On 2015/02/20 08:51:52, Sven Panne wrote:
tl;dr The kBootCodeSizeMultiplier change is OK, but the OOL constant pool is
broken and anything related to it should not be touched at all.

Longer version: We want to remove the OOL constant pool (probably in the next
quarter) and do something different, probably architecture-dependent. The
long-term vision is: Everything which is patched at runtime will live in the type feedback vector, everything else which is constant (e.g. doubles, 64bit
constants, references etc.) will live in the code object and will only be
changed by the GC. More concretely: The ARM port will probably just stay like
it
is with the interleaved constant pools, Intel ports will probably have some
kind
of constant pool at the end of the code object (addressed in a IP-relative way
on x64, addressed via absolute addresses on ia32), etc.

What this means for the PPC port: Don't touch and/or use the OOL constant pool
and do whatever is needed to get things running for you locally to the PPC
port.
Sorry if we didn't communicate this clearly enough, we needed to figure out
internally where we want to go first. :-}

At the very least we are going to need changes in the common code to allow us to

maintain a dedicated constant pool register and slot in the stack frame. This is

because Power architecture does not support pc-relative loads.
We can limit this with a PPC specific guard name but we thought this could also
be leveraged other platforms like MIPs.

Was your concern with the way we guarded changes in the common code (using
existing
out of line flags)  for the base register and slot or that we require
that support at all in the alternate approach.


https://codereview.chromium.org/882263003/

--
--
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.

Reply via email to