[vulkan] Codegen & runtime improvements #5213
Merged
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
for more information, see https://pre-commit.ci
✅ Deploy Preview for docsite-preview canceled.
|
ailzhang
approved these changes
Jun 22, 2022
for more information, see https://pre-commit.ci
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Outdated suggestions cannot be applied.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.
Two changes have been made:
Introduces a "Type compiler" system, that contains an IR that can be built easily from C++ or translated from Taichi SNodes, and then the type IR can be simplified or transformed, then compiles down to SPIR-V. It automatically handles layouts with the concept of a
LayoutContext, and it is now used to handle the generation of the argument buffer struct type in the codegen. The next step would be to use this system on the root buffer, and solve the aliasing problem, by using a type transformer that solves the appropriate storage types for root buffers. This also potentially helps the effort of adding sparse fields support into the SPIR-V codegen. This IR is largely inspired by MLIR (but as we can not upgrade our LLVM yet, it would be hard to directly use MLIR, thus this solution)Changes the runtime code, elimiated the
generate_command_listcall so that we don't need to pass massive amount of arguments into that function. In addition, external pointer access analysis is re-introduced, with this information host shadow buffers are also introduced to reduce device to host transfer overhead. This finally fixed the performance regression seen on stable fluid and cornell box examples when using old GUI.