Better resolution using the hrtimers infrastructure
Hello,
I've been experimenting with the high-resolution timer subsystem on the x86 platform (specifically, P4 2.8 GHz) running Linux 2.6.16.28. (LAPIC and IOAPIC turned on, pre-emptible kernel, HZ=250)
I wrote a small app to create a POSIX timer (timer_create(), timer_settime(), etc) that fires with a given period. The scheduling policy is set to SCHED_RR. After some time, the process writes the distribution of "elapsed time between signals" to a file, and exits. I then post-process this file to output average time between signals, standard deviation, occurences +/- 5 µs and occurences +/- 10 µs.
I first compiled a kernel with a resolution of 5000 ns. I ran 10 concurrent tasks with T = 500+13*i µs RUNTIME = 40000 s
AVG=500.0000119875 STDDEV=4.7221250605
P(495 <= L <= 505) = 0.8177218875
P(490 <= L <= 510) = 0.9630963875
AVG=512.9983118837 STDDEV=4.6444950150
P(508 <= L <= 518) = 0.8248110095
P(503 <= L <= 523) = 0.9648220375
AVG=525.9978541435 STDDEV=4.5872769426
P(521 <= L <= 531) = 0.8294680508
P(516 <= L <= 536) = 0.9662504459
AVG=538.9976678278 STDDEV=4.6550001682
P(534 <= L <= 544) = 0.8233726222
P(529 <= L <= 549) = 0.9647691136
AVG=551.9962816176 STDDEV=4.6202416645
P(547 <= L <= 557) = 0.8261095807
P(542 <= L <= 562) = 0.9654630987
AVG=564.9992072061 STDDEV=4.7382868551
P(560 <= L <= 570) = 0.8201386764
P(555 <= L <= 575) = 0.9631647119
AVG=577.9969566710 STDDEV=4.6799294742
P(573 <= L <= 583) = 0.8199662908
P(568 <= L <= 588) = 0.9637724766
AVG=591.0001112558 STDDEV=4.6366848287
P(586 <= L <= 596) = 0.8273179260
P(581 <= L <= 601) = 0.9651421699
AVG=604.0009058943 STDDEV=4.7022062409
P(599 <= L <= 609) = 0.8253007146
P(594 <= L <= 614) = 0.9627097645
AVG=617.0019377040 STDDEV=4.7598557221
P(612 <= L <= 622) = 0.8164446084
P(607 <= L <= 627) = 0.9616794710
I then recompiled my kernel with a resolution of 800 ns. I ran 10 concurrent tasks with T = 600+13*i µs RUNTIME = 50000 s
AVG=599.9986703520 STDDEV=4.3792286239
P(595 <= L <= 605) = 0.8466944914
P(590 <= L <= 610) = 0.9707216439
AVG=612.9981400476 STDDEV=4.1769930447
P(608 <= L <= 618) = 0.8585605458
P(603 <= L <= 623) = 0.9756196900
AVG=625.9983042536 STDDEV=4.2016958416
P(621 <= L <= 631) = 0.8569019831
P(616 <= L <= 636) = 0.9753284384
AVG=638.9985581093 STDDEV=4.2072766644
P(634 <= L <= 644) = 0.8532895995
P(629 <= L <= 649) = 0.9746377576
AVG=651.9989919428 STDDEV=4.1936623871
P(647 <= L <= 657) = 0.8576286791
P(642 <= L <= 662) = 0.9757251792
AVG=664.9976003076 STDDEV=4.2224641203
P(660 <= L <= 670) = 0.8541156365
P(655 <= L <= 675) = 0.9748352293
AVG=677.9992813471 STDDEV=4.2120672530
P(673 <= L <= 683) = 0.8598737385
P(668 <= L <= 688) = 0.9752509658
AVG=690.9987525239 STDDEV=4.2379530226
P(686 <= L <= 696) = 0.8548963154
P(681 <= L <= 701) = 0.9744069907
AVG=703.9993729190 STDDEV=4.3323158412
P(699 <= L <= 709) = 0.8449093344
P(694 <= L <= 714) = 0.9731234341
AVG=717.0005565641 STDDEV=4.2230279084
P(712 <= L <= 722) = 0.8579009228
P(707 <= L <= 727) = 0.9744286679
The resolution (STDDEV) was slightly better, but I un-scientifically changed two parameters. (I'll have to run more tests.)
Finally, I ran only one task with T = 600 µs
AVG=600.0036125280 STDDEV=1.6630613701
P(595 <= L <= 605) = 0.9997573360
P(590 <= L <= 610) = 0.9998031160
The resolution is markedly better (unsurprisingly).
Is it possible to improve the resolution? Would an HPET-enabled motherboard help? The source code seems to suggest the time stamp counter is used over any HPET. I've read that IBM provides high-quality HPET, whereas some southbridge manufacturers barely meet the Intel spec. Would other hardware (PCI board? PCI-Express board?) help in improving the resolution?
Thanks for reading this far.
Regards.
Quoted 4 lines · John
I've been experimenting with the high-resolution timer subsystem on the x86 platform (specifically, P4 2.8 GHz) running Linux 2.6.16.28. (LAPIC and IOAPIC turned on, pre-emptible kernel, HZ=250)
Quoted 8 lines
I wrote a small app to create a POSIX timer (
timer_create(),timer_settime(), etc) that fires with a given period. The scheduling policy is set toSCHED_RR. After some time, the process writes the distribution of "elapsed time between signals" to a file, and exits. I then post-process this file to output average time between signals, standard deviation, occurences +/- 5 µs and occurences +/- 10 µs.[...]
Given the dearth of replies, I assume that linux-kernel is not the appropriate venue for this type of question.
Could someone point to a better forum?
Regards.
LKML receives hundreds of emails a day (I go two days without checking, and I have over 1000 unread messages). So don't get turned off, if your email is not replied to so quickly or at all. Just try again, or better yet CC people who work on stuff related to your question. I added those that I know about right off hand but I could have missed a few.
-- Steve
Quoted 127 lines · John
Hello,
I've been experimenting with the high-resolution timer subsystem on the x86 platform (specifically, P4 2.8 GHz) running Linux 2.6.16.28. (LAPIC and IOAPIC turned on, pre-emptible kernel, HZ=250)
I wrote a small app to create a POSIX timer (
timer_create(),timer_settime(), etc) that fires with a given period. The scheduling policy is set toSCHED_RR. After some time, the process writes the distribution of "elapsed time between signals" to a file, and exits. I then post-process this file to output average time between signals, standard deviation, occurences +/- 5 µs and occurences +/- 10 µs.I first compiled a kernel with a resolution of 5000 ns. I ran 10 concurrent tasks with T = 500+13*i µs RUNTIME = 40000 s
AVG=500.0000119875 STDDEV=4.7221250605
P(495 <= L <= 505) = 0.8177218875
P(490 <= L <= 510) = 0.9630963875AVG=512.9983118837 STDDEV=4.6444950150
P(508 <= L <= 518) = 0.8248110095
P(503 <= L <= 523) = 0.9648220375AVG=525.9978541435 STDDEV=4.5872769426
P(521 <= L <= 531) = 0.8294680508
P(516 <= L <= 536) = 0.9662504459AVG=538.9976678278 STDDEV=4.6550001682
P(534 <= L <= 544) = 0.8233726222
P(529 <= L <= 549) = 0.9647691136AVG=551.9962816176 STDDEV=4.6202416645
P(547 <= L <= 557) = 0.8261095807
P(542 <= L <= 562) = 0.9654630987AVG=564.9992072061 STDDEV=4.7382868551
P(560 <= L <= 570) = 0.8201386764
P(555 <= L <= 575) = 0.9631647119AVG=577.9969566710 STDDEV=4.6799294742
P(573 <= L <= 583) = 0.8199662908
P(568 <= L <= 588) = 0.9637724766AVG=591.0001112558 STDDEV=4.6366848287
P(586 <= L <= 596) = 0.8273179260
P(581 <= L <= 601) = 0.9651421699AVG=604.0009058943 STDDEV=4.7022062409
P(599 <= L <= 609) = 0.8253007146
P(594 <= L <= 614) = 0.9627097645AVG=617.0019377040 STDDEV=4.7598557221
P(612 <= L <= 622) = 0.8164446084
P(607 <= L <= 627) = 0.9616794710I then recompiled my kernel with a resolution of 800 ns. I ran 10 concurrent tasks with T = 600+13*i µs RUNTIME = 50000 s
AVG=599.9986703520 STDDEV=4.3792286239
P(595 <= L <= 605) = 0.8466944914
P(590 <= L <= 610) = 0.9707216439AVG=612.9981400476 STDDEV=4.1769930447
P(608 <= L <= 618) = 0.8585605458
P(603 <= L <= 623) = 0.9756196900AVG=625.9983042536 STDDEV=4.2016958416
P(621 <= L <= 631) = 0.8569019831
P(616 <= L <= 636) = 0.9753284384AVG=638.9985581093 STDDEV=4.2072766644
P(634 <= L <= 644) = 0.8532895995
P(629 <= L <= 649) = 0.9746377576AVG=651.9989919428 STDDEV=4.1936623871
P(647 <= L <= 657) = 0.8576286791
P(642 <= L <= 662) = 0.9757251792AVG=664.9976003076 STDDEV=4.2224641203
P(660 <= L <= 670) = 0.8541156365
P(655 <= L <= 675) = 0.9748352293AVG=677.9992813471 STDDEV=4.2120672530
P(673 <= L <= 683) = 0.8598737385
P(668 <= L <= 688) = 0.9752509658AVG=690.9987525239 STDDEV=4.2379530226
P(686 <= L <= 696) = 0.8548963154
P(681 <= L <= 701) = 0.9744069907AVG=703.9993729190 STDDEV=4.3323158412
P(699 <= L <= 709) = 0.8449093344
P(694 <= L <= 714) = 0.9731234341AVG=717.0005565641 STDDEV=4.2230279084
P(712 <= L <= 722) = 0.8579009228
P(707 <= L <= 727) = 0.9744286679The resolution (STDDEV) was slightly better, but I un-scientifically changed two parameters. (I'll have to run more tests.)
Finally, I ran only one task with T = 600 µs
AVG=600.0036125280 STDDEV=1.6630613701
P(595 <= L <= 605) = 0.9997573360
P(590 <= L <= 610) = 0.9998031160The resolution is markedly better (unsurprisingly).
Is it possible to improve the resolution? Would an HPET-enabled motherboard help? The source code seems to suggest the time stamp counter is used over any HPET. I've read that IBM provides high-quality HPET, whereas some southbridge manufacturers barely meet the Intel spec. Would other hardware (PCI board? PCI-Express board?) help in improving the resolution?
Thanks for reading this far.
Regards.
-