[RX] randomx linux booster script

[psychocrypt]

Add booster script to enable huge pages and manipulate the caching
behavior.

[fireice-uk]

7 is a better number there.

[fireice-uk]

Non blocking comment - those numbers are correct only for gen1 zen cores. Some testing is needed as they will set collateral bits with unknown function on gen2.

[SChernykh]

Numbers for C001_1020 and C001_1021 registers were taken from R7 3700X (Geekbench performance bias option in BIOS). Funny to see how you copy my work without any clue of what it does or how it was obtained :D AMD doesn't change MSR registers (at least these ones) within same family CPUs, and all Zen, Zen+ and Zen2 CPUs so far are family 17h.

[psychocrypt]

Numbers for C001_1020 and C001_1021 registers were taken from R7 3700X (Geekbench performance bias option in BIOS).
AMD doesn't change MSR registers (at least these ones) within same family CPUs, and all Zen, Zen+ and Zen2 CPUs so far are family 17h.

+1, for the info

Funny to see how you copy my work without any clue of what it does or how it was obtained :D

please stop craying

[fireice-uk]

Funny to see how you copy my work without any clue of what it does or how it was obtained :D

Of course I know - you copied it from RagerX. That's why you have 2 gen2 MSRs and 2 gen1 MSRs. If you don't stop rocking your arse off, we will ban you from this repo.

[SChernykh]

OMG lol. I figured the last two values myself, and the first two were sent to me by xnbya who tested prefetcher options in BIOS on some other AM4 board (that I didn't have). Your RagerX is a joke. Have you even implemented MSR mod there yet?

[SChernykh]

FYI this is the screenshot of #monero-pools IRC from December 11th (everyone with IRC logs can check it): https://imgur.com/a/cE5XqHE
This where I got the info about the first 2 MSR registers and then I figured out the other two on my own, testing my BIOS.

[fireice-uk]

RagerX came out on the 9th -> see, it only works on people who are dumb enough not to fact-check you.

[SChernykh]

Yep, it came out on the 9th and it didn't have MSR mod back then because I tested it and I didn't see anything faster than I already had with xmrig on windows (without the mod). Maybe It has MSR mod now (of course), but I don't really care.

[SChernykh]

It's very convenient to claim you had something before XMRig without proof when everything RagerX is closed source and people have no way of checking it. You're a pathological liar.

[fireice-uk]

You have been hanging out with pathological liars a bit too much.

https://medium.com/@crypto_ryo/senior-nasa-engineer-calls-out-howard-chu-hyc-symas-monero-developer-on-being-a-fraud-db0b46688041

But, hey, if "Look! I got this magic undocumented number of some guy on IRC. I totally don't know where it came from!!" is your story, that's your story.

Follow up question - do you think we would be dumb enough to put all our improvements at once :P. You will need to start looking harder for dodgy guys on IRC to find stuff for you =]

[psychocrypt]

updated

[pjwelsh]

Not much of a dev, but have a suggestion on slight change to script. It includes an example (quoted )awk replacement to the cat|grep|cut chain to remove the false negative error on huge pages:
required_num_huge_pages=1280
num_huge_pages_total="$(awk '/HugePages_Total/ {print $2}' /proc/meminfo)"
num_huge_pages=$(cat /proc/meminfo | grep "HugePages_Free" | sed 's/ {2,}/ /g' | cut -d" " -f2)

if [ $num_huge_pages_total -ge $required_num_huge_pages ] ; then
echo "already active 2 MiB pages"
else
if [ $num_huge_pages -lt $required_num_huge_pages ] ; then
echo "active 2 MiB pages"
echo "execute: sysctl -w vm.nr_hugepages=$required_num_huge_pages"
sysctl -w vm.nr_hugepages="$required_num_huge_pages"
fi
# verify number of huge pages
num_huge_pages=$(cat /proc/meminfo | grep "HugePages_Free" | sed 's/ {2,}/ /g' | cut -d" " -f2)
num_memsets=$((num_huge_pages/required_num_huge_pages))

if [ $num_memsets -eq 0 ] ; then
    echo "Error: not enough 2 MiB pages $num_huge_pages/$required_num_huge_pages" >&2
fi

fi