I would be curious if disablement of Hyperthreading and/or Virtualization give you back some performance.
Twiddling smt doesn’t appear to have any effect.
Haven’t tried to trace the exact execution path causing the problem. After comparing the differences between the two machines (which are running the same rev of the same linux distro) and trying a couple other things, I compared /sys/devices/system/vulnerabilities/*. Noticed one was using several mitigations for Spectre V2, and so tried booting with mitigations=nospectre_v2. That appeared to eliminate the slowdown.
Having done that, I figured I could live with the slowdown (didn’t want to leave the mitigations disabled) and also figured it wasn’t something I was going to address in the T3 code I was writing, so stopped worrying about it.
Brought it up mostly as a thing to be aware of; because I assume it affects some number of random potential players out there. And suggesting they disable kernel protections for a widely-exploited family of vulnerabilities before playing your game doesn’t seem either practical or desirable.
I run Intel 10th gen CPU’s on my two towers. It seems that 10th-14th generation Intel chips had something or another wrong/slightly off with them. Everyone and their dog who remembered the floating point fiasco was happy to throw rocks about it. I’m an Intel guy like I am a Sony guy. I know AMD makes good stuff but for some reason I’m brand loyal to Intel. Hopefully it won’t give me heartburn.
perhaps recompiling with opportune GCC options can solve the issue ? (I admit that par. 3.19 of the GCC manual is, uh, less than clear, but perhaps some options can avoid generating vulnerable code)
Best regards from Italy,
dott. Piergiorgio.
Here’s a StackOverflow write up about it. I don’t use GCC and OP probably knows about it so this is more or less for anyone else.
I haven’t done a lot of experimentation with it, but I would expect compiling the interpreter with -mindirect-branch=thunk would help…but I don’t have any control over the player’s interpreter.
It is useful information to have as a player, though. I was just thinking about it more in terms of something to be aware of when writing a game. Because “cpu vulnerability mitigations might make your game run half as fast in some clients” wasn’t exactly on my bingo card before.
After being absent from the forum for a while, I did not have “TADS 3 slowdown on custom PCs” on my bingo card. Being blind and unable to build my own PC, let alone having the cash or assistance available to do it, this was never on my radar. However, it also makes me wonder how screen reader performance (especially JAWS, which is much more of a performance hog than NVDA) would or could increase this minor slowdown. Of course, it’s one of those things that I, and only I, as a blind techie with a concern for the decrease of performance caused by Electron and web apps as desktop apps in the same vein (especially as a screen reader user), and recently the introduction of AI into almost every facet of computing (or at least, the attempt to, via Microsoft, Apple and their ilk, and even the blind community which has bought into AI hard), will ever care about.