Neko 1.99.9
A portable framework for high-order spectral element flow simulations
Loading...
Searching...
No Matches
vreman_nut.c
Go to the documentation of this file.
1/*
2 Copyright (c) 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#ifdef __APPLE__
36#include <OpenCL/cl.h>
37#else
38#include <CL/cl.h>
39#endif
40
41#include <math.h>
42
43#include <device/device_config.h>
44#include <device/opencl/check.h>
45#include <device/opencl/jit.h>
46#include <device/opencl/prgm_lib.h>
47
48#include "vreman_nut_kernel.cl.h"
49
51void opencl_vreman_nut_compute(void *a11, void *a12, void *a13,
52 void *a21, void *a22, void *a23,
53 void *a31, void *a32, void *a33,
54 void *delta, void *nut, void *mult,
55 real *c, real *eps, int *n) {
56 cl_int err;
57
58 if (vreman_nut_program == NULL) {
59 opencl_kernel_jit(vreman_nut_kernel, (cl_program *) &vreman_nut_program);
60 }
61
62 cl_kernel kernel = clCreateKernel(vreman_nut_program,
63 "vreman_nut_compute_kernel", &err);
64 CL_CHECK(err);
65
66 CL_CHECK(clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *) &a11));
67 CL_CHECK(clSetKernelArg(kernel, 1, sizeof(cl_mem), (void *) &a12));
68 CL_CHECK(clSetKernelArg(kernel, 2, sizeof(cl_mem), (void *) &a13));
69 CL_CHECK(clSetKernelArg(kernel, 3, sizeof(cl_mem), (void *) &a21));
70 CL_CHECK(clSetKernelArg(kernel, 4, sizeof(cl_mem), (void *) &a22));
71 CL_CHECK(clSetKernelArg(kernel, 5, sizeof(cl_mem), (void *) &a23));
72 CL_CHECK(clSetKernelArg(kernel, 6, sizeof(cl_mem), (void *) &a31));
73 CL_CHECK(clSetKernelArg(kernel, 7, sizeof(cl_mem), (void *) &a32));
74 CL_CHECK(clSetKernelArg(kernel, 8, sizeof(cl_mem), (void *) &a33));
75 CL_CHECK(clSetKernelArg(kernel, 9, sizeof(cl_mem), (void *) &delta));
76 CL_CHECK(clSetKernelArg(kernel, 10, sizeof(cl_mem), (void *) &nut));
77 CL_CHECK(clSetKernelArg(kernel, 11, sizeof(cl_mem), (void *) &mult));
78 CL_CHECK(clSetKernelArg(kernel, 12, sizeof(real), c));
79 CL_CHECK(clSetKernelArg(kernel, 13, sizeof(real), eps));
80 CL_CHECK(clSetKernelArg(kernel, 14, sizeof(int), n));
81
82 if (*n > 0) {
83 const int nb = ((*n) + 256 - 1) / 256;
84 const size_t global_item_size = 256 * nb;
85 const size_t local_item_size = 256;
86
87 CL_CHECK(clEnqueueNDRangeKernel((cl_command_queue) glb_cmd_queue, kernel,
88 1, NULL, &global_item_size,
89 &local_item_size, 0, NULL, NULL));
90 }
91
92 CL_CHECK(clReleaseKernel(kernel));
93}
94
100void opencl_vreman_nut_compute_buoy(void *a11, void *a12, void *a13,
101 void *a21, void *a22, void *a23,
102 void *a31, void *a32, void *a33,
103 void *delta, void *nut, void *mult,
104 real *c, real *eps, int *n,
105 void *dTdx, void *dTdy, void *dTdz,
106 real *g, real *ri_c, real *ref_temp) {
107 cl_int err;
108
109 if (vreman_nut_program == NULL) {
110 opencl_kernel_jit(vreman_nut_kernel, (cl_program *) &vreman_nut_program);
111 }
112
113 cl_kernel kernel = clCreateKernel(vreman_nut_program,
114 "vreman_nut_compute_buoy_kernel", &err);
115 CL_CHECK(err);
116
117 const real g1 = g[0];
118 const real g2 = g[1];
119 const real g3 = g[2];
120 const real gmag = sqrt(g1 * g1 + g2 * g2 + g3 * g3);
121 const real n1 = g1 / gmag;
122 const real n2 = g2 / gmag;
123 const real n3 = g3 / gmag;
124
125 CL_CHECK(clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *) &a11));
126 CL_CHECK(clSetKernelArg(kernel, 1, sizeof(cl_mem), (void *) &a12));
127 CL_CHECK(clSetKernelArg(kernel, 2, sizeof(cl_mem), (void *) &a13));
128 CL_CHECK(clSetKernelArg(kernel, 3, sizeof(cl_mem), (void *) &a21));
129 CL_CHECK(clSetKernelArg(kernel, 4, sizeof(cl_mem), (void *) &a22));
130 CL_CHECK(clSetKernelArg(kernel, 5, sizeof(cl_mem), (void *) &a23));
131 CL_CHECK(clSetKernelArg(kernel, 6, sizeof(cl_mem), (void *) &a31));
132 CL_CHECK(clSetKernelArg(kernel, 7, sizeof(cl_mem), (void *) &a32));
133 CL_CHECK(clSetKernelArg(kernel, 8, sizeof(cl_mem), (void *) &a33));
134 CL_CHECK(clSetKernelArg(kernel, 9, sizeof(cl_mem), (void *) &delta));
135 CL_CHECK(clSetKernelArg(kernel, 10, sizeof(cl_mem), (void *) &nut));
136 CL_CHECK(clSetKernelArg(kernel, 11, sizeof(cl_mem), (void *) &mult));
137 CL_CHECK(clSetKernelArg(kernel, 12, sizeof(real), c));
138 CL_CHECK(clSetKernelArg(kernel, 13, sizeof(real), eps));
139 CL_CHECK(clSetKernelArg(kernel, 14, sizeof(int), n));
140 CL_CHECK(clSetKernelArg(kernel, 15, sizeof(cl_mem), (void *) &dTdx));
141 CL_CHECK(clSetKernelArg(kernel, 16, sizeof(cl_mem), (void *) &dTdy));
142 CL_CHECK(clSetKernelArg(kernel, 17, sizeof(cl_mem), (void *) &dTdz));
143 CL_CHECK(clSetKernelArg(kernel, 18, sizeof(real), (void *) &n1));
144 CL_CHECK(clSetKernelArg(kernel, 19, sizeof(real), (void *) &n2));
145 CL_CHECK(clSetKernelArg(kernel, 20, sizeof(real), (void *) &n3));
146 CL_CHECK(clSetKernelArg(kernel, 21, sizeof(real), (void *) &g1));
147 CL_CHECK(clSetKernelArg(kernel, 22, sizeof(real), (void *) &g2));
148 CL_CHECK(clSetKernelArg(kernel, 23, sizeof(real), (void *) &g3));
149 CL_CHECK(clSetKernelArg(kernel, 24, sizeof(real), ri_c));
150 CL_CHECK(clSetKernelArg(kernel, 25, sizeof(real), ref_temp));
151
152 if (*n > 0) {
153 const int nb = ((*n) + 256 - 1) / 256;
154 const size_t global_item_size = 256 * nb;
155 const size_t local_item_size = 256;
156
157 CL_CHECK(clEnqueueNDRangeKernel((cl_command_queue) glb_cmd_queue, kernel,
158 1, NULL, &global_item_size,
159 &local_item_size, 0, NULL, NULL));
160 }
161
162 CL_CHECK(clReleaseKernel(kernel));
163}
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
device_config.h
real
double real
Definition device_config.h:12
opencl_kernel_jit
void opencl_kernel_jit(const char *kernel, cl_program *program)
Definition jit.c:50
CL_CHECK
#define CL_CHECK(err)
Definition check.h:12
prgm_lib.h
vreman_nut_program
void * vreman_nut_program
opencl_vreman_nut_compute
void opencl_vreman_nut_compute(void *a11, void *a12, void *a13, void *a21, void *a22, void *a23, void *a31, void *a32, void *a33, void *delta, void *nut, void *mult, real *c, real *eps, int *n)
Definition vreman_nut.c:51
opencl_vreman_nut_compute_buoy
void opencl_vreman_nut_compute_buoy(void *a11, void *a12, void *a13, void *a21, void *a22, void *a23, void *a31, void *a32, void *a33, void *delta, void *nut, void *mult, real *c, real *eps, int *n, void *dTdx, void *dTdy, void *dTdz, real *g, real *ri_c, real *ref_temp)
Definition vreman_nut.c:100