performance:

Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
C++: 16 iterations in 3.00 seconds = 5.62 seconds per 30 iterations

ооо, если сильно захочется - побенчите ещё lz4 на нём https://github.com/inikep/lzbench

данные https://drive.google.com/file/d/0BwX7dtyRLxThenZpYU9zLTZhR1k/view?usp=sharing

lzbench -t16,16 -elz4 /tmp/tmpfs/silesia.tar

у меня на ноуте Core i7-6700HQ
powersave (ondemand нету)
C++: 27 iterations in 3.00 seconds = 3.33 seconds per 30 iterations
performance
C++: 34 iterations in 3.00 seconds = 2.65 seconds per 30 iterations

ну ясно, примерно в 2 раза

по пассмарку у 6700HQ single-thread rating=1893, а у 10900K - 3195

так что не, всё-таки не верю, что снап 865 = 2/3 коры

Ryzen 5 3600 - вообще
C++: 72 iterations in 3.00 seconds = 1.25 seconds per 30 iterations

это однопоток? Как-то много

однопоток

мб кеш 3го уровня так решает

почему много

а, 6700HQ это ноутбучный же

при governor performance не меняется

ага

тогда норм

Ещё RK3399 для сравнения:

ROCKPro64, пассивное охлаждение:
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
C++: 9 iterations in 3.00 seconds = 10.00 seconds per 30 iterations

NanoPC-T4, активное охлаждение:
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
Found 664579 prime numbers.
C++: 10 iterations in 3.00 seconds = 9.00 seconds per 30 iterations

ага, вполне адекватно

у меня ещё rock64 в принципе валяется, там RK3328, можно затестить, но он совсем тормоз

Snapdragon 850, performance:

~/lzbench$ ./lzbench -t16,16 -elz4 silesia.tar
lzbench 1.8 (64-bit Linux)   Assembled by P.Skibinski
Compressor name         Compress. Decompress. Compr. size  Ratio Filename
memcpy                   5767 MB/s  6658 MB/s   211947520 100.00 silesia.tar
lz4 1.9.2                 290 MB/s  1570 MB/s   100880800  47.60 silesia.tar
done... (cIters=1 dIters=1 cTime=16.0 dTime=16.0 chunkSize=1706MB cSpeed=0MB)

RK3399, активное охлаждение, performance:

~/lzbench$ ./lzbench -t16,16 -elz4  silesia.tar
lzbench 1.8 (64-bit Linux)   Assembled by P.Skibinski
Compressor name         Compress. Decompress. Compr. size  Ratio Filename
memcpy                   4047 MB/s  3811 MB/s   211947520 100.00 silesia.tar
lz4 1.9.2                 185 MB/s  1177 MB/s   100880800  47.60 silesia.tar
done... (cIters=1 dIters=1 cTime=16.0 dTime=16.0 chunkSize=1706MB cSpeed=0MB)