Home
last modified time | relevance | path

Searched refs:thread_queue (Results 1 – 7 of 7) sorted by relevance

/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/gpu/cuda_impl/
Dtopk_impl.cu37 …_Pragma("unroll") for (int i = thread_queue - 1; i > 0; --i) { …
48 template <typename T, typename S, int warp_queue, int thread_queue, int threads_per_block, bool is_…
99 template <typename T, typename S, int warp_queue, int thread_queue, int threads_per_block, bool is_…
115 _Pragma("unroll") for (int i = 0; i < thread_queue; ++i) { in TopKStep()
137 bool needSort = (num_vals == thread_queue); in TopKStep()
141 MergeWarpQueue<T, S, warp_queue, thread_queue, is_descend>(threadK, threadV, warp_K, warp_V); in TopKStep()
144 _Pragma("unroll") for (int i = 0; i < thread_queue; ++i) { in TopKStep()
157 MergeWarpQueue<T, S, warp_queue, thread_queue, is_descend>(threadK, threadV, warp_K, warp_V); in TopKStep()
161 …TopKInBuffer<T, S, warp_queue, thread_queue, threads_per_block, is_descend>(shared_K, shared_V, wa… in TopKStep()
177 template <typename T, typename S, int warp_queue, int thread_queue, int threads_per_block, bool is_…
[all …]
Drcwm_small_impl.cu31 _Pragma("unroll") for (int i = thread_queue - 1; i > 0; --i) { \
42 template <typename T, typename S, typename K, int warp_queue, int thread_queue, int threads_per_blo…
54 T threadK[thread_queue]; // NOLINT in L2Rcwm()
55 S threadV[thread_queue]; // NOLINT in L2Rcwm()
67 _Pragma("unroll") for (int i = 0; i < thread_queue; ++i) { in L2Rcwm()
93 bool needSort = (num_vals == thread_queue); in L2Rcwm()
97 MergeWarpQueue<T, S, warp_queue, thread_queue, is_descend>(threadK, threadV, warp_K, warp_V); in L2Rcwm()
100 _Pragma("unroll") for (int i = 0; i < thread_queue; ++i) { in L2Rcwm()
114 MergeWarpQueue<T, S, warp_queue, thread_queue, is_descend>(threadK, threadV, warp_K, warp_V); in L2Rcwm()
Dtopk_lib.cuh459 template <typename T, typename S, int warp_queue, int thread_queue, bool is_descend>
462 SortWarpByRegister<T, S, thread_queue, !is_descend>(threadK, threadV); in MergeWarpQueue()
472 …MergeWarpByRegister<T, S, kWarpQueueRegisters, thread_queue, !is_descend, false>(warp_KRegisters, … in MergeWarpQueue()
/third_party/mesa3d/src/gallium/drivers/zink/
Dzink_screen.h135 VkQueue thread_queue; //gfx+compute member
Dzink_screen.c1220 vkGetDeviceQueue(screen->dev, screen->gfx_queue, 1, &screen->thread_queue); in init_queue()
1222 screen->thread_queue = screen->queue; in init_queue()
1612 if (n->VKSCR(QueueSubmit)(n->screen->threaded ? n->screen->thread_queue : n->screen->queue, in noop_submit()
Dzink_batch.c622 bs->queue = screen->thread_queue; in zink_end_batch()
Dzink_bo.c698 VkQueue queue = screen->threaded ? screen->thread_queue : screen->queue; in do_commit_single()