream);
581 matchUnrolled<16, 1024, Dist>(query, train, mask, trainIdx, distance, stream);
585 match<16, Dist>(query, train, mask, trainIdx, distance, stream);
623 template <typename T> void matchL1_gpu(const PtrStepSzb& query, const PtrStepSzb& train, const PtrStepSzb& mask,
629 matchDispatcher< L1Dist<T> >(static_cast< PtrStepSz<T> >(query), static_cast< PtrStepSz<T> >(train), SingleMask(mask),
635 matchDispatcher< L1Dist<T> >(static_cast< PtrStepSz<T> >(query), static_cast< PtrStepSz<T> >(train), WithOutMask(),
648 template <typename T> void matchL2_gpu(const PtrStepSzb& query, const PtrStepSzb& train, const PtrStepSzb& mask,
654 matchDispatcher<L2Dist>(static_cast< PtrStepSz<T> >(query), static_cast< PtrStepSz<T> >(train), SingleMask(mask),
660 matchDispatcher<L2Dist>(static_cast< PtrStepSz<T> >(query), static_cast< PtrStepSz<T> >(train), WithOutMask(),
673 template <typename T> void matchHamming_gpu(const PtrStepSzb& query, const PtrStepSzb& train, const PtrStepSzb& mask,
679 matchDispatcher<HammingDist>(static_cast< PtrStepSz<T> >(query), static_cast< PtrStepSz<T> >(train), SingleMask(mask),
685 matchDispatcher<HammingDist>(static_cast< PtrStepSz<T> >(query), static_cast< PtrStepSz<T> >(train), WithOutMask(),
