ext/bcmath: Made the same changes to _bc_do_add as _bc_do_sub

[SakiTakamachi]

benchmark codes:

1:

$num1 = '1.2345678901234567890';
$num2 = '2.12345678901234567890';

for ($i = 0; $i < 4000000; $i++) {
    bcadd($num1, $num2, 20);
}

2:

$num1 = '-12345678901234567890.12345678901234567890';
$num2 = '-212345678901234567890.12345678901234567890';

for ($i = 0; $i < 4000000; $i++) {
    bcadd($num1, $num2, 20);
}

3:

$num1 = '12345678901234567890.00000000000000000000000000000000000000000000000000';
$num2 = '212345678901234567890.00000000000000000000000000000000000000000000000000';

for ($i = 0; $i < 4000000; $i++) {
    bcadd($num1, $num2, 20);
}

before

Time (mean ± σ):     551.0 ms ±  30.3 ms    [User: 547.0 ms, System: 3.1 ms]
Range (min … max):   515.8 ms … 619.9 ms    10 runs

Time (mean ± σ):     666.1 ms ±   7.1 ms    [User: 662.5 ms, System: 2.6 ms]
Range (min … max):   655.2 ms … 679.0 ms    10 runs

Time (mean ± σ):     599.4 ms ±  34.4 ms    [User: 596.0 ms, System: 2.5 ms]
Range (min … max):   565.1 ms … 663.9 ms    10 runs

after

Time (mean ± σ):     515.2 ms ±   5.6 ms    [User: 509.9 ms, System: 4.2 ms]
Range (min … max):   508.5 ms … 526.8 ms    10 runs

Time (mean ± σ):     597.8 ms ±  10.9 ms    [User: 592.1 ms, System: 3.8 ms]
Range (min … max):   586.0 ms … 622.5 ms    10 runs

Time (mean ± σ):     559.5 ms ±  14.5 ms    [User: 554.8 ms, System: 3.6 ms]
Range (min … max):   544.5 ms … 584.3 ms    10 runs

FYI: after the first commit

Time (mean ± σ):     563.6 ms ±  23.4 ms    [User: 559.0 ms, System: 3.5 ms]
Range (min … max):   544.6 ms … 616.3 ms    10 runs

Time (mean ± σ):     637.8 ms ±   5.5 ms    [User: 633.8 ms, System: 3.0 ms]
Range (min … max):   630.0 ms … 646.7 ms    10 runs

Time (mean ± σ):     580.4 ms ±   6.6 ms    [User: 575.3 ms, System: 4.0 ms]
Range (min … max):   574.9 ms … 595.6 ms    10 runs

[SakiTakamachi]

I forgot fixing comments

[SakiTakamachi]

done

[Girgias]

CI is red

[SakiTakamachi]

It looks like the XML test is failing. Merge the latest master

[nielsdos]

Nice work, seems right. Also ran an exhaustive test for the 32-bit case and works fine.
I think it would be good to add a regression test for the issue fixed here though, because the CI didn't catch it (if possible): 64eb41a

[SakiTakamachi]

@nielsdos
I noticed this with Travis.
https://app.travis-ci.com/github/php/php-src/builds/270410835

The other CIs certainly did not detect this.

This is due to a rear overrun, but I can't think of any test to detect this... Do you have any good ideas?

(edit)

I reverted the fix and ran the nightly, but it didn't detect anything.
https://github.com/SakiTakamachi/php-src/actions/runs/9046718030

[SakiTakamachi]

Hmm, I'll merge this for now and add tests later if possible.