CoSSJIT: Combining Static Analysis and Speculation in JIT Compilers
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Just-in-time (JIT) compilers typically sacrifice the precision of program analysis for efficiency, but are capable of performing sophisticated speculative optimizations based on run-time profiles to generate code that is specialized to a given execution. On the contrary, ahead-of-time static compilers can often afford precise flow-sensitive interprocedural analysis, but produce conservative results in scenarios where higher precision could be derived from run-time specialization. In this paper, we propose the first-of-its-kind approach to enrich static analysis with the possibility of speculative optimization during JIT compilation, as well as its usage to perform aggressive stack allocation on a production Java Virtual Machine (JVM). Our approach of combining static analysis with JIT speculation – named CoSSJIT – involves three key contributions. First, we identify the scenarios where a static analysis would make conservative assumptions but a JIT could deliver precision based on run-time speculation. Second, we present the notion of ‘‘speculative conditions’’ and plug them into a static interprocedural dataflow analyzer (whose aim is to identify heap objects that can be allocated on stack), to generate partial results that can be specialized at run-time. Finally, we extend a production JIT compiler to read and enrich static-analysis results with the resolved values of speculative conditions, leading to a practical approach that efficiently combines the best of both worlds. Cherries on the cake: Using CoSSJIT , we obtain 5.7× improvement in stack allocation (translating to performance), while building on a system that ensures functional correctness during JIT compilation.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it