Baunsgaard commented on PR #2171:
URL: https://github.com/apache/systemds/pull/2171#issuecomment-2593418762
This commit merge the BWARE optimizations to transform encode. Attached are
the full logs for various transformations.
Most transformations improve in performance.
Notable results include :
passthough compressed 3x faster.
```txt
before
Transform Encode Perf: rows: 10000000 schema:[UINT4, UINT4, UINT4, UINT4,
UINT4, UINT4, UINT4, UINT4, UINT4, UINT4]
{}
Normal, 207.217+- 11.819 ms,
Compressed, 336.523+- 26.097 ms,
after:
Transform Encode Perf: rows: 10000000 schema:[UINT4, UINT4, UINT4, UINT4,
UINT4, UINT4, UINT4, UINT4, UINT4, UINT4]
{}
Normal, 212.838+- 4.184 ms,
Compressed, 106.941+- 43.943 ms,
```
hash to dummy encode 3 x faster uncompressed.
```txt
Before:
After:
Transform Encode Perf: rows: 10000000 schema:[INT32, INT32, INT32, INT32,
INT32, INT32, INT32, INT32, INT32, INT32]
{ids:true, hash:[1,2,3,4,5,6,7,8,9,10], K:10,
dummycode:[1,2,3,4,5,6,7,8,9,10]}
Normal, 2435.954+-267.029 ms,
Compressed, 472.872+- 35.932 ms,
```
Full logs:
[BeforeSU1.md](https://github.com/user-attachments/files/18427619/BeforeSU1.md)
[afterSU1.md](https://github.com/user-attachments/files/18427621/afterSU1.md)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
