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_path.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_path.h')
-rw-r--r-- | intern/cycles/kernel/kernel_path.h | 55 |
1 files changed, 15 insertions, 40 deletions
diff --git a/intern/cycles/kernel/kernel_path.h b/intern/cycles/kernel/kernel_path.h index 0ba8f694592..e2dbf6e22f7 100644 --- a/intern/cycles/kernel/kernel_path.h +++ b/intern/cycles/kernel/kernel_path.h @@ -42,6 +42,7 @@ #include "kernel_path_state.h" #include "kernel_shadow.h" #include "kernel_emission.h" +#include "kernel_path_common.h" #include "kernel_path_surface.h" #include "kernel_path_volume.h" @@ -305,17 +306,17 @@ ccl_device void kernel_path_ao(KernelGlobals *kg, ShaderData *sd, PathRadiance * sample_cos_hemisphere(ao_N, bsdf_u, bsdf_v, &ao_D, &ao_pdf); - if(dot(sd->Ng, ao_D) > 0.0f && ao_pdf != 0.0f) { + if(dot(ccl_fetch(sd, Ng), ao_D) > 0.0f && ao_pdf != 0.0f) { Ray light_ray; float3 ao_shadow; - light_ray.P = ray_offset(sd->P, sd->Ng); + light_ray.P = ray_offset(ccl_fetch(sd, P), ccl_fetch(sd, Ng)); light_ray.D = ao_D; light_ray.t = kernel_data.background.ao_distance; #ifdef __OBJECT_MOTION__ - light_ray.time = sd->time; + light_ray.time = ccl_fetch(sd, time); #endif - light_ray.dP = sd->dP; + light_ray.dP = ccl_fetch(sd, dP); light_ray.dD = differential3_zero(); if(!shadow_blocked(kg, state, &light_ray, &ao_shadow)) @@ -341,17 +342,17 @@ ccl_device void kernel_branched_path_ao(KernelGlobals *kg, ShaderData *sd, PathR sample_cos_hemisphere(ao_N, bsdf_u, bsdf_v, &ao_D, &ao_pdf); - if(dot(sd->Ng, ao_D) > 0.0f && ao_pdf != 0.0f) { + if(dot(ccl_fetch(sd, Ng), ao_D) > 0.0f && ao_pdf != 0.0f) { Ray light_ray; float3 ao_shadow; - light_ray.P = ray_offset(sd->P, sd->Ng); + light_ray.P = ray_offset(ccl_fetch(sd, P), ccl_fetch(sd, Ng)); light_ray.D = ao_D; light_ray.t = kernel_data.background.ao_distance; #ifdef __OBJECT_MOTION__ - light_ray.time = sd->time; + light_ray.time = ccl_fetch(sd, time); #endif - light_ray.dP = sd->dP; + light_ray.dP = ccl_fetch(sd, dP); light_ray.dD = differential3_zero(); if(!shadow_blocked(kg, state, &light_ray, &ao_shadow)) @@ -381,7 +382,7 @@ ccl_device bool kernel_path_subsurface_scatter(KernelGlobals *kg, ShaderData *sd #ifdef __VOLUME__ Ray volume_ray = *ray; bool need_update_volume_stack = kernel_data.integrator.use_volumes && - sd->flag & SD_OBJECT_INTERSECTS_VOLUME; + ccl_fetch(sd, flag) & SD_OBJECT_INTERSECTS_VOLUME; #endif /* compute lighting with the BSDF closure */ @@ -712,8 +713,8 @@ ccl_device_noinline void kernel_branched_path_surface_indirect_light(KernelGloba RNG *rng, ShaderData *sd, float3 throughput, float num_samples_adjust, PathState *state, PathRadiance *L) { - for(int i = 0; i< sd->num_closure; i++) { - const ShaderClosure *sc = &sd->closure[i]; + for(int i = 0; i< ccl_fetch(sd, num_closure); i++) { + const ShaderClosure *sc = &ccl_fetch(sd, closure)[i]; if(!CLOSURE_IS_BSDF(sc->type)) continue; @@ -764,8 +765,8 @@ ccl_device void kernel_branched_path_subsurface_scatter(KernelGlobals *kg, Ray *ray, float3 throughput) { - for(int i = 0; i< sd->num_closure; i++) { - ShaderClosure *sc = &sd->closure[i]; + for(int i = 0; i< ccl_fetch(sd, num_closure); i++) { + ShaderClosure *sc = &ccl_fetch(sd, closure)[i]; if(!CLOSURE_IS_BSSRDF(sc->type)) continue; @@ -786,7 +787,7 @@ ccl_device void kernel_branched_path_subsurface_scatter(KernelGlobals *kg, #ifdef __VOLUME__ Ray volume_ray = *ray; bool need_update_volume_stack = kernel_data.integrator.use_volumes && - sd->flag & SD_OBJECT_INTERSECTS_VOLUME; + ccl_fetch(sd, flag) & SD_OBJECT_INTERSECTS_VOLUME; #endif /* compute lighting with the BSDF closure */ @@ -1143,32 +1144,6 @@ ccl_device float4 kernel_branched_path_integrate(KernelGlobals *kg, RNG *rng, in #endif -ccl_device_inline void kernel_path_trace_setup(KernelGlobals *kg, ccl_global uint *rng_state, int sample, int x, int y, RNG *rng, Ray *ray) -{ - float filter_u; - float filter_v; - - int num_samples = kernel_data.integrator.aa_samples; - - path_rng_init(kg, rng_state, sample, num_samples, rng, x, y, &filter_u, &filter_v); - - /* sample camera ray */ - - float lens_u = 0.0f, lens_v = 0.0f; - - if(kernel_data.cam.aperturesize > 0.0f) - path_rng_2D(kg, rng, sample, num_samples, PRNG_LENS_U, &lens_u, &lens_v); - - float time = 0.0f; - -#ifdef __CAMERA_MOTION__ - if(kernel_data.cam.shuttertime != -1.0f) - time = path_rng_1D(kg, rng, sample, num_samples, PRNG_TIME); -#endif - - camera_sample(kg, x, y, filter_u, filter_v, lens_u, lens_v, time, ray); -} - ccl_device void kernel_path_trace(KernelGlobals *kg, ccl_global float *buffer, ccl_global uint *rng_state, int sample, int x, int y, int offset, int stride) |