Cache quote unpickling #12242
Cache quote unpickling #12242
Conversation
|
test performance with #quotes please |
|
performance test scheduled: 7 job(s) in queue, 1 running. |
|
In what situations will the same quote be unpickled multiple times? |
|
I will try different variants. The current is using the smallest cashing possible, the cache lives during the expansion of a single macro but skips some due to a bug. This is enough for a recursive macro to be optimized. Though I am not sure if this difference will be noticeable with the current benchmarks. The first step will be to create a benchmark where this difference can be noticed. Also see #12228. |
nicolasstucki
force-pushed
the
add-cache-to-quote-unpickling
branch
from
87aad4c to
c0e7736
Compare
|
Performance test finished successfully: Visit https://dotty-bench2.epfl.ch/12242/ to see the changes. Benchmarks is based on merging with master (3d8b6a1) |
nicolasstucki
force-pushed
the
add-cache-to-quote-unpickling
branch
from
bae29e3 to
bd2fd57
Compare
|
test performance with #quotes please |
|
performance test scheduled: 1 job(s) in queue, 0 running. |
|
Performance test finished successfully: Visit https://dotty-bench2.epfl.ch/12242/ to see the changes. Benchmarks is based on merging with master (007dacd) |
|
test performance with #quotes please |
|
performance test scheduled: 1 job(s) in queue, 0 running. |
|
@nicolasstucki You may need to clear the cache to see the result. |
|
Performance test finished successfully: Visit https://dotty-bench2.epfl.ch/12242/ to see the changes. Benchmarks is based on merging with master (007dacd) |
I noticed |
nicolasstucki
force-pushed
the
add-cache-to-quote-unpickling
branch
from
bd2fd57 to
d6eed31
Compare
|
test performance with #quotes please |
|
performance test scheduled: 1 job(s) in queue, 0 running. |
|
test performance with #quotes please |
|
performance test scheduled: 2 job(s) in queue, 1 running. |
|
Performance test finished successfully: Visit https://dotty-bench2.epfl.ch/12242/ to see the changes. Benchmarks is based on merging with master (bb5e140) |
1 similar comment
|
Performance test finished successfully: Visit https://dotty-bench2.epfl.ch/12242/ to see the changes. Benchmarks is based on merging with master (bb5e140) |
|
test performance with #quotes please |
|
performance test scheduled: 1 job(s) in queue, 1 running. |
|
The per run cache did not seem to help much in https://github.com/lampepfl/dotty/blob/master/tests/bench/power-macro/PowerInlined-1k.scala even though we expand 8 macros |
|
test performance with #quotes please |
2 similar comments
|
test performance with #quotes please |
|
test performance with #quotes please |
|
performance test scheduled: 1 job(s) in queue, 1 running. |
|
test performance with #quotes please |
|
performance test scheduled: 3 job(s) in queue, 1 running. |
|
Performance test finished successfully: Visit https://dotty-bench2.epfl.ch/12242/ to see the changes. Benchmarks is based on merging with master (c14ec43) |
nicolasstucki
force-pushed
the
add-cache-to-quote-unpickling
branch
from
34bfc22 to
94f069e
Compare
|
test performance with #quotes please |
|
performance test scheduled: 8 job(s) in queue, 0 running. |
|
test performance with #quotes please |
|
performance test scheduled: 8 job(s) in queue, 0 running. |
|
test performance with #quotes please |
nicolasstucki
force-pushed
the
add-cache-to-quote-unpickling
branch
from
a394572 to
97d747a
Compare
|
test performance with #quotes please |
|
performance test scheduled: 8 job(s) in queue, 0 running. |
1 similar comment
|
performance test scheduled: 8 job(s) in queue, 0 running. |
|
Performance test finished successfully: Visit https://dotty-bench2.epfl.ch/12242/ to see the changes. Benchmarks is based on merging with master (b1a3e6d) |
nicolasstucki
force-pushed
the
add-cache-to-quote-unpickling
branch
from
97d747a to
00115c6
Compare
|
test performance with #quotes please |
|
performance test scheduled: 1 job(s) in queue, 1 running. |
|
Performance test finished successfully: Visit https://dotty-bench2.epfl.ch/12242/ to see the changes. Benchmarks is based on merging with master (2d2038e) |
| // Copy the cached tree to make sure the all definitions are unique. | ||
| TreeTypeMap(oldOwners = List(owner), newOwners = List(owner)).apply(tree) | ||
| case _ => | ||
| tree |
There was a problem hiding this comment.
In case we have optimizations that make the TreeTypeMap and identity function, do we have a test to guard against such optimizations?
There was a problem hiding this comment.
We already do that optimization outside TreeTypeMap, for example in changeOwner.
We do have tests that would fail if this stops cloning the symbols.
No description provided.