Make AARCH64 harder, better, faster, stronger

- Perform some housekeeping on scalar math function code
- Import ARM's Optimized Routines for SIMD string processing
- Upgrade to latest Chromium zlib and enable more SIMD optimizations
This commit is contained in:
Justine Tunney 2023-05-15 01:51:29 -07:00
parent 550b52abf6
commit cc1732bc42
No known key found for this signature in database
GPG key ID: BE714B4575D6E328
143 changed files with 15661 additions and 1329 deletions

View file

@ -12,8 +12,8 @@
#include "libc/errno.h"
#include "libc/fmt/conv.h"
#include "libc/log/check.h"
#include "libc/mem/mem.h"
#include "libc/mem/gc.internal.h"
#include "libc/mem/mem.h"
#include "libc/runtime/runtime.h"
#include "libc/stdio/stdio.h"
#include "libc/str/str.h"
@ -48,26 +48,62 @@ FLAGS\n\
// clang-format off
// make -j8 o//examples && dd if=/dev/urandom count=100 | tee a | o//examples/compress.com | o//examples/decompress.com >b && sha1sum a b
/*
#!/bin/bash
# data file is o/dbg/third_party/python/python.com
# level 0 147517 compress 495 MB/s decompress 1.4 GB/s
# level 1 80274 compress 29.2 MB/s decompress 303 MB/s
# level 2 79384 compress 33.8 MB/s decompress 212 MB/s
# level 3 78875 compress 28.9 MB/s decompress 224 MB/s
# level 4 78010 compress 27.1 MB/s decompress 319 MB/s <-- sweet spot?
# level 5 77107 compress 19.5 MB/s decompress 273 MB/s
# level 6 75081 compress 10.0 MB/s decompress 99.3 MB/s
# level 7 75022 compress 7.5 MB/s decompress 287 MB/s
# level 8 75016 compress 5.4 MB/s decompress 109 MB/s
# level 9 75016 compress 5.4 MB/s decompress 344 MB/s
# level 1 348739 compress 22.8 MB/s decompress 444 MB/s
# level 2 347549 compress 37.8 MB/s decompress 457 MB/s
# level 3 346902 compress 33.3 MB/s decompress 463 MB/s
# level 4 345671 compress 29.3 MB/s decompress 467 MB/s
# level 5 344392 compress 22.4 MB/s decompress 506 MB/s
# level 6 342105 compress 10.9 MB/s decompress 516 MB/s
# level 7 342046 compress 7.9 MB/s decompress 515 MB/s
# level 8 342009 compress 5.8 MB/s decompress 518 MB/s
# level 9 342001 compress 5.7 MB/s decompress 524 MB/s
# level F 1 362426 compress 48.2 MB/s decompress 488 MB/s
# level F 2 360875 compress 42.7 MB/s decompress 484 MB/s
# level F 3 359992 compress 37.1 MB/s decompress 499 MB/s
# level F 4 358460 compress 32.9 MB/s decompress 503 MB/s
# level F 5 356431 compress 24.0 MB/s decompress 547 MB/s
# level F 6 352274 compress 11.6 MB/s decompress 558 MB/s
# level F 7 352155 compress 8.7 MB/s decompress 554 MB/s
# level F 8 352065 compress 6.3 MB/s decompress 554 MB/s
# level F 9 352051 compress 6.2 MB/s decompress 556 MB/s
# level L 1 348739 compress 41.1 MB/s decompress 446 MB/s
# level L 2 347549 compress 37.4 MB/s decompress 443 MB/s
# level L 3 346902 compress 32.3 MB/s decompress 462 MB/s
# level L 4 351932 compress 28.8 MB/s decompress 511 MB/s
# level L 5 351384 compress 23.6 MB/s decompress 520 MB/s
# level L 6 351328 compress 12.1 MB/s decompress 522 MB/s
# level L 7 351230 compress 7.3 MB/s decompress 518 MB/s
# level L 8 351192 compress 5.7 MB/s decompress 522 MB/s
# level L 9 351182 compress 6.5 MB/s decompress 519 MB/s
# level R 1 388209 compress 83.1 MB/s decompress 371 MB/s
# level R 2 388209 compress 82.3 MB/s decompress 362 MB/s
# level R 3 388209 compress 81.8 MB/s decompress 361 MB/s
# level R 4 388209 compress 81.7 MB/s decompress 364 MB/s
# level R 5 388209 compress 81.7 MB/s decompress 363 MB/s
# level R 6 388209 compress 80.1 MB/s decompress 359 MB/s
# level R 7 388209 compress 80.3 MB/s decompress 354 MB/s
# level R 8 388209 compress 80.3 MB/s decompress 363 MB/s
# level R 9 388209 compress 81.3 MB/s decompress 364 MB/s
# level H 1 390207 compress 87.6 MB/s decompress 371 MB/s
# level H 2 390207 compress 87.5 MB/s decompress 372 MB/s
# level H 3 390207 compress 85.5 MB/s decompress 364 MB/s
# level H 4 390207 compress 87.3 MB/s decompress 375 MB/s
# level H 5 390207 compress 89.0 MB/s decompress 373 MB/s
# level H 6 390207 compress 87.3 MB/s decompress 372 MB/s
# level H 7 390207 compress 87.0 MB/s decompress 368 MB/s
# level H 8 390207 compress 86.2 MB/s decompress 367 MB/s
# level H 9 390207 compress 86.9 MB/s decompress 369 MB/s
m=
make -j8 MODE=$m o/$m/examples || exit
for strategy in ' ' F L R H; do
for level in $(seq 1 9); do
for strategy in F L R H; do
o/$m/examples/compress.com -$strategy$level <o/dbg/third_party/python/python.com | dd count=10000 2>/tmp/info >/tmp/comp
o/$m/examples/compress.com -$level$strategy <o/dbg/third_party/python/python.com | dd count=10000 2>/tmp/info >/tmp/comp
compspeed=$(grep -Po '[.\d]+ \w+/s' /tmp/info)
o/$m/examples/decompress.com </tmp/comp | dd count=10000 2>/tmp/info >/dev/null
decompspeed=$(grep -Po '[.\d]+ \w+/s' /tmp/info)
size=$(o/$m/examples/compress.com -$strategy$level <o/$m/examples/compress.com | wc -c)
size=$(o/$m/examples/compress.com -$level$strategy <o/$m/examples/compress.com | wc -c)
echo "level $strategy $level $size compress $compspeed decompress $decompspeed"
done
done