diff options
author | George Kyriazis <George.Kyriazis@amd.com> | 2015-05-09 17:34:30 +0300 |
---|---|---|
committer | Sergey Sharybin <sergey.vfx@gmail.com> | 2015-05-09 17:52:40 +0300 |
commit | 7f4479da425b2d44a585f1b7b63f91d9dfecef02 (patch) | |
tree | 96ae5e7d4e091f89beedcd37609b3769783a00af /intern/cycles/kernel/kernel_differential.h | |
parent | f680c1b54a28a02fb86271bca649da0660542e9a (diff) |
Cycles: OpenCL kernel split
This commit contains all the work related on the AMD megakernel split work
which was mainly done by Varun Sundar, George Kyriazis and Lenny Wang, plus
some help from Sergey Sharybin, Martijn Berger, Thomas Dinges and likely
someone else which we're forgetting to mention.
Currently only AMD cards are enabled for the new split kernel, but it is
possible to force split opencl kernel to be used by setting the following
environment variable: CYCLES_OPENCL_SPLIT_KERNEL_TEST=1.
Not all the features are supported yet, and that being said no motion blur,
camera blur, SSS and volumetrics for now. Also transparent shadows are
disabled on AMD device because of some compiler bug.
This kernel is also only implements regular path tracing and supporting
branched one will take a bit. Branched path tracing is exposed to the
interface still, which is a bit misleading and will be hidden there soon.
More feature will be enabled once they're ported to the split kernel and
tested.
Neither regular CPU nor CUDA has any difference, they're generating the
same exact code, which means no regressions/improvements there.
Based on the research paper:
https://research.nvidia.com/sites/default/files/publications/laine2013hpg_paper.pdf
Here's the documentation:
https://docs.google.com/document/d/1LuXW-CV-sVJkQaEGZlMJ86jZ8FmoPfecaMdR-oiWbUY/edit
Design discussion of the patch:
https://developer.blender.org/T44197
Differential Revision: https://developer.blender.org/D1200
Diffstat (limited to 'intern/cycles/kernel/kernel_differential.h')
-rw-r--r-- | intern/cycles/kernel/kernel_differential.h | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/intern/cycles/kernel/kernel_differential.h b/intern/cycles/kernel/kernel_differential.h index e5fbd5b450e..ae1e70f0167 100644 --- a/intern/cycles/kernel/kernel_differential.h +++ b/intern/cycles/kernel/kernel_differential.h @@ -18,7 +18,7 @@ CCL_NAMESPACE_BEGIN /* See "Tracing Ray Differentials", Homan Igehy, 1999. */ -ccl_device void differential_transfer(differential3 *dP_, const differential3 dP, float3 D, const differential3 dD, float3 Ng, float t) +ccl_device void differential_transfer(ccl_addr_space differential3 *dP_, const differential3 dP, float3 D, const differential3 dD, float3 Ng, float t) { /* ray differential transfer through homogeneous medium, to * compute dPdx/dy at a shading point from the incoming ray */ @@ -31,7 +31,7 @@ ccl_device void differential_transfer(differential3 *dP_, const differential3 dP dP_->dy = tmpy - dot(tmpy, Ng)*tmp; } -ccl_device void differential_incoming(differential3 *dI, const differential3 dD) +ccl_device void differential_incoming(ccl_addr_space differential3 *dI, const differential3 dD) { /* compute dIdx/dy at a shading point, we just need to negate the * differential of the ray direction */ @@ -40,7 +40,7 @@ ccl_device void differential_incoming(differential3 *dI, const differential3 dD) dI->dy = -dD.dy; } -ccl_device void differential_dudv(differential *du, differential *dv, float3 dPdu, float3 dPdv, differential3 dP, float3 Ng) +ccl_device void differential_dudv(ccl_addr_space differential *du, ccl_addr_space differential *dv, float3 dPdu, float3 dPdv, differential3 dP, float3 Ng) { /* now we have dPdx/dy from the ray differential transfer, and dPdu/dv * from the primitive, we can compute dudx/dy and dvdx/dy. these are |