On 2014/02/05 10:52:42, Rodolph Perfetta (ARM) wrote:
https://codereview.chromium.org/143293003/diff/1/src/a64/lithium-codegen-a64.cc
File src/a64/lithium-codegen-a64.cc (right):
https://codereview.chromium.org/143293003/diff/1/src/a64/lithium-codegen-a64.cc#newcode835
src/a64/lithium-codegen-a64.cc:835: bool
LCodeGen::GenerateDeoptJumpTable() {
Right, but in the mean time a debug check may save us some time when
investigating bugs.
The branch instructions to the jump table entries already check that
they're not
too far, no?
On 2014/02/05 08:57:38, jochen wrote:
> ARM has a check that the table is not too far away. I didn't put the
check
here,
> because eventually, we'll support arbitrary long jumps, right?
https://codereview.chromium.org/143293003/diff/1/src/a64/lithium-codegen-a64.cc#newcode853
src/a64/lithium-codegen-a64.cc:853: __ Mov(__ Tmp0(),
Operand(ExternalReference::ForDeoptEntry(entry)));
In Theory it is never safe to use them outside the macro assembler. It
just so
happen that instruction used here don't need tmps at all. We'll revisit
their
uses later but in the mean time asserting the code sequence in the else
clause
below has a fixed length (therefore doesn't use tmps) would be good.
On 2014/02/05 10:34:17, jochen wrote:
> On 2014/02/05 10:25:41, ulan wrote:
> > I am not sure that we can use Tmp0, Tmp1 here. Rodolph, when is it
safe to
use
> > these registers?
>
> Tmp1() was uses in the old code (see line 1009) so I assumed it's safe.
https://codereview.chromium.org/143293003/
--
--
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/groups/opt_out.
