cl_int cl_convert_rgbf_to_yuvf(cl_environment_t *pEnv,
float *r,
float *g,
float *b,
float *y,
float *u,
float *v,
float *a, // constants
cl_uint numPixels)
{
cl_int err = CL_SUCCESS;
const size_t numBytes = numPixels * sizeof(float);
cl_kernel_param_t params[] = {
{CL_KPARAM_BUFFER_1D, numBytes, r, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, numBytes, g, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, numBytes, b, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, sizeof(float) * 9, a, NULL, CL_MEM_READ_ONLY}, // constants only have 9 elements
{CL_KPARAM_BUFFER_1D, numBytes, y, NULL, CL_MEM_WRITE_ONLY},
{CL_KPARAM_BUFFER_1D, numBytes, u, NULL, CL_MEM_WRITE_ONLY},
{CL_KPARAM_BUFFER_1D, numBytes, v, NULL, CL_MEM_WRITE_ONLY},
};
cl_kernel_call_t call = {
"kernel_rgb2yuv",
params, dimof(params),
1,
{0,0,0},
{numPixels, 0, 0},
{1,1,1},
CL_SUCCESS,0,0,0
};
err = clCallKernel(pEnv, &call, 1);
return err;
}
cl_int nbodies(cl_environment_t *pEnv,
cl_float *m,
cl_float4 *a,
cl_float4 *v,
cl_float4 *p,
cl_float *t,
cl_uint numBodies)
{
cl_uint n = sizeof(cl_float)*numBodies;
cl_uint n4 = sizeof(cl_float4)*numBodies;
cl_kernel_param_t params[] = {
{CL_KPARAM_BUFFER_1D, n, m, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, n4, a, NULL, CL_MEM_READ_WRITE},
{CL_KPARAM_BUFFER_1D, n4, v, NULL, CL_MEM_READ_WRITE},
{CL_KPARAM_BUFFER_1D, n4, p, NULL, CL_MEM_READ_WRITE},
{CL_KPARAM_BUFFER_1D, n, t, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(numBodies), &numBodies, NULL, CL_MEM_READ_ONLY},
};
cl_kernel_call_t call = {
"kernel_nbody",
params, dimof(params),
1,
{0,0,0},
{numBodies, 0, 0},
{1,1,1},
CL_SUCCESS,0,0,0
};
return clCallKernel(pEnv, &call,1);
}
cl_int imgfilter1d_opt(cl_environment_t *pEnv,
cl_uint width,
cl_uint height,
cl_uchar *pSrc,
cl_int srcStride,
cl_uchar *pDst,
cl_int dstStride,
cl_char *op,
cl_uint opDim,
cl_uint range,
cl_uint limit)
{
cl_int err = CL_SUCCESS;
cl_uint numSrcBytes = srcStride * height;
cl_uint numDstBytes = dstStride * height;
cl_uint numOpBytes = 2 * opDim * opDim;
cl_kernel_param_t params[] = {
{CL_KPARAM_BUFFER_0D, sizeof(cl_uint), &width, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_uint), &height, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, numSrcBytes, pSrc, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_int), &srcStride, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, numDstBytes, pDst, NULL, CL_MEM_WRITE_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_int), &dstStride, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, numOpBytes, op, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_uint), &opDim, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_uint), &range, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_uint), &limit, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_LOCAL, sizeof(cl_uchar)*18, NULL, NULL, CL_MEM_READ_WRITE},
{CL_KPARAM_LOCAL, sizeof(cl_char)*18, NULL, NULL, CL_MEM_READ_WRITE},
};
cl_kernel_call_t call = {
"kernel_edge_filter_opt",
params, dimof(params),
2,
{0,0,0},
{width, height, 0},
{1,1,1},
CL_SUCCESS, 0,0,0
};
err = clCallKernel(pEnv, &call,1);
return err;
}
cl_int distance(cl_environment_t *pEnv,
cl_float4 *p,
cl_float *d,
cl_uint numPoints)
{
cl_uint n = sizeof(float) * numPoints;
cl_uint n4 = sizeof(cl_float4) * numPoints;
cl_kernel_param_t params[] = {
{CL_KPARAM_BUFFER_1D, n4,p,NULL,CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, n ,d,NULL,CL_MEM_WRITE_ONLY},
};
cl_kernel_call_t call = {
"kernel_magnitude",
params, dimof(params),
1,
{0,0,0},
{numPoints, 0, 0},
{1,1,1},
CL_SUCCESS, 0,0,0
};
return clCallKernel(pEnv, &call,1);
}
cl_int convert_uyvy_to_2bgr(cl_environment_t *pEnv, cl_uchar *pUYVY, cl_uchar *pBGR, cl_uint width, cl_uint height, cl_int srcStride, cl_int dstStride)
{
cl_int err = CL_SUCCESS;
cl_kernel_param_t params[] = {
{CL_KPARAM_BUFFER_1D, srcStride*height, pUYVY, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, dstStride*height, pBGR, NULL, CL_MEM_WRITE_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_uint), &width, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_uint), &height, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_int), &srcStride, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_int), &dstStride, NULL, CL_MEM_READ_ONLY},
};
cl_kernel_call_t call = {
"kernel_uyvy_to_2bgr",
params, dimof(params),
2,
{0,0,0},
{width/2, height, 0}, // width/2 since it is macropixel packed!
{1,1,1},
CL_SUCCESS, 0,0,0
};
err = clCallKernel(pEnv, &call, 1);
return err;
}
cl_int convert_yuv420_to_luma(cl_environment_t *pEnv, cl_uchar *pYUV420, cl_uchar *pLuma, cl_uint width, cl_uint height, cl_int srcStride, cl_int dstStride)
{
cl_int err = CL_SUCCESS;
cl_kernel_param_t params[] = {
{CL_KPARAM_BUFFER_1D, srcStride*height, pYUV420, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, dstStride*height, pLuma, NULL, CL_MEM_WRITE_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_uint), &width, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_uint), &height, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_int), &srcStride, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(cl_int), &dstStride, NULL, CL_MEM_READ_ONLY},
};
cl_kernel_call_t call = {
"kernel_yuv420_to_luma",
params, dimof(params),
2,
{0,0,0},
{width, height, 0},
{1,1,1},
CL_SUCCESS, 0,0,0
};
err = clCallKernel(pEnv, &call, 1);
return err;
}
cl_int kernel_spacetime(cl_environment_t *pEnv,
cl_spacetime_grid_t *grid,
cl_body_t *body) {
cl_int err;
cl_double distance[GRID_SIZE];
cl_kernel_param_t params[] = {
{CL_KPARAM_BUFFER_0D, sizeof(precision4), &body->m_position, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_0D, sizeof(precision), &body->m_mass, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, sizeof(precision4)*GRID_SIZE, grid->m_position, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, sizeof(precision)*GRID_SIZE, grid->m_mass, NULL, CL_MEM_READ_ONLY},
{CL_KPARAM_BUFFER_1D, sizeof(precision4)*GRID_SIZE, grid->m_budge, NULL, CL_MEM_WRITE_ONLY},
};
cl_kernel_call_t calls[] = {
{
"kernel_distort",
params, dimof(params),
1, // num dimensions
{0, 0, 0}, // offsets
{GRID_SIZE, 0, 0}, // dimensionality
{1, 1, 1}, // local work size
CL_SUCCESS,
//0, {}, {}
}
};
err = clCallKernel(pEnv, calls, dimof(calls));
printf("Distance from grid[0] to body is %lf, pull is %lf\n", distance[0], spacetime->m_grid.m_pull[0]);
if (err != CL_SUCCESS)
return err;
else {
for (size_t e = 0; e < dimof(calls); e++)
if (calls[e].err != CL_SUCCESS)
return calls[e].err;
}
return err;
}
