buffer: fix performance regression

[targos]

Checklist

• make -j8 test (UNIX), or vcbuild test nosign (Windows) passes
• commit message follows commit guidelines

Affected core subsystem(s)

buffer

Description of change

V8 5.4 changed the way that the default constructor of derived classes
is called. It introduced a significant performance regression in the
buffer module for the creation of pooled buffers. This commit forces the
definition back to how it was.

Ref: https://bugs.chromium.org/p/v8/issues/detail?id=4890

Benchmark results using the script from #8738:

6.6.0:
m.js n=1024 len=10 source="array": 2940.0523751381074
m.js n=1024 len=2048 source="array": 243.34844074199046
m.js n=1024 len=10 source="arraybuffer": 5469.801373510926
m.js n=1024 len=2048 source="arraybuffer": 5897.122441515577
m.js n=1024 len=10 source="arraybuffer-middle": 5715.82551047164
m.js n=1024 len=2048 source="arraybuffer-middle": 5651.568910575927
m.js n=1024 len=10 source="buffer": 2627.6210998395145
m.js n=1024 len=2048 source="buffer": 659.2472687131358
m.js n=1024 len=10 source="uint8array": 2598.5564648340537
m.js n=1024 len=2048 source="uint8array": 270.0350801781541
m.js n=1024 len=10 source="string": 1847.099925407467
m.js n=1024 len=2048 source="string": 335.8807107007938
m.js n=1024 len=10 source="string-base64": 1628.3220387388276
m.js n=1024 len=2048 source="string-base64": 293.5605053750516
Current master:
m.js n=1024 len=10 source="array": 1105.7778268274712
m.js n=1024 len=2048 source="array": 262.61636308568154
m.js n=1024 len=10 source="arraybuffer": 1463.1770393435074
m.js n=1024 len=2048 source="arraybuffer": 1438.3519347799318
m.js n=1024 len=10 source="arraybuffer-middle": 1289.6443262079608
m.js n=1024 len=2048 source="arraybuffer-middle": 1465.8917985524079
m.js n=1024 len=10 source="buffer": 1154.3281747402962
m.js n=1024 len=2048 source="buffer": 714.9198271277221
m.js n=1024 len=10 source="uint8array": 1188.4816701911184
m.js n=1024 len=2048 source="uint8array": 312.9837225134168
m.js n=1024 len=10 source="string": 1021.2184715065113
m.js n=1024 len=2048 source="string": 415.72725788425214
m.js n=1024 len=10 source="string-base64": 746.3019883767605
m.js n=1024 len=2048 source="string-base64": 272.66673662077363
This PR:
m.js n=1024 len=10 source="array": 4360.13359704818
m.js n=1024 len=2048 source="array": 210.9601264473078
m.js n=1024 len=10 source="arraybuffer": 6096.237297560867
m.js n=1024 len=2048 source="arraybuffer": 6204.587073036097
m.js n=1024 len=10 source="arraybuffer-middle": 6931.443006081272
m.js n=1024 len=2048 source="arraybuffer-middle": 5500.454679186303
m.js n=1024 len=10 source="buffer": 3741.9594463683297
m.js n=1024 len=2048 source="buffer": 782.6012936368813
m.js n=1024 len=10 source="uint8array": 4495.957363782579
m.js n=1024 len=2048 source="uint8array": 188.2792093087443
m.js n=1024 len=10 source="string": 2287.6058600223178
m.js n=1024 len=2048 source="string": 461.90557855442626
m.js n=1024 len=10 source="string-base64": 2026.5954755341668
m.js n=1024 len=2048 source="string-base64": 298.25954061659525

[addaleax]

Yup, I can reproduce the benchmarking results.

Thanks and LGTM!

[jasnell]

LGTM

[targos]

CI: https://ci.nodejs.org/job/node-test-pull-request/4255/

[targos]

/cc @indutny@trevnorris

[trevnorris]

LGTM

[imyller]

LGTM

[indutny]

Ahaha! Nice

[indutny]

LGTM

[Fishrock123]

Wow, interesting.

[jbergstroem]

Wow, silly. LGTM 👍

[targos]

Landed in 558a884