Helping thread :)

[doofoo24]

i done a couple of test using only 4x4:lzma with different setting '4x4:lzma:a1:mfbt4:d64m:fb273:mc10000:lc8' '4x4:lzma:a1:mfbt4:d128m:fb273:mc10000:lc8' '4x4:lzma:a1:mfbt4:d200m:fb273:mc10000:lc8'...
i found that using 4 core 8 threads any setting other than d:64 only 40% cpu utilized ?
using 4x4:lzma:a1:mfbt4:d64m:fb273:mc10000:lc8 it seem to me the best option in scaling cpu with 2 core only only use 384mb in decompression and use the 2 core to tha max...
using on 2 core with 4 threads and 4 core without hyperthreading use the same 756mb of memory and in 4 core 8 threads it use 1024mb and the cpu 100%...
when try to use 4x4:lzma:a1:mfbt4:d128m:fb273:mc10000:lc8 or 4x4:lzma:a1:mfbt4:d200m:fb273:mc10000:lc8 the cpu usage on 4 core 8 threads only 40% whille on 2 core 4 threads and 4 core without hyperthreading cpu usage is 85%? (in decompression not compression)
i am not an expert so no need for smart remarks.
so the best scaling in decompression (in compression on i7 it use around 57% of cpu with 1684mb of memory) is to use 4x4:lzma:a1:mfbt4:d64m:fb273:mc10000:lc8 it work ( i mean that it use the cpu 100% on all setting like pentium-i3-i5-i7)
if anyone find it useful great.
*using 4x4:t does not work for me in decompression i tried 4x4:t4 but when decompress it use 14% of i7 not 50% of i7 4 core 8 threads...

summary
if you use pzlib or precompmt use it with this
4x4:lzma:a1:mfbt4:d64m:fb273:mc10000:lc8 is super fast in decompression but it use the cpu to the max and require more memory but modern system use 64bit so no problem there, for old system and limited ram don't use 4x4 use lzma:a1:mfbt4:d185m:fb273:mc10000:lc8

if srep only use it with zstd the new version 1.1.2

add this line to arc.ini
[External compressor:zstd]
header = 0
packcmd = zstd -22 --ultra <stdin> <stdout>

in compression it use around limit core and only 769mb in compression
in decompression zstd alone use only 128mb and srep better to limit it in inno to 128mb so the total memory use is 256mb it super fast and less demanding on the cpu...