Updates:
Status: Assigned
Owner: [email protected]
Comment #3 on issue 4374 by [email protected]: Math.max inside
asm.js produces the wrong result with --turbo_inlining
https://code.google.com/p/v8/issues/detail?id=4374
Yes, this is known issue, the C++ fallback implementation if %Arguments
and %ArgumentsLength (i.e. implemented in runtime-scopes.cc at the bottom)
are not composable with regard to inlining. There are two parts to this
problem that needs to be fixed before shipping general-purpose inlining.
1) A certain class of intrinsics can only be lowered after all inlining is
finished. So once JSIntrinsicLowering supports %_Arguments
and %_ArgumentsLength, it imposes a certain phase ordering on the pipeline.
This ordering restriction needs to made obvious, otherwise it will turn
into a treasure trove of bugs.
2) The fallback implementations in C++ need to be adapted so that they
compose well with respect to inlining. They should both use deoptimization
information to reconstruct the original arguments, similar to what
Accessors::FunctionGetArguments is doing.
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
--
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
