Umh, okay, thanks a lot!
Maybe I should elaborate my problem, hopefully I get some further help:
I (more or less with the help of a colleague) patched parts of the v8 code
(mostly the *String*) and added a few functions to the String prototype.
Let's call one of the function *isPatched(), *which checks whether a string
is patched or not (by setting the heap of the string to a different,
self-defined map).
Assume the following code:
*var s = "patchedString"for (var i = 0; i < s.length; i++) {
s[i].isPatched()}*
This is *true* for two times, and then turns *false* from the third
iteration on. I fixed this behaviour by calling *GenerateMiss(masm) *as the
first command in *KeyedLoadIC::GenerateMegamorphic (**ic-x64:286*), but
this is probably a very bad hack and I want to avoid that. Moreover, it did
not fix the same problem with other "added" functions, where the IC also
somehow interferes (return two times the right answer, from the third time
on the wrong answer)
My guess is, that the two first lookups for the characters are cache misses
(as shown in the ic-trace) and the "right" patched string is accessed. From
the third character lookup on, it is somehow a cache hit and a non-patched
string in the cache is accessed.
Is there any way to avoid this behaviour or to find a better way of having
a "cache hit on the wrong string"?
Thanks a lot in advance!
On Tuesday, March 1, 2016 at 2:34:35 PM UTC+1, Jakob Kummerow wrote:
>
> --nouse-ic (which is equivalent to changing the flag's default and
> recompiling) turns off those parts of the IC system that can reasonably be
> turned off (in particular, compilation of specialized handlers). You'll
> observe that --trace-ic's output changes quite a bit when you do that.
>
> Turning off any and all uses of inline caches is not possible. The fact
> that we have FLAG_use_ic at all is mostly a historical artifact.
>
>
> On Tue, Mar 1, 2016 at 1:53 PM, Martin Junker <nitram....@gmail.com
> <javascript:>> wrote:
>
>> Hello,
>>
>> I am using v8 4.5.103.30 and trying to disable inline-caching. I changed
>> the *use_ic *in src/flag-definitions.h to false and compiled v8.
>> Then I started d8 with the --trace-ic flag and kind of expected to not
>> have any IC's there (as the use of IC is disabled).
>> Am I right with that assumption? If not, how can I disable IC?
>>
>> Many thanks!
>>
>> Martin
>>
>> --
>> --
>> v8-users mailing list
>> v8-u...@googlegroups.com <javascript:>
>> http://groups.google.com/group/v8-users
>> ---
>> You received this message because you are subscribed to the Google Groups
>> "v8-users" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to v8-users+u...@googlegroups.com <javascript:>.
>> For more options, visit https://groups.google.com/d/optout.
>>
>
>
--
--
v8-users mailing list
v8-users@googlegroups.com
http://groups.google.com/group/v8-users
---
You received this message because you are subscribed to the Google Groups
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to v8-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
