Some AMD CPUs To Benefit From New Micro-Optimization In Linux 6.6

[phoronix]

Phoronix: Some AMD CPUs To Benefit From New Micro-Optimization In Linux 6.6

One of the patches to be picked up by the Linux 6.6 kernel this week brings back REP MOSQ for user-access on CPUs without Enhanced REP MOVSB (ERMS) support. In turn this can equate to some performance benefits on AMD CPUs lacking ERMS...

https://www.phoronix.com/news/Linux-6.6-Optimize-Lacking-ERMS

[Kjell]

Do I have to enable something to benefit from this on my 5950X?

[billyswong]

Do I have to enable something to benefit from this on my 5950X?

5950X still doesn't have that "ERMS" CPU feature?

[willmore]

3700X doesn't show it in /proc/cpuinfo FWIW.

[CHECK_12345]

I have Ryzen 7 1700 and did "cat /proc/cpuinfo | grep erms" and there was no output

When I do "perf top", and hide all non-kernel functions, I could see the "rep_movs_alternative" kernel function being used when playing audio

I guess this patch will give me a performance benefit on a Ryzen 7 1700!!!

I think this patch will not benefit Ryzen 5XXX and newer because they already have ERMS and FSRM

[DanglingPointer]

5800x ...

$ cat /proc/cpuinfo | grep erms
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif v_spec_ctrl umip pku ospke vaes vpclmulqdq rdpid overflow_recov succor smca fsrm

[EphemeralEft]

Slightly hurting performance on some higher end hardware in order to increase performance on some lower end hardware? That’s going to ruffle some feathers. I expect at least 2 protest forks. Maybe “Debtwo” with x86_64_v4.