Try further optimize ReconstructSome for leopard 8+16 #274
Conversation
liamsi
force-pushed
the
elias-orijtech/master
branch
from
7a3b1b5
to
b4df876
Compare
liamsi
changed the title
liamsi
force-pushed
the
elias-orijtech/master
branch
from
b4df876
to
08a1be0
Compare
| errorBits.set(i) | ||
| } | ||
| } | ||
| for i := 0; i < r.dataShards; i++ { | ||
| if len(shards[i]) == 0 { | ||
| if len(shards[i]) == 0 /*&& (recoverAll || (required != nil && required[i]))*/ { |
There was a problem hiding this comment.
Uncommenting this leads to more significant improvements in BenchmarkDecode1K but will lead to a bunch of test failures (I think mostly because ReconstructData expects to reconstruct all original missing data). This can probably be fixed.
| errorBits.set(i) | ||
| } | ||
| } | ||
| for i := 0; i < r.dataShards; i++ { | ||
| if len(shards[i]) == 0 { | ||
| if len(shards[i]) == 0 /*&& (recoverAll || (required != nil && required[i]))*/ { |
There was a problem hiding this comment.
same as in leopard.go:
https://github.com/klauspost/reedsolomon/pull/274/files#r1511442291
|
Note that I suggest adding some more detailed tests. |
|
Yeah, that makes sense. I was hoping that @elias-orijtech could try to optimize, test, and benchmark this further. I think particularly the Line 410 in 2d78890 I was also surprised it is constructed at all if we already know with the line if it is needed or not (I'm guessing it is rather inexpensive but it seems somewhat unnecessary. |
|
Closing in favour of #272 |
Trying:
This might lead to failing tests. But if not this should be much faster than before.
builds on changes introduced in #272