Enable SSE2 compression path to work on MSVC

[TrianglesPCT]

The compile time detection uses the __SSE2__ predefined macro, but this does not exist on MSVC--resulting it that compiler always using scalar path.

On MSVC you can detect SSE support by checking for _M_AMD64, as when targeting x64 it requires SSE2 as a baseline, this patch adds this check.

[facebook-github-bot]

Hi @TrianglesPCT!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at [email protected]. Thanks!

[TrianglesPCT]

.

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Facebook open source project. Thanks!

[TrianglesPCT]

Switch to unaligned load

[Cyan4973]

Thanks @TrianglesPCT ,

did you attempt to measure performance differences provided by your PR ?

[TrianglesPCT]

Thanks @TrianglesPCT ,

did you attempt to measure performance differences provided by your PR ?

I removed the AVX2 stuff for a later PR

[Cyan4973]

Well, even without AVX2,
this PR is presumed to help performance, because it's supposed to help SSE2 detection ?

[TrianglesPCT]

All it is doing is normalizing MSVC to target the same code path as GCC/Clang when SSE2 is set as a requirement for your compile target. Not adding any new code..

The defined(__SSE2__) check doesn't work on MSVC because that define doesn't exist even if SSE2 is enabled.

If you want specific performance comparison I can do that might be few days until I have time for that.

[terrelln]

If you want specific performance comparison I can do that might be few days until I have time for that.

That would be great! But that doesn't need to block merging this PR. You can easily measure performance with the builtin benchmark tool.

zstd -b5e12 silesia.tar

[ghost]

Will there be a hot fix for v1.5.0?

[wolfpld]

You can easily measure performance with the builtin benchmark tool.

The difference is rather spectacular. Results for 5950X.

Before:

 5#silesia.tar       : 211957760 ->  63806593 (3.322), 111.0 MB/s , 977.5 MB/s
 6#silesia.tar       : 211957760 ->  62980544 (3.365), 107.9 MB/s ,1001.6 MB/s
 7#silesia.tar       : 211957760 ->  61482185 (3.447),  73.1 MB/s ,1055.9 MB/s
 8#silesia.tar       : 211957760 ->  60914308 (3.480),  57.2 MB/s ,1076.1 MB/s
 9#silesia.tar       : 211957760 ->  59928587 (3.537),  52.1 MB/s ,1100.2 MB/s
10#silesia.tar       : 211957760 ->  59296205 (3.575),  50.3 MB/s ,1103.6 MB/s
11#silesia.tar       : 211957760 ->  59152928 (3.583),  47.7 MB/s ,1106.8 MB/s
12#silesia.tar       : 211957760 ->  58640204 (3.615),  30.6 MB/s ,1123.8 MB/s

After:

 5#silesia.tar       : 211957760 ->  63806593 (3.322), 160.8 MB/s , 980.0 MB/s
 6#silesia.tar       : 211957760 ->  62980544 (3.365), 153.3 MB/s ,1004.7 MB/s
 7#silesia.tar       : 211957760 ->  61482185 (3.447), 111.9 MB/s ,1056.3 MB/s
 8#silesia.tar       : 211957760 ->  60914308 (3.480),  90.9 MB/s ,1079.5 MB/s
 9#silesia.tar       : 211957760 ->  59928587 (3.537),  74.7 MB/s ,1101.9 MB/s
10#silesia.tar       : 211957760 ->  59296205 (3.575),  71.3 MB/s ,1107.0 MB/s
11#silesia.tar       : 211957760 ->  59152928 (3.583),  65.0 MB/s ,1113.2 MB/s
12#silesia.tar       : 211957760 ->  58640204 (3.615),  51.3 MB/s ,1129.6 MB/s

[senhuang42]

This is a great fix, thanks for the contribution!