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_shadow.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_shadow.h')
-rw-r--r-- | intern/cycles/kernel/kernel_shadow.h | 54 |
1 files changed, 38 insertions, 16 deletions
diff --git a/intern/cycles/kernel/kernel_shadow.h b/intern/cycles/kernel/kernel_shadow.h index d7c4fa02bcf..5119b7a30c2 100644 --- a/intern/cycles/kernel/kernel_shadow.h +++ b/intern/cycles/kernel/kernel_shadow.h @@ -180,19 +180,37 @@ ccl_device_inline bool shadow_blocked(KernelGlobals *kg, PathState *state, Ray * * potentially transparent, and only in that case start marching. this gives * one extra ray cast for the cases were we do want transparency. */ -ccl_device_inline bool shadow_blocked(KernelGlobals *kg, PathState *state, Ray *ray, float3 *shadow) +/* The arguments sd_mem and isect_mem are meaningful only for OpenCL split kernel. Other uses can just pass a NULL */ +ccl_device_inline bool shadow_blocked(KernelGlobals *kg, ccl_addr_space PathState *state, ccl_addr_space Ray *ray_input, float3 *shadow +#ifdef __SPLIT_KERNEL__ + , ShaderData *sd_mem, Intersection *isect_mem +#endif + ) { *shadow = make_float3(1.0f, 1.0f, 1.0f); - if(ray->t == 0.0f) + if(ray_input->t == 0.0f) return false; - Intersection isect; - bool blocked = scene_intersect(kg, ray, PATH_RAY_SHADOW_OPAQUE, &isect, NULL, 0.0f, 0.0f); +#ifdef __SPLIT_KERNEL__ + Ray private_ray = *ray_input; + Ray *ray = &private_ray; +#else + Ray *ray = ray_input; +#endif + +#ifdef __SPLIT_KERNEL__ + Intersection *isect = isect_mem; +#else + Intersection isect_object; + Intersection *isect = &isect_object; +#endif + + bool blocked = scene_intersect(kg, ray, PATH_RAY_SHADOW_OPAQUE, isect, NULL, 0.0f, 0.0f); #ifdef __TRANSPARENT_SHADOWS__ if(blocked && kernel_data.integrator.transparent_shadows) { - if(shader_transparent_shadow(kg, &isect)) { + if(shader_transparent_shadow(kg, isect)) { float3 throughput = make_float3(1.0f, 1.0f, 1.0f); float3 Pend = ray->P + ray->D*ray->t; int bounce = state->transparent_bounce; @@ -204,9 +222,8 @@ ccl_device_inline bool shadow_blocked(KernelGlobals *kg, PathState *state, Ray * if(bounce >= kernel_data.integrator.transparent_max_bounce) return true; - if(!scene_intersect(kg, ray, PATH_RAY_SHADOW_TRANSPARENT, &isect, NULL, 0.0f, 0.0f)) + if(!scene_intersect(kg, ray, PATH_RAY_SHADOW_TRANSPARENT, isect, NULL, 0.0f, 0.0f)) { - #ifdef __VOLUME__ /* attenuation for last line segment towards light */ if(ps.volume_stack[0].shader != SHADER_NONE) @@ -218,39 +235,44 @@ ccl_device_inline bool shadow_blocked(KernelGlobals *kg, PathState *state, Ray * return false; } - if(!shader_transparent_shadow(kg, &isect)) + if(!shader_transparent_shadow(kg, isect)) return true; #ifdef __VOLUME__ /* attenuation between last surface and next surface */ if(ps.volume_stack[0].shader != SHADER_NONE) { Ray segment_ray = *ray; - segment_ray.t = isect.t; + segment_ray.t = isect->t; kernel_volume_shadow(kg, &ps, &segment_ray, &throughput); } #endif /* setup shader data at surface */ - ShaderData sd; - shader_setup_from_ray(kg, &sd, &isect, ray, state->bounce+1, bounce); +#ifdef __SPLIT_KERNEL__ + ShaderData *sd = sd_mem; +#else + ShaderData sd_object; + ShaderData *sd = &sd_object; +#endif + shader_setup_from_ray(kg, sd, isect, ray, state->bounce+1, bounce); /* attenuation from transparent surface */ - if(!(sd.flag & SD_HAS_ONLY_VOLUME)) { - shader_eval_surface(kg, &sd, 0.0f, PATH_RAY_SHADOW, SHADER_CONTEXT_SHADOW); - throughput *= shader_bsdf_transparency(kg, &sd); + if(!(ccl_fetch(sd, flag) & SD_HAS_ONLY_VOLUME)) { + shader_eval_surface(kg, sd, 0.0f, PATH_RAY_SHADOW, SHADER_CONTEXT_SHADOW); + throughput *= shader_bsdf_transparency(kg, sd); } if(is_zero(throughput)) return true; /* move ray forward */ - ray->P = ray_offset(sd.P, -sd.Ng); + ray->P = ray_offset(ccl_fetch(sd, P), -ccl_fetch(sd, Ng)); if(ray->t != FLT_MAX) ray->D = normalize_len(Pend - ray->P, &ray->t); #ifdef __VOLUME__ /* exit/enter volume */ - kernel_volume_stack_enter_exit(kg, &sd, ps.volume_stack); + kernel_volume_stack_enter_exit(kg, sd, ps.volume_stack); #endif bounce++; |