diff options
author | Mai Lavelle <mai.lavelle@gmail.com> | 2017-02-22 16:10:02 +0300 |
---|---|---|
committer | Mai Lavelle <mai.lavelle@gmail.com> | 2017-03-08 08:52:41 +0300 |
commit | 230c00d872b817b0c4de85647464e4a12197c6aa (patch) | |
tree | 3659069562c7fff395c54faa464eff57c20c9676 /intern/cycles/kernel/kernel_passes.h | |
parent | 520b53364c73c75c4ff400d639dad13630f0e6fc (diff) |
Cycles: OpenCL split kernel refactor
This does a few things at once:
- Refactors host side split kernel logic into a new device
agnostic class `DeviceSplitKernel`.
- Removes tile splitting, a new work pool implementation takes its place and
allows as many threads as will fit in memory regardless of tile size, which
can give performance gains.
- Refactors split state buffers into one buffer, as well as reduces the
number of arguments passed to kernels. Means there's less code to deal
with overall.
- Moves kernel logic out of OpenCL kernel files so they can later be used by
other device types.
- Replaced OpenCL specific APIs with new generic versions
- Tiles can now be seen updating during rendering
Diffstat (limited to 'intern/cycles/kernel/kernel_passes.h')
-rw-r--r-- | intern/cycles/kernel/kernel_passes.h | 12 |
1 files changed, 6 insertions, 6 deletions
diff --git a/intern/cycles/kernel/kernel_passes.h b/intern/cycles/kernel/kernel_passes.h index 7aec47e4957..7790cce067b 100644 --- a/intern/cycles/kernel/kernel_passes.h +++ b/intern/cycles/kernel/kernel_passes.h @@ -19,16 +19,16 @@ CCL_NAMESPACE_BEGIN ccl_device_inline void kernel_write_pass_float(ccl_global float *buffer, int sample, float value) { ccl_global float *buf = buffer; -#if defined(__SPLIT_KERNEL__) && defined(__WORK_STEALING__) +#if defined(__SPLIT_KERNEL__) atomic_add_and_fetch_float(buf, value); #else *buf = (sample == 0)? value: *buf + value; -#endif // __SPLIT_KERNEL__ && __WORK_STEALING__ +#endif /* __SPLIT_KERNEL__ */ } ccl_device_inline void kernel_write_pass_float3(ccl_global float *buffer, int sample, float3 value) { -#if defined(__SPLIT_KERNEL__) && defined(__WORK_STEALING__) +#if defined(__SPLIT_KERNEL__) ccl_global float *buf_x = buffer + 0; ccl_global float *buf_y = buffer + 1; ccl_global float *buf_z = buffer + 2; @@ -39,12 +39,12 @@ ccl_device_inline void kernel_write_pass_float3(ccl_global float *buffer, int sa #else ccl_global float3 *buf = (ccl_global float3*)buffer; *buf = (sample == 0)? value: *buf + value; -#endif // __SPLIT_KERNEL__ && __WORK_STEALING__ +#endif /* __SPLIT_KERNEL__ */ } ccl_device_inline void kernel_write_pass_float4(ccl_global float *buffer, int sample, float4 value) { -#if defined(__SPLIT_KERNEL__) && defined(__WORK_STEALING__) +#if defined(__SPLIT_KERNEL__) ccl_global float *buf_x = buffer + 0; ccl_global float *buf_y = buffer + 1; ccl_global float *buf_z = buffer + 2; @@ -57,7 +57,7 @@ ccl_device_inline void kernel_write_pass_float4(ccl_global float *buffer, int sa #else ccl_global float4 *buf = (ccl_global float4*)buffer; *buf = (sample == 0)? value: *buf + value; -#endif // __SPLIT_KERNEL__ && __WORK_STEALING__ +#endif /* __SPLIT_KERNEL__ */ } ccl_device_inline void kernel_write_data_passes(KernelGlobals *kg, ccl_global float *buffer, PathRadiance *L, |