mirror of
https://github.com/firestar5683/StarPilot.git
synced 2026-07-06 22:22:11 +08:00
Dean Lee
9be81a6f99
* cleaup clutil.h * clutil.c -> clutil.cc * std::hash * try simplify functions * using namespace group help functions * continue * add function get_platform_info * continue * remove caching * struct CLContext * std::cout & CL_ERR_TO_STR * remove CLU_NO_CACHE * Revert "struct CLContext" This reverts commit 882e413f227e4c1ba4bb83f956cc67739a189221. * cl_get_error_string : remove undefied error codes * replace all printf with std::cout * util::read_file * cleanup * revert CL_ERR_TO_STR * simplify * add macro CL_ERR_TO_STR * cl_get_error_string: fix undeclared errors * log_size should not +1 * use std::stirng log * apply review old-commit-hash: 3c08d0f1f7e7c9f376fd69fd46067c6f6ff59d39
62 lines
2.2 KiB
C
62 lines
2.2 KiB
C
#include <string.h>
|
|
#include <assert.h>
|
|
|
|
#include "clutil.h"
|
|
|
|
#include "loadyuv.h"
|
|
|
|
void loadyuv_init(LoadYUVState* s, cl_context ctx, cl_device_id device_id, int width, int height) {
|
|
memset(s, 0, sizeof(*s));
|
|
|
|
s->width = width;
|
|
s->height = height;
|
|
|
|
char args[1024];
|
|
snprintf(args, sizeof(args),
|
|
"-cl-fast-relaxed-math -cl-denorms-are-zero "
|
|
"-DTRANSFORMED_WIDTH=%d -DTRANSFORMED_HEIGHT=%d",
|
|
width, height);
|
|
cl_program prg = cl_program_from_file(ctx, device_id, "transforms/loadyuv.cl", args);
|
|
|
|
s->loadys_krnl = CL_CHECK_ERR(clCreateKernel(prg, "loadys", &err));
|
|
s->loaduv_krnl = CL_CHECK_ERR(clCreateKernel(prg, "loaduv", &err));
|
|
|
|
// done with this
|
|
CL_CHECK(clReleaseProgram(prg));
|
|
}
|
|
|
|
void loadyuv_destroy(LoadYUVState* s) {
|
|
CL_CHECK(clReleaseKernel(s->loadys_krnl));
|
|
CL_CHECK(clReleaseKernel(s->loaduv_krnl));
|
|
}
|
|
|
|
void loadyuv_queue(LoadYUVState* s, cl_command_queue q,
|
|
cl_mem y_cl, cl_mem u_cl, cl_mem v_cl,
|
|
cl_mem out_cl) {
|
|
CL_CHECK(clSetKernelArg(s->loadys_krnl, 0, sizeof(cl_mem), &y_cl));
|
|
CL_CHECK(clSetKernelArg(s->loadys_krnl, 1, sizeof(cl_mem), &out_cl));
|
|
|
|
const size_t loadys_work_size = (s->width*s->height)/8;
|
|
CL_CHECK(clEnqueueNDRangeKernel(q, s->loadys_krnl, 1, NULL,
|
|
&loadys_work_size, NULL, 0, 0, NULL));
|
|
|
|
const size_t loaduv_work_size = ((s->width/2)*(s->height/2))/8;
|
|
cl_int loaduv_out_off = (s->width*s->height);
|
|
|
|
CL_CHECK(clSetKernelArg(s->loaduv_krnl, 0, sizeof(cl_mem), &u_cl));
|
|
CL_CHECK(clSetKernelArg(s->loaduv_krnl, 1, sizeof(cl_mem), &out_cl));
|
|
CL_CHECK(clSetKernelArg(s->loaduv_krnl, 2, sizeof(cl_int), &loaduv_out_off));
|
|
|
|
CL_CHECK(clEnqueueNDRangeKernel(q, s->loaduv_krnl, 1, NULL,
|
|
&loaduv_work_size, NULL, 0, 0, NULL));
|
|
|
|
loaduv_out_off += (s->width/2)*(s->height/2);
|
|
|
|
CL_CHECK(clSetKernelArg(s->loaduv_krnl, 0, sizeof(cl_mem), &v_cl));
|
|
CL_CHECK(clSetKernelArg(s->loaduv_krnl, 1, sizeof(cl_mem), &out_cl));
|
|
CL_CHECK(clSetKernelArg(s->loaduv_krnl, 2, sizeof(cl_int), &loaduv_out_off));
|
|
|
|
CL_CHECK(clEnqueueNDRangeKernel(q, s->loaduv_krnl, 1, NULL,
|
|
&loaduv_work_size, NULL, 0, 0, NULL));
|
|
}
|