GUILE 2: Increase initial heap

GUILE 2: Increase initial heap

This reduces the GC overhead for initialization from 0.3s to 0.12s.

Total time goes from 1.7 to 1.2s, because LilyPond starts out with a
larger heap after initialization.

[hanwenn, 2020-01-29 08:53:54]

overwrite

[lemzwerg, 2020-01-29 10:24:12]

LGTM

https://codereview.appspot.com/547540043/diff/561380043/lily/main.cc
File lily/main.cc (right):

https://codereview.appspot.com/547540043/diff/561380043/lily/main.cc#newcode772
lily/main.cc:772: sane_putenv("GC_INITIAL_HEAP_SIZE", "40M", false);
Do you think it makes sense to mention this environment variable in the lilypond
documentation?  Is there ever a reason for the user to change the value?

[Dan Eble, 2020-01-29 11:55:40]

On 2020/01/29 10:24:12, lemzwerg wrote:
> 
> Is there ever a reason for the user to change the value?

Say, to speed up testing?

[hanwenn, 2020-01-30 10:03:08]

On 2020/01/29 10:24:12, lemzwerg wrote:
> LGTM
> 
> https://codereview.appspot.com/547540043/diff/561380043/lily/main.cc
> File lily/main.cc (right):
> 
>
https://codereview.appspot.com/547540043/diff/561380043/lily/main.cc#newcode772
> lily/main.cc:772: sane_putenv("GC_INITIAL_HEAP_SIZE", "40M", false);
> Do you think it makes sense to mention this environment variable in the
lilypond
> documentation?  Is there ever a reason for the user to change the value?

It is too early to tell. Ideally, we would tune GC so it does the right thing.

[hanwenn, 2020-01-30 10:04:30]

For deployments (say, our builds that always run on the same dedicated machine),
it makes sense to set this to a max size.

Servers at Google usually set the equivalent parameter for Java upfront as well,
because they get a fixed allocation of RAM.