enable no-copy stack switching implementation #29578
Conversation
JeffBezanson
force-pushed
the
jb/stackmem
branch
from
c7083f0 to
4e560e0
Compare
|
32bit failure on both Win and Linux, looks like we are running out of (virtual) memory? |
|
Ah, of course the limits will have to be much lower on 32-bit. |
JeffBezanson
force-pushed
the
jb/stackmem
branch
6 times, most recently
from
de2798a to
81e9b34
Compare
Falls back to stack copying if there are too many stack mappings. Also adds some code to free unused pooled stacks.
JeffBezanson
force-pushed
the
jb/stackmem
branch
from
81e9b34 to
8a5866a
Compare
|
🎉 |
| # else | ||
| # define MAX_STACK_MAPPINGS 500 | ||
| # endif | ||
| #endif |
There was a problem hiding this comment.
we should probably make these the default values for a runtime variable, since several of these are dependent on the amount of memory available in the system (currently these are set for the size of our CI machines, but user machines may be much larger or smaller), so it may make sense for some users to need to configure these
|
@JeffBezanson: Is there some formal way to propose a change for inclusion into the 1.1 release? This change will fix a long-standing issue in Gtk.jl (JuliaGraphics/Gtk.jl#161) but Gtk.jl can just move on (and remove the sigatom workaround) once this is in an official Julia release. |
|
It is on master so it's included in 1.1. |
|
Everything on master goes into the next minor release; backports go into patch releases. |
|
Ah great, I was a little confused by the patch releases but it makes of course much sense that way. |
Falls back to stack copying if there are too many stack mappings. Confirmed the
readtest passes with this. Hopefully this makes the new implementation robust enough to use by default. Also reduced the default stack size from 8MB to 4MB; 8MB is rather large.