Use CompilerCaching 0.5 specialization entries - #310
Merged
Conversation
maleadt
force-pushed
the
tb/compilercaching-0.5
branch
from
September 5, 2026 19:48
70ec2c7 to
bf22b3e
Compare
maleadt
changed the title
CompilerCaching 0.5 exposes const-specialized inference through typed entries. Read each entry's source and return type together, and extract its CuTileResults from const-specialized lookups before passing them to the compilation phases. This removes the manual analysis-results walk and the untyped source lookup. Cover cold and warm lookups, separate results for different constants, and specialization-specific return types in a host-side regression.
maleadt
force-pushed
the
tb/compilercaching-0.5
branch
from
September 5, 2026 20:43
bf22b3e to
1b21e53
Compare
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Use CompilerCaching 0.5's typed specialization entries to retrieve const-specialized source, return types, and compilation results. This removes the manual analysis-results walk and adapts the const-specialized
lookupresult while preserving the driver's(ci, CuTileResults)interface.Add a host-side regression covering cold and warm lookups, separate results for different constants, and specialization-specific return types.
CI build 632 failed during dependency resolution on all three Julia versions: cuTile requires CompilerCaching 0.5, but GPUCompiler only allows 0.4. CUDA/CUDACore require GPUCompiler. This PR needs a GPUCompiler release containing JuliaGPU/GPUCompiler.jl#922.
Rebased onto current
main. Local validation used CompilerCaching 0.5.0, CUDA 6.3.1, and GPUCompiler with the compat change from #922, in isolated environments:host/inference_cache,codegen/integration, anddevice/core: 166 passed on Julia 1.11.9; 168 passed on 1.12.7; 168 passed on 1.13.0-rc4. Device tests ran on an RTX 5080.