Neko 1.99.9
A portable framework for high-order spectral element flow simulations
Loading...
Searching...
No Matches
ax_helm_svv_one_sided.cu
Go to the documentation of this file.
1/*
2 Copyright (c) 2025-2026, 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 <stdlib.h>
37#include <device/device_config.h>
38#include <device/cuda/check.h>
39#include "ax_helm_svv_one_sided_kernel.h"
40
41extern "C" {
42
46void cuda_ax_helm_svv_one_sided(
47 void *w, void *u, void *dx, void *dy, void *dz, void *h1,
48 void *drdx, void *drdy, void *drdz,
49 void *dsdx, void *dsdy, void *dsdz,
50 void *dtdx, void *dtdy, void *dtdz,
51 void *jacinv, void *w3, void *h1_svv,
52 void *filter_r, void *filter_s, void *filter_t,
53 int *nelv, int *lx) {
54
55 const dim3 threads(*lx, *lx, 1);
56 const dim3 blocks(*nelv, 1, 1);
57 const cudaStream_t stream = (cudaStream_t) glb_cmd_queue;
58 const size_t shared_size =
59 2 * (*lx) * (*lx) * (*lx) * sizeof(real);
60 static bool shared_configured[17] = {false};
61
62#define LAUNCH(LX) \
63 ax_helm_svv_one_sided_kernel<real, LX> \
64 <<<blocks, threads, shared_size, stream>>>( \
65 (real *) w, (real *) u, \
66 (real *) dx, (real *) dy, (real *) dz, (real *) h1, \
67 (real *) drdx, (real *) drdy, (real *) drdz, \
68 (real *) dsdx, (real *) dsdy, (real *) dsdz, \
69 (real *) dtdx, (real *) dtdy, (real *) dtdz, \
70 (real *) jacinv, (real *) w3, (real *) h1_svv, \
71 (real *) filter_r, (real *) filter_s, (real *) filter_t); \
72 CUDA_CHECK(cudaGetLastError())
73
74#define CASE(LX) \
75 case LX: \
76 LAUNCH(LX); \
77 break
78
79// Double precision exceeds the default 48 KiB shared-memory limit at LX >= 15.
80#define CASE_LARGE(LX) \
81 case LX: \
82 if (!shared_configured[LX]) { \
83 CUDA_CHECK(cudaFuncSetAttribute( \
84 ax_helm_svv_one_sided_kernel<real, LX>, \
85 cudaFuncAttributeMaxDynamicSharedMemorySize, shared_size)); \
86 shared_configured[LX] = true; \
87 } \
88 LAUNCH(LX); \
89 break
90
91 switch (*lx) {
92 CASE(2);
93 CASE(3);
94 CASE(4);
95 CASE(5);
96 CASE(6);
97 CASE(7);
98 CASE(8);
99 CASE(9);
100 CASE(10);
101 CASE(11);
102 CASE(12);
103 CASE(13);
104 CASE(14);
105 CASE_LARGE(15);
106 CASE_LARGE(16);
107 default:
108 fprintf(stderr, __FILE__ ": size not supported: %d\n", *lx);
109 exit(1);
110 }
111}
112
113}
cuda_ax_helm_svv_one_sided
void cuda_ax_helm_svv_one_sided(void *w, void *u, void *dx, void *dy, void *dz, void *h1, void *drdx, void *drdy, void *drdz, void *dsdx, void *dsdy, void *dsdz, void *dtdx, void *dtdy, void *dtdz, void *jacinv, void *w3, void *h1_svv, void *filter_r, void *filter_s, void *filter_t, int *nelv, int *lx)
Definition ax_helm_svv_one_sided.cu:46
CASE
#define CASE(LX)
CASE_LARGE
#define CASE_LARGE(LX)
ale_add_kinematics_kernel
__global__ void ale_add_kinematics_kernel(const int n, T *__restrict__ wx, T *__restrict__ wy, T *__restrict__ wz, const T *__restrict__ x_ref, const T *__restrict__ y_ref, const T *__restrict__ z_ref, const T *__restrict__ phi, const T *__restrict__ x, const T *__restrict__ y, const T *__restrict__ z, const kinematics_params_t kin_params)
Definition ale_kinematics_kernel.h:52
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
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
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
w
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ w
Definition ax_helm_full_kernel.h:48
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
dz
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dz
Definition ax_helm_full_kernel.h:51
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
dx
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dx
Definition ax_helm_full_kernel.h:49
u
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ u
Definition ax_helm_full_kernel.h:46
dy
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dy
Definition ax_helm_full_kernel.h:50
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
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
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
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
h1
__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__ h1
Definition ax_helm_full_kernel.h:52
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
nelv
__global__ void 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 int nelv
Definition ax_helm_kernel.h:185
w3
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ w3
Definition cdtp_kernel.h:117
device_config.h
real
double real
Definition device_config.h:12
ax_helm_svv_one_sided_kernel.h