aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

+added Makefile and some fixes

Browse Source
This commit is contained in:
Evghenii
2013-11-13 14:16:48 +01:00
parent dededd1929
commit c0c1cc1ba7
9 changed files with 596 additions and 117 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
examples_cuda/mandelbrot_tasks3d/mandelbrot_tasks3d.ispc
View File

@@ -31,12 +31,21 @@
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#ifdef __NVPTX__
#define taskIndex0 blockIndex0()
#define taskIndex1 blockIndex1()
#define taskCount0 blockCount0()
#define taskCount1 blockCount1()
#define programCount warpSize()
#define programIndex laneIndex()
#endif
static inline int
mandel(float c_re, float c_im, int count) {
float z_re = c_re, z_im = c_im;
int i;
for (i = 0; i < count; ++i) {
if (z_re * z_re + z_im * z_im > 4.)
if (z_re * z_re + z_im * z_im > 4.0f)
break;
float new_re = z_re*z_re - z_im*z_im;
@@ -65,13 +74,16 @@ mandelbrot_scanline(uniform float x0, uniform float dx,
const uniform int ystart = taskIndex1 * yspan;
const uniform int yend = min(ystart + yspan, height);
for (uniform int yi = ystart; yi < yend; yi++)
for (uniform int xi = xstart; xi < xend; xi += programCount)
{
const float x = x0 + (xi + programIndex) * dx;
const float y = y0 + yi * dy;
foreach (yi = ystart ... yend, xi = xstart ... xend) {
float x = x0 + xi * dx;
float y = y0 + yi * dy;
int index = yi * width + xi;
output[index] = mandel(x, y, maxIterations);
const int res = mandel(x,y,maxIterations);
const int index = yi * width + (xi + programIndex);
if (xi + programIndex < xend)
output[index] = res;
}
}
@@ -84,8 +96,8 @@ mandelbrot_ispc(uniform float x0, uniform float y0,
uniform int maxIterations, uniform int output[]) {
uniform float dx = (x1 - x0) / width;
uniform float dy = (y1 - y0) / height;
const uniform int xspan = 16; /* make sure it is big enough to avoid false-sharing */
const uniform int yspan = 16;
const uniform int xspan = 32; /* make sure it is big enough to avoid false-sharing */
const uniform int yspan = 4;
#if 1