X86: Fast CRC32 computation using PCLMULQDQ instruction

[frankdjx]

Based on:
"Fast CRC Computation for Generic Polynomials Using PCLMULQDQ Instruction"
V. Gopal, E. Ozturk, et al., 2009, http://intel.ly/2ySEwL0

Benchmarking with below synthetic code, the time improve to 0.002s from 0.088s on i7-8700.

function simple_crc32() {
  $a = "foo";
  for ($i = 0; $i < 10000; $i++) {
    crc32($a);
    $a .= "o";
  }
}

Signed-off-by: Frank Du [email protected]

[frankdjx]

The batch size of SIMD FOLD operation is 16(128 bit) and 64(4 fold), the new added test case cover "16, 16+3, 32, 32+3, 48, 48+3, 64, 64+3, 80, 80+3, 128, 128+3, 144, 144+3" size.

[frankdjx]

Ping @cmb69

Can someone help to review this? I don't know who are the right reviewer. This is my first commit to PHP, kindly let me know if I miss any steps.

[cmb69]

Everything fine from your side. Not sure who'd be best for reviewing; maybe @laruence?

[weltling]

It's a great functional improvement. I'd just suggest to clean up a bit so the code is explicitly generated only on a supported platform.

Thanks.

[frankdjx]

Can we merge this or more actions need? Thanks.

[Girgias]

Can we merge this or more actions need? Thanks.

Could you squash and rebase? Makes life easier when trying to merge for us. :)

[frankdjx]

Can we merge this or more actions need? Thanks.

Could you squash and rebase? Makes life easier when trying to merge for us. :)

Done. And CI is green also.