Raising: unroll budget for constant-bound loops - #3004
Merged
Conversation
Nested constant-bound loops unroll multiplicatively: mfem's generic max-degree EA fallback kernels reached millions of ops and a 2 GB object (one embedded module was a gigabyte of text). Estimate the unrolled op count and, past a 64k budget, try the while raising first even when unrolling is preferred: the diffusion EA object shrinks from 2 GB to 78 MB and still raises fully. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016zErYp7upmqr4NHfhod9UD
unroll_budget (default 65536) is the nested unrolled op count past which a constant-bound affine.for raises as a while; -1 disables the budget. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016zErYp7upmqr4NHfhod9UD
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016zErYp7upmqr4NHfhod9UD
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Nested constant-bound loops unroll multiplicatively: mfem's generic max-degree EA fallback kernels (runtime degree, loops bounded by
MAX_D1D-style constants) reached millions of ops —bilininteg_diffusion_ea.cppproduced a 2 GB object whose largest embedded module was a gigabyte of textual IR with SSA ids past 2.2M. Estimate the unrolled op count recursively and, past a budget, try the while raising first even when unrolling is preferred (prefer_while_raising=false): the diffusion EA object shrinks to 78 MB, convection EA 935 MB → 30 MB, all still raising fully.The budget is a pass option,
unroll_budget(default 65536, the nested unrolled op count past which a constant-boundaffine.forraises as a while);-1disables it. The test covers the default, a tight budget that pushes a 4-trip loop to a while, and-1unrolling the 100000-trip loop.Part of #2968.
🤖 Generated with Claude Code
https://claude.ai/code/session_016zErYp7upmqr4NHfhod9UD