diff options
| author | Brecht Van Lommel <brechtvanlommel@gmail.com> | 2017-08-06 18:17:40 +0300 |
|---|---|---|
| committer | Brecht Van Lommel <brechtvanlommel@gmail.com> | 2017-08-12 21:43:34 +0300 |
| commit | 8f971083531c8f4bf2e6b3892e615d0616e152db (patch) | |
| tree | c497b682830807ca544b1fd45f8b99ff67dec9c3 | |
| parent | 601f94a3c2024343c4c473c5cf079f6c02a2ab77 (diff) | |
Cycles: optimize CPU split kernel data init.
| -rw-r--r-- | intern/cycles/kernel/split/kernel_data_init.h | 20 |
1 files changed, 16 insertions, 4 deletions
diff --git a/intern/cycles/kernel/split/kernel_data_init.h b/intern/cycles/kernel/split/kernel_data_init.h index 6f3297de342..2c042dfde6f 100644 --- a/intern/cycles/kernel/split/kernel_data_init.h +++ b/intern/cycles/kernel/split/kernel_data_init.h @@ -124,14 +124,25 @@ void KERNEL_FUNCTION_FULL_NAME(data_init)( /* zero the tiles pixels and initialize rng_state if this is the first sample */ if(start_sample == 0) { - parallel_for(kg, i, sw * sh * kernel_data.film.pass_stride) { - int pixel = i / kernel_data.film.pass_stride; - int pass = i % kernel_data.film.pass_stride; + int pass_stride = kernel_data.film.pass_stride; + +#ifdef __KERNEL_CPU__ + for(int y = sy; y < sy + sh; y++) { + int index = offset + y * stride; + memset(buffer + (sx + index) * pass_stride, 0, sizeof(float) * pass_stride * sw); + for(int x = sx; x < sx + sw; x++) { + rng_state[index + x] = hash_int_2d(x, y); + } + } +#else + parallel_for(kg, i, sw * sh * pass_stride) { + int pixel = i / pass_stride; + int pass = i % pass_stride; int x = sx + pixel % sw; int y = sy + pixel / sw; - int index = (offset + x + y*stride) * kernel_data.film.pass_stride + pass; + int index = (offset + x + y*stride) * pass_stride + pass; *(buffer + index) = 0.0f; } @@ -143,6 +154,7 @@ void KERNEL_FUNCTION_FULL_NAME(data_init)( int index = (offset + x + y*stride); *(rng_state + index) = hash_int_2d(x, y); } +#endif } #endif /* KERENL_STUB */ |