Bug 122796
| Summary: | Fix code comment in nativeForGenerator() about ARM64 calling convention | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Daniel Bates <dbates> |
| Component: | JavaScriptCore | Assignee: | Nobody <webkit-unassigned> |
| Status: | NEW | ||
| Severity: | Normal | CC: | barraclough, fpizlo, funnyjokes1.com, ggaren, maxwellstella6, oliver |
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| Bug Depends on: | 122762 | ||
| Bug Blocks: | |||
Daniel Bates
As remarked by Oliver Hunt in bug #122762, comment 15, the following code comment in function nativeForGenerator() in file Source/JavaScriptCore/jit/ThunkGenerators.cpp incorrectly describes the ARM64 calling convention:
// Calling convention: f(edi, esi, edx, ecx, ...);
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Daniel Bates
Can someone save me some time and let me know what the correct calling convention for ARM64 is? Then I can update the comment in the patch for bug #122762. Otherwise, when I have a moment I'll look to fix up this comment.
Filip Pizlo
The best place to look for the calling convention is GPRInfo.h. Look for argumentGPR0, argumentGPR1, etc.
Stellamaxwell
The nativeForGenerator() this fix is making ready for transmitting javascript direction if conceivable. To do that, we have to decide javascript direction is upheld in the given CPU. We include a capacity nativeForGenerator(), which is capable to gather. CPU highlights if important in Linux, we can utilize an assistant vector to get the data without parsing nativeForGenerator(). https://www.courseworkempire.co.uk