diff options
author | Lukas Stockner <lukas.stockner@freenet.de> | 2018-04-01 03:10:27 +0300 |
---|---|---|
committer | Lukas Stockner <lukas.stockner@freenet.de> | 2018-05-24 19:44:56 +0300 |
commit | b4a8b813993138cc831d4c04f0f94bdafc51e7fe (patch) | |
tree | 72b5fcfe3cdfdf6c4a9db68d896414ef0d5cdb0c /intern/cycles/kernel/filter | |
parent | 68627626854c27c2135cab72b48b648cb638c8cb (diff) |
Cycles Denoising: Don't use atomics in the accumulation kernel on CPUs
The GPU kernel needs to use atomics for accumulation since all offsets are processed in
parallel, but on CPUs that's not the case, so we can disable them there for a considerable speedup.
Diffstat (limited to 'intern/cycles/kernel/filter')
-rw-r--r-- | intern/cycles/kernel/filter/filter_reconstruction.h | 5 |
1 files changed, 5 insertions, 0 deletions
diff --git a/intern/cycles/kernel/filter/filter_reconstruction.h b/intern/cycles/kernel/filter/filter_reconstruction.h index b7bf322f9ce..58740d5b06a 100644 --- a/intern/cycles/kernel/filter/filter_reconstruction.h +++ b/intern/cycles/kernel/filter/filter_reconstruction.h @@ -61,8 +61,13 @@ ccl_device_inline void kernel_filter_construct_gramian(int x, int y, make_int2(x+dx, y+dy), buffer + q_offset, pass_stride, *rank, design_row, transform, stride); +#ifdef __KERNEL_GPU__ math_trimatrix_add_gramian_strided(XtWX, (*rank)+1, design_row, weight, stride); math_vec3_add_strided(XtWY, (*rank)+1, design_row, weight * q_color, stride); +#else + math_trimatrix_add_gramian(XtWX, (*rank)+1, design_row, weight); + math_vec3_add(XtWY, (*rank)+1, design_row, weight * q_color); +#endif } ccl_device_inline void kernel_filter_finalize(int x, int y, |