3 ms·
As of today still most of the software I use seems to get stuck in one core of my CPU when calculating stuff so I'd argue that by enabling HT I'm losing perform
by azahk 8y ago
As of today still most of the software I use seems to get stuck in one core of my CPU when calculating stuff so I'd argue that by enabling HT I'm losing performance.
- vbezhenar 8y agoThat's not really clear from the article, but I suspect that performance drop is from multithreaded software. And sure, if you are observing performance degradation with hyper-threading, there's 0 reason to keep it enabled. OpenBSD guys suggest to turn it off after all.
- hermanradtke 8y agoThe article claims a 30% perf slowdown with PHP. The majority of PHP workloads are single threaded. That causes me to suspect this impacts single core workloads too.
- bonzini 8y agoYes, on some processors basically it's as if all userspace was recompiled to use retpolines. Enabling STIBP promises that one thread cannot influence the indirect branch predictor's operation when the sibling runs; on those processors the indirect branch predictor is completely disabled, which is an inefficient but valid way to implement STIBP semantics.
- monocasa 8y agoThat's interesting. 30% is generally the back of the envelope number for what hyperthreads get you in a OoOE core.