Neko 0.9.1
A portable framework for high-order spectral element flow simulations
Loading...
Searching...
No Matches
opr_cdtp.c
Go to the documentation of this file.
1/*
2 Copyright (c) 2021-2024, 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 <stdio.h>
42#include <device/device_config.h>
43#include <device/opencl/jit.h>
44#include <device/opencl/prgm_lib.h>
45#include <device/opencl/check.h>
46
47#include "cdtp_kernel.cl.h"
48
52void opencl_cdtp(void *dtx, void *x,
53 void *dr, void *ds, void *dt,
54 void *dxt, void *dyt, void *dzt,
55 void *w3, int *nel, int *lx) {
56 cl_int err;
57
58 if (cdtp_program == NULL)
59 opencl_kernel_jit(cdtp_kernel, (cl_program *) &cdtp_program);
60
61 const size_t global_item_size = 256 * (*nel);
62 const size_t local_item_size = 256;
63
64#define STR(X) #X
65#define CASE(LX) \
66 case LX: \
67 { \
68 cl_kernel kernel = clCreateKernel(cdtp_program, \
69 STR(cdtp_kernel_lx##LX), &err); \
70 CL_CHECK(err); \
71 \
72 CL_CHECK(clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *) &dtx)); \
73 CL_CHECK(clSetKernelArg(kernel, 1, sizeof(cl_mem), (void *) &x)); \
74 CL_CHECK(clSetKernelArg(kernel, 2, sizeof(cl_mem), (void *) &dr)); \
75 CL_CHECK(clSetKernelArg(kernel, 3, sizeof(cl_mem), (void *) &ds)); \
76 CL_CHECK(clSetKernelArg(kernel, 4, sizeof(cl_mem), (void *) &dt)); \
77 CL_CHECK(clSetKernelArg(kernel, 5, sizeof(cl_mem), (void *) &dxt)); \
78 CL_CHECK(clSetKernelArg(kernel, 6, sizeof(cl_mem), (void *) &dyt)); \
79 CL_CHECK(clSetKernelArg(kernel, 7, sizeof(cl_mem), (void *) &dzt)); \
80 CL_CHECK(clSetKernelArg(kernel, 8, sizeof(cl_mem), (void *) &w3)); \
81 \
82 CL_CHECK(clEnqueueNDRangeKernel((cl_command_queue) glb_cmd_queue, \
83 kernel, 1, NULL, &global_item_size, \
84 &local_item_size, 0, NULL, NULL)); \
85 } \
86 break
87
88 switch(*lx) {
89 CASE(2);
90 CASE(3);
91 CASE(4);
92 CASE(5);
93 CASE(6);
94 CASE(7);
95 CASE(8);
96 CASE(9);
97 CASE(10);
98 CASE(11);
99 CASE(12);
100 }
101}
102
dr
__global__ void const T *__restrict__ const T *__restrict__ dr
Definition cdtp_kernel.h:107
ds
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ ds
Definition cdtp_kernel.h:108
x
__global__ void const T *__restrict__ x
Definition cdtp_kernel.h:106
dt
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dt
Definition cdtp_kernel.h:109
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:113
dzt
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dzt
Definition cdtp_kernel.h:112
dyt
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dyt
Definition cdtp_kernel.h:111
dxt
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dxt
Definition cdtp_kernel.h:110
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
opencl_kernel_jit
void opencl_kernel_jit(const char *kernel, cl_program *program)
Definition jit.c:50
opencl_cdtp
void opencl_cdtp(void *dtx, void *x, void *dr, void *ds, void *dt, void *dxt, void *dyt, void *dzt, void *w3, int *nel, int *lx)
Definition opr_cdtp.c:52
CASE
#define CASE(LX)
prgm_lib.h
cdtp_program
void * cdtp_program