mirror of
https://github.com/blender/blender
synced 2026-10-25 03:07:01 +03:00
5fc85d1cb7
The new intersect_mnee kernel runs before shade_surface, and shade_surface_mnee is eliminated. That large kernel was causing problems for some GPU compilers. MNEE state is packed into a shadow path state to avoid significantly increasing the path state size. This shadow state is then either turned into an actual shadow ray state or discarded in shade_surface. MNEE was re-enabled on HIP RDNA2 as it works again now. Texture cache misses now also work correctly with MNEE. This adds some extra code to the regular shade_surface kernel even when MNEE is not used, to use the MNEE sampled point instead of sampling a light. But there seems to be no significant performance impact. Co-authored-by: Sergey Sharybin <sergey@blender.org> Pull Request: https://projects.blender.org/blender/blender/pulls/158698
121 lines
4.5 KiB
C
121 lines
4.5 KiB
C
/* SPDX-FileCopyrightText: 2011-2022 Blender Foundation
|
|
*
|
|
* SPDX-License-Identifier: Apache-2.0 */
|
|
|
|
#pragma once
|
|
|
|
#include "kernel/integrator/intersect_closest.h"
|
|
#include "kernel/integrator/intersect_dedicated_light.h"
|
|
#include "kernel/integrator/intersect_mnee.h"
|
|
#include "kernel/integrator/intersect_shadow.h"
|
|
#include "kernel/integrator/intersect_subsurface.h"
|
|
#include "kernel/integrator/intersect_volume_stack.h"
|
|
#include "kernel/integrator/shade_background.h"
|
|
#include "kernel/integrator/shade_dedicated_light.h"
|
|
#include "kernel/integrator/shade_light.h"
|
|
#include "kernel/integrator/shade_shadow.h"
|
|
#include "kernel/integrator/shade_surface.h"
|
|
#include "kernel/integrator/shade_volume.h"
|
|
|
|
CCL_NAMESPACE_BEGIN
|
|
|
|
ccl_device void integrator_megakernel(KernelGlobals kg,
|
|
IntegratorState state,
|
|
ccl_global float *ccl_restrict render_buffer)
|
|
{
|
|
/* Each kernel indicates the next kernel to execute, so here we simply
|
|
* have to check what that kernel is and execute it. */
|
|
while (true) {
|
|
/* Handle any shadow paths before we potentially create more shadow paths. */
|
|
const uint32_t shadow_queued_kernel = INTEGRATOR_STATE(
|
|
&state->shadow, shadow_path, queued_kernel);
|
|
if (shadow_queued_kernel) {
|
|
switch (shadow_queued_kernel) {
|
|
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_SHADOW:
|
|
integrator_intersect_shadow(kg, &state->shadow);
|
|
continue;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_SHADOW:
|
|
integrator_shade_shadow(kg, &state->shadow, render_buffer);
|
|
continue;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_LIGHT_NEE:
|
|
integrator_shade_light_nee(kg, &state->shadow, render_buffer);
|
|
continue;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADOW_PATH_MNEE_PENDING:
|
|
/* Not a real kernel, only a state to keep it alive until
|
|
* shade_surface uses this shadow path. */
|
|
break;
|
|
default:
|
|
kernel_assert(0);
|
|
break;
|
|
}
|
|
}
|
|
|
|
/* Handle any AO paths before we potentially create more AO paths. */
|
|
const uint32_t ao_queued_kernel = INTEGRATOR_STATE(&state->ao, shadow_path, queued_kernel);
|
|
if (ao_queued_kernel) {
|
|
switch (ao_queued_kernel) {
|
|
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_SHADOW:
|
|
integrator_intersect_shadow(kg, &state->ao);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_SHADOW:
|
|
integrator_shade_shadow(kg, &state->ao, render_buffer);
|
|
break;
|
|
default:
|
|
kernel_assert(0);
|
|
break;
|
|
}
|
|
continue;
|
|
}
|
|
|
|
/* Then handle regular path kernels. */
|
|
const uint32_t queued_kernel = INTEGRATOR_STATE(state, path, queued_kernel);
|
|
if (queued_kernel) {
|
|
switch (queued_kernel) {
|
|
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_CLOSEST:
|
|
integrator_intersect_closest(kg, state, render_buffer);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_BACKGROUND:
|
|
integrator_shade_background(kg, state, render_buffer);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE:
|
|
integrator_shade_surface(kg, state, render_buffer);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_VOLUME:
|
|
integrator_shade_volume(kg, state, render_buffer);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_VOLUME_RAY_MARCHING:
|
|
integrator_shade_volume_ray_marching(kg, state, render_buffer);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE:
|
|
integrator_shade_surface_raytrace(kg, state, render_buffer);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_LIGHT_FORWARD:
|
|
integrator_shade_light_forward(kg, state, render_buffer);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_SHADE_DEDICATED_LIGHT:
|
|
integrator_shade_dedicated_light(kg, state, render_buffer);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_SUBSURFACE:
|
|
integrator_intersect_subsurface(kg, state);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_VOLUME_STACK:
|
|
integrator_intersect_volume_stack(kg, state);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_DEDICATED_LIGHT:
|
|
integrator_intersect_dedicated_light(kg, state);
|
|
break;
|
|
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_MNEE:
|
|
integrator_intersect_mnee(kg, state);
|
|
break;
|
|
default:
|
|
kernel_assert(0);
|
|
break;
|
|
}
|
|
continue;
|
|
}
|
|
|
|
break;
|
|
}
|
|
}
|
|
|
|
CCL_NAMESPACE_END
|