Neko 1.99.1
A portable framework for high-order spectral element flow simulations
Loading...
Searching...
No Matches
opr_cfl.cu
Go to the documentation of this file.
1/*
2 Copyright (c) 2022-2025, The Neko Authors
3 All rights reserved.
4
5 Redistribution and use in source and binary forms, with or without
6 modification, are permitted provided that the following conditions
7 are met:
8
9 * Redistributions of source code must retain the above copyright
10 notice, this list of conditions and the following disclaimer.
11
12 * Redistributions in binary form must reproduce the above
13 copyright notice, this list of conditions and the following
14 disclaimer in the documentation and/or other materials provided
15 with the distribution.
16
17 * Neither the name of the authors nor the names of its
18 contributors may be used to endorse or promote products derived
19 from this software without specific prior written permission.
20
21 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
29 CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
31 ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32 POSSIBILITY OF SUCH DAMAGE.
33*/
34
35#include <stdio.h>
36#include "cfl_kernel.h"
37#include <device/device_config.h>
38#include <device/cuda/check.h>
39
40
41#ifdef HAVE_NVSHMEM
42#include <nvshmem.h>
43#include <nvshmemx.h>
44#endif
45
46extern "C" {
47
48#include <math/bcknd/device/device_mpi_reduce.h>
49#include <math/bcknd/device/device_mpi_op.h>
50
51#ifdef HAVE_NCCL
52#include <math/bcknd/device/device_nccl_reduce.h>
53#include <math/bcknd/device/device_nccl_op.h>
54#endif
55
59 void *cfl_d = NULL;
60
64 real cuda_cfl(real *dt, void *u, void *v, void *w,
65 void *drdx, void *dsdx, void *dtdx,
66 void *drdy, void *dsdy, void *dtdy,
67 void *drdz, void *dsdz, void *dtdz,
68 void *dr_inv, void *ds_inv, void *dt_inv,
69 void *jacinv, int *nel, int *lx) {
70
71 const dim3 nthrds(1024, 1, 1);
72 const dim3 nblcks((*nel), 1, 1);
73 const cudaStream_t stream = (cudaStream_t) glb_cmd_queue;
74
75 if (cfl_d == NULL) {
76#ifdef HAVE_NVSHMEM
77 cfl_d = (real *) nvshmem_malloc(sizeof(real));
78#else
79 CUDA_CHECK(cudaMalloc(&cfl_d, (*nel) * sizeof(real)));
80#endif
81 }
82
83#define CASE(LX) \
84 case LX: \
85 cfl_kernel<real, LX, 1024> \
86 <<<nblcks, nthrds, 0, stream>>> \
87 (*dt, (real *) u, (real *) v, (real *) w, \
88 (real *) drdx, (real *) dsdx, (real *) dtdx, \
89 (real *) drdy, (real *) dsdy, (real *) dtdy, \
90 (real *) drdz, (real *) dsdz, (real *) dtdz, \
91 (real *) dr_inv, (real *) ds_inv, (real *) dt_inv, \
92 (real *) jacinv, (real *) cfl_d); \
93 CUDA_CHECK(cudaGetLastError()); \
94 break
95
96 switch(*lx) {
97 CASE(2);
98 CASE(3);
99 CASE(4);
100 CASE(5);
101 CASE(6);
102 CASE(7);
103 CASE(8);
104 CASE(9);
105 CASE(10);
106 default:
107 {
108 fprintf(stderr, __FILE__ ": size not supported: %d\n", *lx);
109 exit(1);
110 }
111 }
112
113 cfl_reduce_kernel<real><<<1, 1024, 0, stream>>> ((real *) cfl_d, (*nel));
114 CUDA_CHECK(cudaGetLastError());
115
116 real cfl;
117#ifdef HAVE_NCCL
118 device_nccl_allreduce(cfl_d, cfl_d, 1, sizeof(real),
119 DEVICE_NCCL_MAX, stream);
120 CUDA_CHECK(cudaMemcpyAsync(&cfl, cfl_d, sizeof(real),
121 cudaMemcpyDeviceToHost, stream));
122 cudaStreamSynchronize(stream);
123#elif HAVE_NVSHMEM
124 if (sizeof(real) == sizeof(float)) {
125 nvshmemx_float_max_reduce_on_stream(NVSHMEM_TEAM_WORLD,
126 (float *) cfl_d,
127 (float *) cfl_d, 1, stream);
128 }
129 else if (sizeof(real) == sizeof(double)) {
130 nvshmemx_double_max_reduce_on_stream(NVSHMEM_TEAM_WORLD,
131 (double *) cfl_d,
132 (double *) cfl_d, 1, stream);
133 }
134 CUDA_CHECK(cudaMemcpyAsync(&cfl, cfl_d, sizeof(real),
135 cudaMemcpyDeviceToHost, stream));
136 cudaStreamSynchronize(stream);
137#elif HAVE_DEVICE_MPI
138 cudaStreamSynchronize(stream);
139 device_mpi_allreduce(cfl_d, &cfl, 1, sizeof(real), DEVICE_MPI_MAX);
140#else
141 CUDA_CHECK(cudaMemcpyAsync(&cfl, cfl_d, sizeof(real),
142 cudaMemcpyDeviceToHost, stream));
143 cudaStreamSynchronize(stream);
144#endif
145 return cfl;
146 }
147}
drdy
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ drdy
Definition ax_helm_full_kernel.h:54
drdz
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ drdz
Definition ax_helm_full_kernel.h:55
dsdz
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dsdz
Definition ax_helm_full_kernel.h:58
dsdy
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dsdy
Definition ax_helm_full_kernel.h:57
w
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ w
Definition ax_helm_full_kernel.h:48
dtdy
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dtdy
Definition ax_helm_full_kernel.h:60
u
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ u
Definition ax_helm_full_kernel.h:46
drdx
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ drdx
Definition ax_helm_full_kernel.h:53
dtdz
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dtdz
Definition ax_helm_full_kernel.h:61
dsdx
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dsdx
Definition ax_helm_full_kernel.h:56
v
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ v
Definition ax_helm_full_kernel.h:47
dtdx
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dtdx
Definition ax_helm_full_kernel.h:59
jacinv
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ jacinv
Definition ax_helm_full_kernel.h:62
dt
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dt
Definition cdtp_kernel.h:109
CUDA_CHECK
#define CUDA_CHECK(err)
Definition check.h:6
dirichlet_apply_scalar_kernel
__global__ void dirichlet_apply_scalar_kernel(const int *__restrict__ msk, T *__restrict__ x, const T g, const int m)
Definition dirichlet_kernel.h:42
device_config.h
real
double real
Definition device_config.h:12
device_mpi_op.h
DEVICE_MPI_MAX
#define DEVICE_MPI_MAX
Definition device_mpi_op.h:10
device_mpi_allreduce
void device_mpi_allreduce(void *buf_d, void *buf, int count, int nbytes, int op)
Definition device_mpi_reduce.c:46
device_mpi_reduce.h
device_nccl_op.h
DEVICE_NCCL_MAX
#define DEVICE_NCCL_MAX
Definition device_nccl_op.h:10
device_nccl_allreduce
void device_nccl_allreduce(void *sbuf_d, void *rbuf_d, int count, int nbytes, int op, void *stream)
Definition device_nccl_reduce.c:45
device_nccl_reduce.h
cfl_kernel.h
cfl_d
void * cfl_d
Definition opr_cfl.cu:59
CASE
#define CASE(LX)
cuda_cfl
real cuda_cfl(real *dt, void *u, void *v, void *w, void *drdx, void *dsdx, void *dtdx, void *drdy, void *dsdy, void *dtdy, void *drdz, void *dsdz, void *dtdz, void *dr_inv, void *ds_inv, void *dt_inv, void *jacinv, int *nel, int *lx)
Definition opr_cfl.cu:64