lexical-binding causes a lot more GC, why?

lexical-binding causes a lot more GC, why?

[Le Wang]

[-- Attachment #1: Type: text/plain, Size: 288 bytes --]

Hi all,

We found, while doing optimizations in flx (
https://github.com/lewang/flx/pull/76#issuecomment-151689064) that enabling
lexical-binding caused a lot of GC runs.

I thought lexical-binding should always result in perf gains, does anyone
know why this could be happening?

-- 
Le

[-- Attachment #2: Type: text/html, Size: 496 bytes --]

Re: lexical-binding causes a lot more GC, why?

[John Wiegley]

>>>>> Le Wang <l26wang@gmail.com> writes:

> I thought lexical-binding should always result in perf gains, does anyone
> know why this could be happening?

Hi Le, would it be possible to create a test that takes the memory-use-counts
before some activity, then after, with and without lexical-binding enabled, so
that we could fail if lexical-binding has bad behavior? I think that should be
more efficient, so I'd love to pin this down in an isolated case.

John

Re: lexical-binding causes a lot more GC, why?

[Paul Eggert]

On 10/28/2015 08:12 AM, Le Wang wrote:
>
> I thought lexical-binding should always result in perf gains

Isn't lexical binding implemented by passing around a list of bindings?  
That might explain the performance problems you see. (Sorry, I haven't 
looked at the actual code; this is just my vague recollection.)

Re: lexical-binding causes a lot more GC, why?

[Artur Malabarba]

[-- Attachment #1: Type: text/plain, Size: 447 bytes --]

On 30 Oct 2015 9:14 pm, "Paul Eggert" <eggert@cs.ucla.edu> wrote:
>
> On 10/28/2015 08:12 AM, Le Wang wrote:
>>
>>
>> I thought lexical-binding should always result in perf gains
>
>
> Isn't lexical binding implemented by passing around a list of bindings?
That might explain the performance problems you see. (Sorry, I haven't
looked at the actual code; this is just my vague recollection.)

I believe that's only the case for non compiled code.

[-- Attachment #2: Type: text/html, Size: 632 bytes --]