Suggestion about AcceleratedCRC32C()

[VinceCui]

I think the Process(unaligned bytes) in function in file port_posix_sse.cc maybe wrong.
The original code does m times STEP1, where m = p%8.
If p=10,then m=2,and the new p will be equal to 12, which is not aligned by 8.
Thannks for reading.

[googlebot]

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please visit https://cla.developers.google.com/ to sign.

Once you've signed, please reply here (e.g. I signed it!) and we'll verify. Thanks.

---

• If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Check your existing CLA data and verify that your email is set on your git commits.
• If your company signed a CLA, they designated a Point of Contact who decides which employees are authorized to participate. You may need to contact the Point of Contact for your company and ask to be added to the group of authorized contributors. If you don't know who your Point of Contact is, direct the project maintainer to go/cla#troubleshoot.
• In order to pass this check, please resolve this problem and have the pull request author add another comment and the bot will run again.

[VinceCui]

I signed it!

[googlebot]

CLAs look good, thanks!

[pwnall]

Can you please submit a test that the original code fails, but your test passes? You can use the non-accelerated implementation to get a reference value.

[VinceCui]

I know the original code can get the right result. However, the original code can’t do a right alignment work.
For example, when you use the original code, if p = 0x0003, then reinterpret_cast<uintptr_t>(p) % 8 will be 3. That means you should do 5 times STEP1 so that p will be added to 0x0008, which is aligned. But the original code will perform 3 times STEP1, which can just add p to 0x0006.
My suggestion will always let the p aligned to a multiple of 8, which is the purpose of that code.
Of course, the original code can get the right result because it just do a wrong alignment work, which makes the accelerate slow. As far as I know, the right alignment work can make the memory access faster.
Thanks for reading. May you have a good summer!

[pwnall]

@VincentVessalius Thank you for explaining! The explanation makes sense, and a fix is on the way.

[pwnall]

This landed in 2964b80. Thank you very much for your contribution!