Draft: vulkan: Add bfloat16 support #12554
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
github-actions
bot
added
Vulkan
|
The tooling for Vulkan bfloat16 is not all merged yet - see KhronosGroup/SPIRV-Tools#6057 and KhronosGroup/glslang#3905. So if anybody wants to try this locally, you'd need to build a custom glslc. I'll update when it's all merged NVIDIA will release a Vulkan developer driver with support for this extension, hopefully tomorrow. |
jeffbolznv
commented
Mar 24, 2025
| @@ -221,7 +222,8 @@ void string_to_spv_func(const std::string& _name, const std::string& in_fname, c | |||
| std::string target_env = (name.find("_cm2") != std::string::npos) ? "--target-env=vulkan1.3" : "--target-env=vulkan1.2"; | |||
|
|
|||
| // disable spirv-opt for coopmat shaders for https://github.com/ggerganov/llama.cpp/issues/10734 | |||
| std::string opt_level = coopmat ? "" : "-O"; | |||
| // XXX do not submit - temporarily disabled for bfloat | |||
| std::string opt_level = "";//coopmat ? "" : "-O"; | |||
There was a problem hiding this comment.
I don't intend to submit this line as-is. spirv-opt needs a bit of work, hopefully that will be resolved soon.
This adds bfloat16 matrix multiply support based on VK_KHR_shader_bfloat16. The extension is required for coopmat multiply support, but matrix-vector multiply trivially promotes bf16 to fp32 and doesn't require the extension. The copy/get_rows shaders also don't require the extension. It's probably possible to fall back to non-coopmat and promote to fp32 when the extension isn't supported, but this change doesn't do that. The coopmat support also requires a glslc that supports the extension, which currently requires a custom build.
jeffbolznv
force-pushed
the
bfloat16_rebase
branch
from
9ffb02b to
9cd10f8
Compare
|
I've rebased this, and added a fallback (promote to fp32) when bf16 coopmat support isn't available. Still waiting on the tooling to be merged, so still Draft for now. |
…pport Compile a variant of the scalar mul_mm shader that will promote the bf16 values to float, and use that when either the bf16 extension or the coopmat extensions aren't available.
jeffbolznv
force-pushed
the
bfloat16_rebase
branch
from
02160db to
21e8793
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This adds bfloat16 matrix multiply support based on VK_KHR_shader_bfloat16. The extension is required for coopmat multiply support, but matrix-vector multiply trivially promotes bf16 to fp32 and doesn't require the extension. The copy/get_rows shaders also don't require the extension.
It's probably possible to fall back to non-coopmat and promote to fp32 when the extension isn't supported, but this change doesn't do that.
The coopmat support also requires a glslc that supports the extension, which currently requires a custom build.