Home
last modified time | relevance | path

Searched refs:NB (Results 1 – 4 of 4) sorted by relevance

/libCEED/include/ceed/jit-source/magma/
H A Dmagma-basis-interp-deriv-nontensor.h13 template <typename T, int Q_COMP, int P, int Q, int NB>
19 const int nblocks = (n + NB - 1) / NB; in magma_basis_nontensor_device_n()
20 const int myn = min(NB, n - id * NB); in magma_basis_nontensor_device_n()
22 dB += id * P * NB; in magma_basis_nontensor_device_n()
23 dC += id * Q * NB; in magma_basis_nontensor_device_n()
26 CeedScalar *sB = shared_data + ty * P * NB; in magma_basis_nontensor_device_n()
27 CeedScalar *sA = shared_data + blockDim.y * P * NB; in magma_basis_nontensor_device_n()
31 read_B_g2s_1D_nosync<CeedScalar, Q, P, NB>(tx, myn, dB, sB); in magma_basis_nontensor_device_n()
40 CeedScalar rC[NB]; in magma_basis_nontensor_device_n()
41 mul_rAsBrC_1D_nosync<CeedScalar, Q, P, NB>(rA, sB, rC); in magma_basis_nontensor_device_n()
[all …]
H A Dmagma-common-nontensor.h69 template <typename T, int P, int Q, int NB>
73 if (n != NB) { in read_B_g2s_1D_nosync()
79 for (i = 0; i < Q * NB - P; i += P) { in read_B_g2s_1D_nosync()
93 template <typename T, int P, int Q, int NB>
94 static __device__ __inline__ void write_C_r2g_1D_nosync(const int tx, const int n, T rC[NB], T *dC)… in write_C_r2g_1D_nosync() argument
95 if (n != NB) { in write_C_r2g_1D_nosync()
101 for (int i = 0; i < NB; i++) { in write_C_r2g_1D_nosync()
112 template <typename T, int P, int Q, int NB>
113 static __device__ __inline__ void sum_C_r2g_1D_nosync(const int tx, const int n, T rC[NB], T *dC) { in sum_C_r2g_1D_nosync() argument
114 if (n != NB) { in sum_C_r2g_1D_nosync()
[all …]
/libCEED/backends/magma/tuning/
H A DREADME.md6 blocking factor parameter, `NB`, which varies with `P` and `Q` as well as the
10 generate the optimal `NB` selections for a new target architecture.
15 `NB` from 1 to 32 and saved to `a100_rtc.h`, is:
/libCEED/backends/magma/
H A Dceed-magma-basis.c382 CeedInt q_comp, NB, M, K; in CeedBasisApplyNonTensorCore_Magma() local
411 NB = impl->NB_interp_t[iN]; in CeedBasisApplyNonTensorCore_Magma()
414 NB = impl->NB_interp[iN]; in CeedBasisApplyNonTensorCore_Magma()
419 NB = impl->NB_deriv_t[iN]; in CeedBasisApplyNonTensorCore_Magma()
422 NB = impl->NB_deriv[iN]; in CeedBasisApplyNonTensorCore_Magma()
426 CeedInt grid = CeedDivUpInt(N, num_t_col * NB); in CeedBasisApplyNonTensorCore_Magma()
428 CeedInt shared_mem_B = num_t_col * K * NB * sizeof(CeedScalar); in CeedBasisApplyNonTensorCore_Magma()