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/split/kernel_lamp_emission.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/split/kernel_lamp_emission.h')
-rw-r--r-- | intern/cycles/kernel/split/kernel_lamp_emission.h | 58 |
1 files changed, 36 insertions, 22 deletions
diff --git a/intern/cycles/kernel/split/kernel_lamp_emission.h b/intern/cycles/kernel/split/kernel_lamp_emission.h index 3bd0e361078..a9820af3001 100644 --- a/intern/cycles/kernel/split/kernel_lamp_emission.h +++ b/intern/cycles/kernel/split/kernel_lamp_emission.h @@ -14,7 +14,7 @@ * limitations under the License. */ -#include "kernel_split_common.h" +CCL_NAMESPACE_BEGIN /* Note on kernel_lamp_emission * This is the 3rd kernel in the ray-tracing logic. This is the second of the @@ -36,28 +36,39 @@ * sw -------------------------------------------------| | * sh -------------------------------------------------| | */ -ccl_device void kernel_lamp_emission( - KernelGlobals *kg, - ccl_global float3 *throughput_coop, /* Required for lamp emission */ - PathRadiance *PathRadiance_coop, /* Required for lamp emission */ - ccl_global Ray *Ray_coop, /* Required for lamp emission */ - ccl_global PathState *PathState_coop, /* Required for lamp emission */ - Intersection *Intersection_coop, /* Required for lamp emission */ - ccl_global char *ray_state, /* Denotes the state of each ray */ - int sw, int sh, - ccl_global char *use_queues_flag, /* Used to decide if this kernel should use - * queues to fetch ray index - */ - int ray_index) +ccl_device void kernel_lamp_emission(KernelGlobals *kg) { - if(IS_STATE(ray_state, ray_index, RAY_ACTIVE) || - IS_STATE(ray_state, ray_index, RAY_HIT_BACKGROUND)) + /* We will empty this queue in this kernel. */ + if(ccl_global_id(0) == 0 && ccl_global_id(1) == 0) { + kernel_split_params.queue_index[QUEUE_ACTIVE_AND_REGENERATED_RAYS] = 0; + } + /* Fetch use_queues_flag. */ + ccl_local char local_use_queues_flag; + if(ccl_local_id(0) == 0 && ccl_local_id(1) == 0) { + local_use_queues_flag = *kernel_split_params.use_queues_flag; + } + ccl_barrier(CCL_LOCAL_MEM_FENCE); + + int ray_index = ccl_global_id(1) * ccl_global_size(0) + ccl_global_id(0); + if(local_use_queues_flag) { + ray_index = get_ray_index(kg, ray_index, + QUEUE_ACTIVE_AND_REGENERATED_RAYS, + kernel_split_state.queue_data, + kernel_split_params.queue_size, + 1); + if(ray_index == QUEUE_EMPTY_SLOT) { + return; + } + } + + if(IS_STATE(kernel_split_state.ray_state, ray_index, RAY_ACTIVE) || + IS_STATE(kernel_split_state.ray_state, ray_index, RAY_HIT_BACKGROUND)) { - PathRadiance *L = &PathRadiance_coop[ray_index]; - ccl_global PathState *state = &PathState_coop[ray_index]; + PathRadiance *L = &kernel_split_state.path_radiance[ray_index]; + ccl_global PathState *state = &kernel_split_state.path_state[ray_index]; - float3 throughput = throughput_coop[ray_index]; - Ray ray = Ray_coop[ray_index]; + float3 throughput = kernel_split_state.throughput[ray_index]; + Ray ray = kernel_split_state.ray[ray_index]; #ifdef __LAMP_MIS__ if(kernel_data.integrator.use_lamp_mis && !(state->flag & PATH_RAY_CAMERA)) { @@ -65,7 +76,7 @@ ccl_device void kernel_lamp_emission( Ray light_ray; light_ray.P = ray.P - state->ray_t*ray.D; - state->ray_t += Intersection_coop[ray_index].t; + state->ray_t += kernel_split_state.isect[ray_index].t; light_ray.D = ray.D; light_ray.t = state->ray_t; light_ray.time = ray.time; @@ -74,10 +85,13 @@ ccl_device void kernel_lamp_emission( /* intersect with lamp */ float3 emission; - if(indirect_lamp_emission(kg, kg->sd_input, state, &light_ray, &emission)) { + if(indirect_lamp_emission(kg, kernel_split_state.sd_DL_shadow, state, &light_ray, &emission)) { path_radiance_accum_emission(L, throughput, emission, state->bounce); } } #endif /* __LAMP_MIS__ */ } } + +CCL_NAMESPACE_END + |