/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/x64/gemm/ |
H A D | gemm_utils.hpp | 53 while (nthr_m * nthr_n > 4 * nthrs) { in calc_nthr_2d() 63 while (nthr_m * nthr_n > nthrs) { in calc_nthr_2d() 84 while (nthr_m * nthr_n < 0.95 * nthrs) { in calc_nthr_2d() 104 if ((nthr_m * nthr_n > nthrs)) { in calc_nthr_2d() 107 nthr_m = (int)sqrt((double)nthrs); in calc_nthr_2d() 110 nthr_n = nthrs / nthr_m; in calc_nthr_2d() 112 while ((nthr_m > 1) && (nthr_m * nthr_n != nthrs)) { in calc_nthr_2d() 114 nthr_n = nthrs / nthr_m; in calc_nthr_2d() 117 nthr_n = (int)sqrt((double)nthrs); in calc_nthr_2d() 120 nthr_m = nthrs / nthr_n; in calc_nthr_2d() [all …]
|
H A D | gemm_partition.hpp | 32 static inline void partition_1d(const int ithr, const int nthrs, const dim_t n, in partition_1d() argument 35 dim_t band = n / nthrs; in partition_1d() 37 dim_t tail = n - (nthrs - 1) * band; in partition_1d() 39 tail = n - (nthrs - 1) * band; in partition_1d() 41 if (ithr < (nthrs - 1)) in partition_1d() 56 static inline void partition_2d(const int ithr, int *nthrs, const int ithr_i, in partition_2d() argument 96 *nthrs = mthr_used * nthr_used; in partition_2d() 98 if (ithr < *nthrs) { in partition_2d() 128 if (ithr < *nthrs) { in partition_2d()
|
H A D | gemm_driver.cpp | 1047 = ((n > nthrs * N2D_MAX) || (n <= nthrs * N2D_MAX / 2)) in set_thread_opts_nopack() 1082 int nthrs_n = nthrs; in set_thread_opts_nopack() 1146 if (m > n && (m >= nthrs * veclen || n < nthrs)) { in set_thread_opts_nopack() 1231 nthrs--; in set_thread_opts_pack() 1289 if (nocopy_checker(nthrs, arg)) { in set_thread_opts() 1446 dim_t band = (sizeM + nthrs - 1) / nthrs; in parallel_a_copy() 1544 if (m > 10 * n && n < *nthrs) in adjust_thread_count() 1551 int i = *nthrs; in adjust_thread_count() 1557 *nthrs = 1; in adjust_thread_count() 1567 *nthrs = 1; in adjust_thread_count() [all …]
|
H A D | gemm_threading.hpp | 52 int nthrs() const { return nthrs_m * nthrs_n * nthrs_k; } in nthrs() function 75 partition_1d(ithr, nthrs(), m, off_m, size_m); in get_thread_slice() 80 partition_1d(ithr, nthrs(), n, off_n, size_n); in get_thread_slice() 84 int nthr_eff = nthrs(); in get_thread_slice()
|
H A D | gemm_pack_storage.hpp | 56 int nthr() const { return single_nocopy() ? 1 : threading().nthrs(); } in nthr()
|
/dports/math/onednn/oneDNN-2.5.1/src/cpu/x64/gemm/ |
H A D | gemm_utils.hpp | 53 while (nthr_m * nthr_n > 4 * nthrs) { in calc_nthr_2d() 63 while (nthr_m * nthr_n > nthrs) { in calc_nthr_2d() 84 while (nthr_m * nthr_n < 0.95 * nthrs) { in calc_nthr_2d() 104 if ((nthr_m * nthr_n > nthrs)) { in calc_nthr_2d() 107 nthr_m = (int)sqrt((double)nthrs); in calc_nthr_2d() 110 nthr_n = nthrs / nthr_m; in calc_nthr_2d() 112 while ((nthr_m > 1) && (nthr_m * nthr_n != nthrs)) { in calc_nthr_2d() 114 nthr_n = nthrs / nthr_m; in calc_nthr_2d() 117 nthr_n = (int)sqrt((double)nthrs); in calc_nthr_2d() 120 nthr_m = nthrs / nthr_n; in calc_nthr_2d() [all …]
|
H A D | gemm_partition.hpp | 32 static inline void partition_1d(const int ithr, const int nthrs, const dim_t n, in partition_1d() argument 35 dim_t band = n / nthrs; in partition_1d() 37 dim_t tail = n - (nthrs - 1) * band; in partition_1d() 39 tail = n - (nthrs - 1) * band; in partition_1d() 41 if (ithr < (nthrs - 1)) in partition_1d() 56 static inline void partition_2d(const int ithr, int *nthrs, const int ithr_i, in partition_2d() argument 96 *nthrs = mthr_used * nthr_used; in partition_2d() 98 if (ithr < *nthrs) { in partition_2d() 128 if (ithr < *nthrs) { in partition_2d()
|
H A D | gemm_driver.cpp | 1047 = ((n > nthrs * N2D_MAX) || (n <= nthrs * N2D_MAX / 2)) in set_thread_opts_nopack() 1082 int nthrs_n = nthrs; in set_thread_opts_nopack() 1146 if (m > n && (m >= nthrs * veclen || n < nthrs)) { in set_thread_opts_nopack() 1231 nthrs--; in set_thread_opts_pack() 1289 if (nocopy_checker(nthrs, arg)) { in set_thread_opts() 1446 dim_t band = (sizeM + nthrs - 1) / nthrs; in parallel_a_copy() 1544 if (m > 10 * n && n < *nthrs) in adjust_thread_count() 1551 int i = *nthrs; in adjust_thread_count() 1557 *nthrs = 1; in adjust_thread_count() 1567 *nthrs = 1; in adjust_thread_count() [all …]
|
H A D | gemm_threading.hpp | 52 int nthrs() const { return nthrs_m * nthrs_n * nthrs_k; } in nthrs() function 75 partition_1d(ithr, nthrs(), m, off_m, size_m); in get_thread_slice() 80 partition_1d(ithr, nthrs(), n, off_n, size_n); in get_thread_slice() 84 int nthr_eff = nthrs(); in get_thread_slice()
|
H A D | gemm_pack_storage.hpp | 56 int nthr() const { return single_nocopy() ? 1 : threading().nthrs(); } in nthr()
|
/dports/net/nng/nng-1.5.2/src/core/ |
H A D | taskq.c | 238 int nthrs; in nni_taskq_sys_init() local 241 nthrs = nni_plat_ncpu() * 2; in nni_taskq_sys_init() 243 nthrs = NNG_NUM_TASKQ_THREADS; in nni_taskq_sys_init() 246 if (nthrs > NNG_MAX_TASKQ_THREADS) { in nni_taskq_sys_init() 247 nthrs = NNG_MAX_TASKQ_THREADS; in nni_taskq_sys_init() 251 return (nni_taskq_init(&nni_taskq_systq, nthrs)); in nni_taskq_sys_init()
|
/dports/devel/folly/folly-2021.12.27.00/folly/test/ |
H A D | SpinLockTest.cpp | 98 int nthrs = sysconf(_SC_NPROCESSORS_ONLN) * 2; in correctnessTest() local 101 for (int i = 0; i < nthrs; ++i) { in correctnessTest() 111 int nthrs = sysconf(_SC_NPROCESSORS_ONLN) + 4; in trylockTest() local 115 for (int i = 0; i < nthrs; ++i) { in trylockTest()
|
H A D | PackedSyncPtrTest.cpp | 91 const int nthrs = 297; variable 109 for (intptr_t i = 0; i < nthrs; i++) { in TEST() 122 EXPECT_EQ(idsFound.size(), nthrs); // check they are all there in TEST()
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/ |
H A D | nchw_pooling.hpp | 156 size_t nthrs = dnnl_get_max_threads(); in init_scratchpad() local 160 src_sz_ * nthrs * channel_block_size_); in init_scratchpad() 162 dst_sz_ * nthrs * channel_block_size_); in init_scratchpad() 172 dim_t nthrs = dnnl_get_max_threads(); in calculate_channel_block_size() local 173 dim_t C_per_thr = nstl::min(MB() * IC() / nthrs, IC()); in calculate_channel_block_size()
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/gemm/f32/ |
H A D | gemm_utils_f32.cpp | 40 void calc_nthr_nocopy_avx(dim_t m, dim_t n, dim_t k, int nthrs, int *nthrs_m, in calc_nthr_nocopy_avx() argument 44 if (nthrs == 1) { in calc_nthr_nocopy_avx() 58 nthr = nthrs; in calc_nthr_nocopy_avx() 157 void calc_nthr_nocopy_avx512_common(dim_t m, dim_t n, dim_t k, int nthrs, in calc_nthr_nocopy_avx512_common() argument 162 if (nthrs == 1) { in calc_nthr_nocopy_avx512_common() 175 nthr = nthrs; in calc_nthr_nocopy_avx512_common() 180 nthr = nthrs; in calc_nthr_nocopy_avx512_common()
|
H A D | gemm_utils_f32.hpp | 55 void calc_nthr_nocopy_avx512_common(dim_t m, dim_t n, dim_t k, int nthrs, 59 void calc_nthr_nocopy_avx(dim_t m, dim_t n, dim_t k, int nthrs, int *nthrs_m,
|
/dports/math/onednn/oneDNN-2.5.1/src/cpu/gemm/f32/ |
H A D | gemm_utils_f32.cpp | 40 void calc_nthr_nocopy_avx(dim_t m, dim_t n, dim_t k, int nthrs, int *nthrs_m, in calc_nthr_nocopy_avx() argument 44 if (nthrs == 1) { in calc_nthr_nocopy_avx() 58 nthr = nthrs; in calc_nthr_nocopy_avx() 157 void calc_nthr_nocopy_avx512_common(dim_t m, dim_t n, dim_t k, int nthrs, in calc_nthr_nocopy_avx512_common() argument 162 if (nthrs == 1) { in calc_nthr_nocopy_avx512_common() 175 nthr = nthrs; in calc_nthr_nocopy_avx512_common() 180 nthr = nthrs; in calc_nthr_nocopy_avx512_common()
|
H A D | gemm_utils_f32.hpp | 55 void calc_nthr_nocopy_avx512_common(dim_t m, dim_t n, dim_t k, int nthrs, 59 void calc_nthr_nocopy_avx(dim_t m, dim_t n, dim_t k, int nthrs, int *nthrs_m,
|
/dports/devel/folly/folly-2021.12.27.00/folly/experimental/test/ |
H A D | RelaxedConcurrentPriorityQueueTest.cpp | 1060 std::vector<int> nthrs = {1, 2, 4, 8, 12, 14, 16, 28, 32, 56}; in TEST() local 1072 for (int i : nthrs) { in TEST() 1087 std::vector<int> nthrs = {1, 2, 4, 8, 12, 16, 24}; in TEST() local 1099 for (int m : nthrs) { in TEST() 1100 for (int n : nthrs) { in TEST() 1117 for (int m : nthrs) { in TEST() 1118 for (int n : nthrs) { in TEST() 1136 for (int m : nthrs) { in TEST() 1137 for (int n : nthrs) { in TEST()
|
/dports/devel/folly/folly-2021.12.27.00/folly/synchronization/test/ |
H A D | SmallLocksTest.cpp | 121 const int nthrs = 17; in doPslTest() local 123 for (int i = 0; i < nthrs; ++i) { in doPslTest() 151 int nthrs = sysconf(_SC_NPROCESSORS_ONLN) * 2; in TEST() local 153 for (int i = 0; i < nthrs; ++i) { in TEST()
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/x64/gemm/f32/ |
H A D | jit_avx512_common_gemm_f32.hpp | 29 dnnl_status_t jit_avx512_common_gemm_f32(int nthrs, const char *transa,
|
H A D | jit_avx_gemm_f32.hpp | 29 dnnl_status_t jit_avx_gemm_f32(int nthrs, const char *transa,
|
/dports/math/onednn/oneDNN-2.5.1/src/cpu/x64/gemm/f32/ |
H A D | jit_avx512_common_gemm_f32.hpp | 29 dnnl_status_t jit_avx512_common_gemm_f32(int nthrs, const char *transa,
|
H A D | jit_avx_gemm_f32.hpp | 29 dnnl_status_t jit_avx_gemm_f32(int nthrs, const char *transa,
|
/dports/math/onednn/oneDNN-2.5.1/src/cpu/x64/ |
H A D | jit_uni_deconv_zp_pad_str_kernel.cpp | 322 const int nthrs = (work_amount / jcp.nthr) > parallelization_ratio_thr in compute_deconv_zp_pad_str_comp_ker() local 326 parallel(nthrs, [&](const int ithr, const int nthr) { in compute_deconv_zp_pad_str_comp_ker()
|