Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Wed, 15 Nov 2017, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 11:43 PM, Ingo Molnar wrote: > Again, try that "cat" example again, and time it. > > (And yes, I also know that "cpu*/cpufreq" is a symlink, but the direct > names are illogical, and no faster for me. Try it

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Wed, 15 Nov 2017, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 11:43 PM, Ingo Molnar wrote: > Again, try that "cat" example again, and time it. > > (And yes, I also know that "cpu*/cpufreq" is a symlink, but the direct > names are illogical, and no faster for me. Try it if you like: > >

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 11:43 PM, Ingo Molnar wrote: > > I also think that /proc/cpuinfo is a pretty bad interface for many uses - I > personally only very rarely need the cpuinfo of _all_ CPUs. > > We we should eventually have /proc/cpu/N/info or so, so that 99% of the times >

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 11:43 PM, Ingo Molnar wrote: > > I also think that /proc/cpuinfo is a pretty bad interface for many uses - I > personally only very rarely need the cpuinfo of _all_ CPUs. > > We we should eventually have /proc/cpu/N/info or so, so that 99% of the times > cpuinfo is needed

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, 14 Nov 2017, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner wrote: > > Current head + Raphaels patch: > > > > real0m0.029s > > user0m0.000s > > sys 0m0.010s > > > > So that patch is actually slower. > > Oh it definitely is

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, 14 Nov 2017, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner wrote: > > Current head + Raphaels patch: > > > > real0m0.029s > > user0m0.000s > > sys 0m0.010s > > > > So that patch is actually slower. > > Oh it definitely is expected to be slower,

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Wed, Nov 15, 2017 at 08:43:58AM +0100, Ingo Molnar wrote: > > * Rafael J. Wysocki wrote: > > > On Wednesday, November 15, 2017 1:06:12 AM CET Linus Torvalds wrote: > > > On Tue, Nov 14, 2017 at 4:04 PM, Linus Torvalds > > > wrote: > > > >

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Wed, Nov 15, 2017 at 08:43:58AM +0100, Ingo Molnar wrote: > > * Rafael J. Wysocki wrote: > > > On Wednesday, November 15, 2017 1:06:12 AM CET Linus Torvalds wrote: > > > On Tue, Nov 14, 2017 at 4:04 PM, Linus Torvalds > > > wrote: > > > > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner >

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

* Rafael J. Wysocki wrote: > On Wednesday, November 15, 2017 1:06:12 AM CET Linus Torvalds wrote: > > On Tue, Nov 14, 2017 at 4:04 PM, Linus Torvalds > > wrote: > > > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner > > >

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

* Rafael J. Wysocki wrote: > On Wednesday, November 15, 2017 1:06:12 AM CET Linus Torvalds wrote: > > On Tue, Nov 14, 2017 at 4:04 PM, Linus Torvalds > > wrote: > > > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner > > > wrote: > > >> Current head + Raphaels patch: > > >> > > >> real

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 4:30 PM, Rafael J. Wysocki wrote: > > OK, so may I queue it up? > > I don't think I can get that to work substantially faster anyway ... Yeah, I think it's ok. If somebody is really doing that /proc/cpuinfo in a loop, the rate limiting should kick in.

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 4:30 PM, Rafael J. Wysocki wrote: > > OK, so may I queue it up? > > I don't think I can get that to work substantially faster anyway ... Yeah, I think it's ok. If somebody is really doing that /proc/cpuinfo in a loop, the rate limiting should kick in. And if you do it

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Wednesday, November 15, 2017 1:06:12 AM CET Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 4:04 PM, Linus Torvalds > wrote: > > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner wrote: > >> Current head + Raphaels patch: > >> > >> real

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Wednesday, November 15, 2017 1:06:12 AM CET Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 4:04 PM, Linus Torvalds > wrote: > > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner wrote: > >> Current head + Raphaels patch: > >> > >> real0m0.029s > >> user0m0.000s > >> sys 0m0.010s >

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 4:04 PM, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner wrote: >> Current head + Raphaels patch: >> >> real0m0.029s >> user0m0.000s >> sys 0m0.010s >> >> So that patch is actually

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 4:04 PM, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner wrote: >> Current head + Raphaels patch: >> >> real0m0.029s >> user0m0.000s >> sys 0m0.010s >> >> So that patch is actually slower. > > Oh it definitely is expected to be slower,

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Wednesday, November 15, 2017 12:53:24 AM CET Thomas Gleixner wrote: > On Tue, 14 Nov 2017, Linus Torvalds wrote: > > > On Tue, Nov 14, 2017 at 2:47 PM, Rafael J. Wysocki > > wrote: > > > > > > So what about the one below? It works for me as expected. > > > > Can

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Wednesday, November 15, 2017 12:53:24 AM CET Thomas Gleixner wrote: > On Tue, 14 Nov 2017, Linus Torvalds wrote: > > > On Tue, Nov 14, 2017 at 2:47 PM, Rafael J. Wysocki > > wrote: > > > > > > So what about the one below? It works for me as expected. > > > > Can somebody with 100+ cores

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner wrote: > Current head + Raphaels patch: > > real0m0.029s > user0m0.000s > sys 0m0.010s > > So that patch is actually slower. Oh it definitely is expected to be slower, because it does the IPI to all the cores and

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner wrote: > Current head + Raphaels patch: > > real0m0.029s > user0m0.000s > sys 0m0.010s > > So that patch is actually slower. Oh it definitely is expected to be slower, because it does the IPI to all the cores and actually gets their

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, 14 Nov 2017, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 2:47 PM, Rafael J. Wysocki wrote: > > > > So what about the one below? It works for me as expected. > > Can somebody with 100+ cores test this? Ingo? You had a 160 core > machine that took forever, didn't

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, 14 Nov 2017, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 2:47 PM, Rafael J. Wysocki wrote: > > > > So what about the one below? It works for me as expected. > > Can somebody with 100+ cores test this? Ingo? You had a 160 core > machine that took forever, didn't you.. On a 144 CPUs

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 2:47 PM, Rafael J. Wysocki wrote: > > So what about the one below? It works for me as expected. Can somebody with 100+ cores test this? Ingo? You had a 160 core machine that took forever, didn't you.. Linus

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Tue, Nov 14, 2017 at 2:47 PM, Rafael J. Wysocki wrote: > > So what about the one below? It works for me as expected. Can somebody with 100+ cores test this? Ingo? You had a 160 core machine that took forever, didn't you.. Linus

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Saturday, November 11, 2017 12:09:18 AM CET Rafael J. Wysocki wrote: > On Fri, Nov 10, 2017 at 8:11 PM, Linus Torvalds > wrote: > > On Thu, Nov 9, 2017 at 2:30 PM, Rafael J. Wysocki wrote: > >> > >> c_start() can run aperfmperf_snapshot_khz()

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Saturday, November 11, 2017 12:09:18 AM CET Rafael J. Wysocki wrote: > On Fri, Nov 10, 2017 at 8:11 PM, Linus Torvalds > wrote: > > On Thu, Nov 9, 2017 at 2:30 PM, Rafael J. Wysocki wrote: > >> > >> c_start() can run aperfmperf_snapshot_khz() on all CPUs upfront (say > >> in parallel), then

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Fri, Nov 10, 2017 at 8:11 PM, Linus Torvalds wrote: > On Thu, Nov 9, 2017 at 2:30 PM, Rafael J. Wysocki wrote: >> >> c_start() can run aperfmperf_snapshot_khz() on all CPUs upfront (say >> in parallel), then wait for a while (say 5 ms; the

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Fri, Nov 10, 2017 at 8:11 PM, Linus Torvalds wrote: > On Thu, Nov 9, 2017 at 2:30 PM, Rafael J. Wysocki wrote: >> >> c_start() can run aperfmperf_snapshot_khz() on all CPUs upfront (say >> in parallel), then wait for a while (say 5 ms; the current 20 ms wait >> is overkill) and then

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Thu, Nov 9, 2017 at 2:30 PM, Rafael J. Wysocki wrote: > > c_start() can run aperfmperf_snapshot_khz() on all CPUs upfront (say > in parallel), then wait for a while (say 5 ms; the current 20 ms wait > is overkill) and then aperfmperf_snapshot_khz() can be run once on > each

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Thu, Nov 9, 2017 at 2:30 PM, Rafael J. Wysocki wrote: > > c_start() can run aperfmperf_snapshot_khz() on all CPUs upfront (say > in parallel), then wait for a while (say 5 ms; the current 20 ms wait > is overkill) and then aperfmperf_snapshot_khz() can be run once on > each CPU in

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On 11/10/17 at 08:25P, Ingo Molnar wrote: > > * Rafael J. Wysocki wrote: > > > Hi Linus, > > > > On 11/9/2017 11:38 AM, WANG Chao wrote: > > > Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused > > > a serious performance issue when reading

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On 11/10/17 at 08:25P, Ingo Molnar wrote: > > * Rafael J. Wysocki wrote: > > > Hi Linus, > > > > On 11/9/2017 11:38 AM, WANG Chao wrote: > > > Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused > > > a serious performance issue when reading from /proc/cpuinfo on system >

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

* Rafael J. Wysocki wrote: > Hi Linus, > > On 11/9/2017 11:38 AM, WANG Chao wrote: > > Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused > > a serious performance issue when reading from /proc/cpuinfo on system > > with aperfmperf. > > > >

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

* Rafael J. Wysocki wrote: > Hi Linus, > > On 11/9/2017 11:38 AM, WANG Chao wrote: > > Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused > > a serious performance issue when reading from /proc/cpuinfo on system > > with aperfmperf. > > > > For each cpu,

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On 11/10/17 at 12:04P, WANG Chao wrote: > On 11/10/17 at 01:06P, Rafael J. Wysocki wrote: > > On Thursday, November 9, 2017 11:30:54 PM CET Rafael J. Wysocki wrote: > > > On Thu, Nov 9, 2017 at 5:06 PM, Rafael J. Wysocki > > > wrote: > > > > Hi Linus, > > > > > > > >

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On 11/10/17 at 12:04P, WANG Chao wrote: > On 11/10/17 at 01:06P, Rafael J. Wysocki wrote: > > On Thursday, November 9, 2017 11:30:54 PM CET Rafael J. Wysocki wrote: > > > On Thu, Nov 9, 2017 at 5:06 PM, Rafael J. Wysocki > > > wrote: > > > > Hi Linus, > > > > > > > > On 11/9/2017 11:38 AM, WANG

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On 11/10/17 at 01:06P, Rafael J. Wysocki wrote: > On Thursday, November 9, 2017 11:30:54 PM CET Rafael J. Wysocki wrote: > > On Thu, Nov 9, 2017 at 5:06 PM, Rafael J. Wysocki > > wrote: > > > Hi Linus, > > > > > > On 11/9/2017 11:38 AM, WANG Chao wrote: > > >> > > >>

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On 11/10/17 at 01:06P, Rafael J. Wysocki wrote: > On Thursday, November 9, 2017 11:30:54 PM CET Rafael J. Wysocki wrote: > > On Thu, Nov 9, 2017 at 5:06 PM, Rafael J. Wysocki > > wrote: > > > Hi Linus, > > > > > > On 11/9/2017 11:38 AM, WANG Chao wrote: > > >> > > >> Commit 941f5f0f6ef5 (x86:

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Thursday, November 9, 2017 11:30:54 PM CET Rafael J. Wysocki wrote: > On Thu, Nov 9, 2017 at 5:06 PM, Rafael J. Wysocki > wrote: > > Hi Linus, > > > > On 11/9/2017 11:38 AM, WANG Chao wrote: > >> > >> Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo)

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Thursday, November 9, 2017 11:30:54 PM CET Rafael J. Wysocki wrote: > On Thu, Nov 9, 2017 at 5:06 PM, Rafael J. Wysocki > wrote: > > Hi Linus, > > > > On 11/9/2017 11:38 AM, WANG Chao wrote: > >> > >> Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused > >> a serious

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Thu, Nov 9, 2017 at 5:06 PM, Rafael J. Wysocki wrote: > Hi Linus, > > On 11/9/2017 11:38 AM, WANG Chao wrote: >> >> Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused >> a serious performance issue when reading from /proc/cpuinfo on system >>

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

On Thu, Nov 9, 2017 at 5:06 PM, Rafael J. Wysocki wrote: > Hi Linus, > > On 11/9/2017 11:38 AM, WANG Chao wrote: >> >> Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused >> a serious performance issue when reading from /proc/cpuinfo on system >> with aperfmperf. >> >> For

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

Hi Linus, On 11/9/2017 11:38 AM, WANG Chao wrote: Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused a serious performance issue when reading from /proc/cpuinfo on system with aperfmperf. For each cpu, arch_freq_get_on_cpu() sleeps 20ms to get its frequency. On a system

Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

Hi Linus, On 11/9/2017 11:38 AM, WANG Chao wrote: Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused a serious performance issue when reading from /proc/cpuinfo on system with aperfmperf. For each cpu, arch_freq_get_on_cpu() sleeps 20ms to get its frequency. On a system

[PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused a serious performance issue when reading from /proc/cpuinfo on system with aperfmperf. For each cpu, arch_freq_get_on_cpu() sleeps 20ms to get its frequency. On a system with 64 cpus, it takes 1.5s to finish running `cat

[PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused a serious performance issue when reading from /proc/cpuinfo on system with aperfmperf. For each cpu, arch_freq_get_on_cpu() sleeps 20ms to get its frequency. On a system with 64 cpus, it takes 1.5s to finish running `cat