The VM already knew that: leveraging compile-time knowledge to optimize gradual typing
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
Programmers in dynamic languages wishing to constrain and understand the behavior of their programs may turn to gradually-typed languages, which allow types to be specified optionally and check values at the boundary between dynamic and static code. Unfortunately, the performance cost of these run-time checks can be severe, slowing down execution by at least 10x when checks are present. Modern virtual machines (VMs) for dynamic languages use speculative techniques to improve performance: If a particular value was seen once, it is likely that similar values will be seen in the future. They combine optimization-relevant properties of values into cacheable “shapes”, then use a single shape check to subsume checks for each property. Values with the same memory layout or the same field types have the same shape. This greatly reduces the amount of type checking that needs to be performed at run-time to execute dynamic code. While very valuable to the VM’s optimization, these checks do little to benefit the programmer aside from improving performance. We present in this paper a design for intrinsic object contracts, which makes the obligations of gradually-typed languages’ type checks an intrinsic part of object shapes, and thus can subsume run-time type checks into existing shape checks, eliminating redundant checks entirely. With an implementation on a VM for JavaScript used as a target for SafeTypeScript’s soundness guarantees, we demonstrate slowdown averaging 7% in fully-typed code relative to unchecked code, and no more than 45% in pessimal configurations.
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 | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.002 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.008 | 0.003 |
| 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