Cache line reservation: exploring a scheme for cache-friendly object allocation
Bibliographic record
Abstract
This thesis presents a novel idea for object allocation, cache line reservation (CLR), whose goal is to reduce data cache misses.Certain objects are allocated from "reserved" cache lines, so that they do not evict other objects that will be needed later.We discuss what kinds of allocations can benefit from CLR, as well as sources of overhead.Prototypes using CLR were implemented in the IBM R J9 Java TM virtual machine (JVM) and its Testarossa just-in-time (JIT) compiler.A performance study of our prototypes was conducted using various benchmarks.CLR can offer a benefit in specialized microbenchmarks when allocating long-lived objects that are accessed in infrequent bursts.In other benchmarks such as SPECjbb2005 and SPECjvm2008, we show that CLR can reduce cache misses when allocating a large number of short-lived objects, but not provide a performance improvement due to the introduced overhead.We measure and quantify this overhead in the current implementation and suggest areas for future development.CLR is not limited to Java applications, so other static and dynamic compilers could benefit from it in the future.me the independence to work on research that I find interesting.Managers at IBM, Marcel Mitran and Emilia Tung, for allowing me to use IBM resources for my research.Vijay Sundaresan, Nikola Grčevski and Daryl Maier for providing me with the initial CLR idea, and offering regular advice about the technical aspects of the project, and being always available for questions.Yan Luo, for answering numerous questions about the workings of the J9 JVM.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.003 | 0.009 |
| Open science | 0.006 | 0.004 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.005 | 0.002 |
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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".