Closed jacob-hughes closed 2 years ago
Let's just quickly check that doing so doesn't change any performance numbers (you should be able to tell from 1 or 2 benchmarks). If it still looks good, please squash.
OK, pretend I wrote my earlier comment now :)
It still looks good. Squashed!
bors r+
Build failed:
bors try
bors try
bors try
This one runs a SOM test suite as part of CI which I forgot about and didn't run locally. It seems to unearthed a memory corruption in the String
implementation which I am investigating.
bors try
This one is ready to squash now.
Please squash.
Squashed
bors r+
Build succeeded:
You may want to consider to run the tests on top of SomSom (SOM implemented in SOM) in addition to how you're running them now.
Might be better at triggering GC-related issues than a normal test run:
/som -G -cp core-lib/Smalltalk:core-lib/TestSuite:core-lib/SomSom/src/compiler:core-lib/SomSom/src/vm:core-lib/SomSom/src/vmobjects:core-lib/SomSom/src/interpreter:core-lib/SomSom/src/primitives \
core-lib/SomSom/tests/SomSomTests.som
The core lib also got a few more tests in general since the last update, but those are less likely to be of much use.
Thanks @smarr. I'll add this to the CI.
@smarr I think we're doing that at https://github.com/softdevteam/yksom/blob/master/.buildbot.sh#L37 ? But it's so long since I've looked at this that I might be completely wrong!
Oh, you're right. I didn't realize that the stdio was just the last view lines. I should have from the line numbers, I guess my brain just filtered those out.
Hehe no problem :)
This gives us a nice performance boost on types which use
Drop
for allocation only.