[
https://issues.apache.org/jira/browse/PARQUET-1841?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17082924#comment-17082924
]
Frank Du commented on PARQUET-1841:
-----------------------------------
Thanks. Seems we can speed up the spacing with "_AVX-512 Vector Byte
Manipulation Instructions 2 (VBMI2)" which introduced from Ice Lake(the Sunny
Cove core). Put valid_bits maps to k, values_read to a, set src to zero. If we
implement with this Intrinsic but it can only benefit the latest X86 device
shipped from this year, can it be up-streamed?_
!image-2020-04-14-15-01-48-222.png!
> [C++] Experiment to see if using SIMD shuffle operations for DecodeSpaced
> improves performance
> ----------------------------------------------------------------------------------------------
>
> Key: PARQUET-1841
> URL: https://issues.apache.org/jira/browse/PARQUET-1841
> Project: Parquet
> Issue Type: Improvement
> Components: parquet-cpp
> Reporter: Micah Kornfield
> Priority: Major
> Attachments: image-2020-04-14-15-01-48-222.png
>
>
> Followup from PARQUET-1840 for current benchmarks it seems that doing
> removing the memset somehow either has no impact or is slightly worse. We
> should investigate using SIMD operations to speed up spacing.
>
> As part of this we can see if moving the memset to only cover uninitialized
> values after moving all required values provides any speedup.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
