blender
/
blender
136
407
Fork 589
Code Issues 6.3k Pull Requests 768 Projects 19 Wiki Activity

EEVEE-Next: Fix motion blur large geom on iGPU HD520 devices #120915

Merged
Jeroen Bakker merged 8 commits from Jeroen-Bakker/blender:eevee/fix/copy-velocity-groups into main 2024-05-06 17:02:37 +02:00
Conversation 4 Commits 8 Files Changed 2 +15 -8
2 changed files with 15 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
source/blender/draw/engines/eevee_next/eevee_velocity.cc
View File

@ -279,7 +279,9 @@ void VelocityModule::geometry_steps_fill()
copy_ps.push_constant("start_offset", geom.ofs);
copy_ps.push_constant("vertex_stride", int(format->stride / 4));
copy_ps.push_constant("vertex_count", geom.len);
copy_ps.dispatch(int3(divide_ceil_u(geom.len, VERTEX_COPY_GROUP_SIZE), 1, 1));
uint group_len_x = divide_ceil_u(geom.len, VERTEX_COPY_GROUP_SIZE);
uint verts_per_thread = divide_ceil_u(group_len_x, GPU_max_work_group_count(0));
copy_ps.dispatch(int3(group_len_x / verts_per_thread, 1, 1));
}
}

19
source/blender/draw/engines/eevee_next/shaders/eevee_vertex_copy_comp.glsl
View File

@ -2,14 +2,19 @@
*
* SPDX-License-Identifier: GPL-2.0-or-later */
#pragma BLENDER_REQUIRE(gpu_shader_math_base_lib.glsl)
Jeroen-Bakker marked this conversation as resolved Outdated
Clément Foucault commented 2024-04-30 11:09:17 +02:00
Outdated
Review
Copy Link

Do not use this header. Use gpu_shader_math_base_lib.glsl instead. common_math_lib is legacy.

Do not use this header. Use `gpu_shader_math_base_lib.glsl` instead. `common_math_lib` is legacy.
void main()
{
uint vertex_id = gl_GlobalInvocationID.x;
if (vertex_id >= uint(vertex_count)) {
return;
uint vertices_per_thread = divide_ceil(uint(vertex_count), uint(gl_WorkGroupSize.x)) /
gl_NumWorkGroups.x;
uint vertex_start = min(gl_GlobalInvocationID.x * vertices_per_thread, uint(vertex_count));
uint vertex_end = min(vertex_start + vertices_per_thread, uint(vertex_count));
for (uint vertex_id = vertex_start; vertex_id < vertex_end; vertex_id++) {
out_buf[start_offset + vertex_id] = vec4(in_buf[vertex_id * vertex_stride + 0],
in_buf[vertex_id * vertex_stride + 1],
in_buf[vertex_id * vertex_stride + 2],
1.0);
}
out_buf[start_offset + vertex_id] = vec4(in_buf[vertex_id * vertex_stride + 0],
in_buf[vertex_id * vertex_stride + 1],
in_buf[vertex_id * vertex_stride + 2],
1.0);
}