Bug 89940
| Summary: | JSC should try to make profiling deterministic because otherwise reproducing failures is nearly impossible | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Filip Pizlo <fpizlo> |
| Component: | JavaScriptCore | Assignee: | Filip Pizlo <fpizlo> |
| Status: | RESOLVED FIXED | ||
| Severity: | Normal | ||
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | All | ||
| OS: | All | ||
| Bug Depends on: | 89802 | ||
| Bug Blocks: | |||
Filip Pizlo
We should in future do more exploration of the stochastic profiling from https://bugs.webkit.org/show_bug.cgi?id=89802. But without some sort of record-replay mode, this sort of randomness greatly increases the work required to narrow in on compiler bugs.
This bug is concerned with just removing the random part of http://trac.webkit.org/changeset/121215. It still leaves the ability to periodically compute least-upper-bounds of value profiles by setting an artificially low execution count threshold.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Filip Pizlo
Rubber-stamped in person by Gavin Barraclough.
Filip Pizlo
Landed in http://trac.webkit.org/changeset/121218