Miguel Pozo pragma37
  • Joined on 2018-04-26
Miguel Pozo pushed to pull-gpu-batch-compilation at pragma37/blender 2024-05-29 16:45:49 +02:00
e5f80b1c13 Fix cmd path
Miguel Pozo pushed to test-parallel-compilation-workers-2 at pragma37/blender 2024-05-29 16:43:20 +02:00
28abd99509 Fix cmd path
Miguel Pozo pushed to test-parallel-compilation-workers-2 at pragma37/blender 2024-05-28 18:38:22 +02:00
f316674c13 Compute and geometry shaders support
Miguel Pozo pushed to test-parallel-compilation-workers-2 at pragma37/blender 2024-05-28 15:47:48 +02:00
21bafe2123 Merge branch 'main' into test-parallel-compilation-workers-2
2dab2b34c2 Fix vulkan builds
ceeaea0c8a Fix: Switch to scheduled trigger for appropriate permission scope in workflow
b15bc1b5bb Extensions: add the blender version to the remote URL
171319d9ef Extensions: send access token for HTTP requests
Compare 189 commits »
Miguel Pozo pushed to test-parallel-compilation-workers-2 at pragma37/blender 2024-05-27 19:38:27 +02:00
942f2200f5 Rename SpecializationConstat::default_value > value
4c2ce68e42 Share SpecializationConstant implementations
5ffddad2ca Squash-merge pull-gpu-batch-compilation
Compare 3 commits »
Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 18:13:38 +02:00
GPU: Add GPU_shader_batch_create_from_infos

But it is better to not "hide" static variables inside functions.

I mean, but that's my point. For this specific case I really appreciate the safety of limiting the scope of this variable as…

Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 18:08:38 +02:00
GPU: Add GPU_shader_batch_create_from_infos

I'd like to hear @Sergey 's take on this first.

Miguel Pozo closed pull request blender/blender#121471 2024-05-27 17:26:29 +02:00
WIP: GPU: Add batch shader compilation API and per backend ShaderCompiler
Miguel Pozo commented on pull request blender/blender#121471 2024-05-27 17:26:27 +02:00
WIP: GPU: Add batch shader compilation API and per backend ShaderCompiler

Replaced by #122232.

Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 17:09:31 +02:00
GPU: Add GPU_shader_batch_create_from_infos

When shader compilation fails, what will be reported to the user and which process is doing the reporting.

When compilation fails either at the subprocess level or locally after loading the…

Miguel Pozo pushed to pull-gpu-batch-compilation at pragma37/blender 2024-05-27 17:05:48 +02:00
d683e11637 Fix typo
Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 17:04:05 +02:00
GPU: Add GPU_shader_batch_create_from_infos

I've cleaned the class interface and documented the enums 30efe59382. But I kept the names the same since to me it's easier to understand the order this way. And I prefer all the compilation steps…

Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 16:56:34 +02:00
GPU: Add GPU_shader_batch_create_from_infos

I replaced it with a sizeof('\0') to make it explicit.

Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 16:55:37 +02:00
GPU: Add GPU_shader_batch_create_from_infos

Are you sure about this? I don't get the reasoning. I can't find any related code style, and if you look for _start in the code it's almost always paired with an _end.

Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 16:52:50 +02:00
GPU: Add GPU_shader_batch_create_from_infos

What's your reasoning here? I prefer to make it as local as possible, since two workers with the same id will wreak havoc, so this seems a lot safer to me. This shouldn't be needed anywhere else.

Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 16:50:04 +02:00
GPU: Add GPU_shader_batch_create_from_infos

Even then, I think the worst thing that can happen is that it reads garbage, the validation fails and the shader is compiled again. For writing both should write the same result, but even if that…

Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 16:43:55 +02:00
GPU: Add GPU_shader_batch_create_from_infos

It's 128 bit in total (vertex + shader). An easy extra measure could be to append the string lengths to the hash string.

Of course, hash collisions are always possible in theory.

Miguel Pozo commented on pull request blender/blender#122232 2024-05-27 16:40:20 +02:00
GPU: Add GPU_shader_batch_create_from_infos

You're right, I've added multiple error messages, just in case.