Avoid Chrome `max_old_space_size` config requirement?

[John Muehlhausen]

I have a memory-intensive process that runs fine on Firefox (everywhere) and Chrome (Max+Linux) and Safari.  However, Chrome on Windows "aw, snap"'s me most-- but not all-- of the time.  I was finally able to resolve this with `--max_old_space_size=4096` Chrome command line parameter, but I don't want to place this requirement on my users.

My guess is that gc is not keeping up with me on Windows in particular (fewer points where gc wants to run? ) or else that the defaults differ per platform.  Any ideas about how to nudge garbage collection to happen more often?  Any other thoughts?

Thanks!

[John Muehlhausen]

It seems as if my entire heap is being duplicated (due to ALLOW_MEMORY_GROWTH) and then the old one is garbage collected?  Is that how it works?  Seems like I cannot set an initial size more than 1GB and the "growth" increment is less than 1GB.  What I'd really like is to set 2GB (the max?) initially, but it seems as if this doesn't work?

[Alon Zakai]

Yes, when memory grows in asm.js we create a new one, and the old one will be GC'd eventually, but when depends on the browser. So both will be in memory at once for a while, which can be a problem.

Setting 2GB on Chrome is an issue currently, but I think you can set 2GB - 64K, so you can get close.

--
You received this message because you are subscribed to the Google Groups "emscripten-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to emscripten-disc...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[John Muehlhausen]

I tried doing a malloc/memset/free of 1800000000ull at the beginning of my program, which seemed to fix the problem.  Once the heap grows this large, is it ever downsized?  If it is downsized then I’m not sure why this worked, because it seems like there would be no effect from the temporary usage.

I don’t think I can set 2gb - 64k because I think the initial memory was limited to powers of two for a security reason?  So 1gb is the largest initial....

You received this message because you are subscribed to a topic in the Google Groups "emscripten-discuss" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/emscripten-discuss/1V1aoe7Pxlg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to emscripten-disc...@googlegroups.com.

[Alon Zakai]

Memory never shrinks, once grown.

There's no power of 2 requirement - asm.js requires multiples of 16MB, and wasm 64Kb.

[John Muehlhausen]

Thank you... avoiding growth and setting 2gb - 64kb ended up being the most reliable approach.

On Monday, March 18, 2019 at 9:01:52 PM UTC-5, Alon Zakai wrote:

Memory never shrinks, once grown.

There's no power of 2 requirement - asm.js requires multiples of 16MB, and wasm 64Kb.

On Mon, Mar 18, 2019 at 5:42 PM John Muehlhausen <j...@jgm.org> wrote:

I tried doing a malloc/memset/free of 1800000000ull at the beginning of my program, which seemed to fix the problem.  Once the heap grows this large, is it ever downsized?  If it is downsized then I’m not sure why this worked, because it seems like there would be no effect from the temporary usage.

I don’t think I can set 2gb - 64k because I think the initial memory was limited to powers of two for a security reason?  So 1gb is the largest initial....

On Mon, Mar 18, 2019 at 7:16 PM Alon Zakai <alon...@gmail.com> wrote:

Yes, when memory grows in asm.js we create a new one, and the old one will be GC'd eventually, but when depends on the browser. So both will be in memory at once for a while, which can be a problem.

Setting 2GB on Chrome is an issue currently, but I think you can set 2GB - 64K, so you can get close.

On Mon, Mar 18, 2019 at 4:56 PM John Muehlhausen <j...@jgm.org> wrote:

It seems as if my entire heap is being duplicated (due to ALLOW_MEMORY_GROWTH) and then the old one is garbage collected?  Is that how it works?  Seems like I cannot set an initial size more than 1GB and the "growth" increment is less than 1GB.  What I'd really like is to set 2GB (the max?) initially, but it seems as if this doesn't work?

On Monday, March 18, 2019 at 4:39:18 PM UTC-5, John Muehlhausen wrote:

I have a memory-intensive process that runs fine on Firefox (everywhere) and Chrome (Max+Linux) and Safari.  However, Chrome on Windows "aw, snap"'s me most-- but not all-- of the time.  I was finally able to resolve this with `--max_old_space_size=4096` Chrome command line parameter, but I don't want to place this requirement on my users.

My guess is that gc is not keeping up with me on Windows in particular (fewer points where gc wants to run? ) or else that the defaults differ per platform.  Any ideas about how to nudge garbage collection to happen more often?  Any other thoughts?

Thanks!

--
You received this message because you are subscribed to the Google Groups "emscripten-discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to emscripten-discuss+unsub...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "emscripten-discuss" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/emscripten-discuss/1V1aoe7Pxlg/unsubscribe.

To unsubscribe from this group and all its topics, send an email to emscripten-discuss+unsub...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "emscripten-discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to emscripten-discuss+unsub...@googlegroups.com.