* Performance difference in mainline and downstream
@ 2021-04-30 7:49 Mighty M
2021-05-03 10:19 ` Tony Lindgren
0 siblings, 1 reply; 6+ messages in thread
From: Mighty M @ 2021-04-30 7:49 UTC (permalink / raw)
To: linux-omap
I have a Samsung Galaxy Tab 2 which has an OMAP 4430 clocked at 1Ghz,
with that there is quite some difference in performance when comparing
the mainline and downstream kernels. To confirm my suspicions I ran
UnixBench(https://github.com/kdlucas/byte-unixbench) on both of them,
the results are as follows -
1) Downstream (running SailfishOS) - Kernel Source -
https://github.com/CyanogenMod/android_kernel_samsung_espresso10, I
get a result of 580.4 in dual-core and 292.9 in single core. I
compiled UnixBench manually.
2) Mainline (running PostMarketOS) - Kernel Source -
https://github.com/MightyM17/linux_openpvrsgx/tree/letux-pvrsrvkm-5.12-rc1,
when using UnixBench package from PostMarketOS repo, I get a score of
352.7 dual core and 192.7 single core, and compiling it myself and
then running gives the score 415.1 dual core and 209.9 single core.
The downstream scores are quite high even beating Droid 4 which has
CPU clocked at 1.2Ghz when Tab 2 has it at 1 GHz.
Why is there so much difference in mainline and downstream and what
can be done to improve the performance.
Regards.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Performance difference in mainline and downstream
2021-04-30 7:49 Performance difference in mainline and downstream Mighty M
@ 2021-05-03 10:19 ` Tony Lindgren
2021-05-04 6:02 ` Mighty M
0 siblings, 1 reply; 6+ messages in thread
From: Tony Lindgren @ 2021-05-03 10:19 UTC (permalink / raw)
To: Mighty M; +Cc: linux-omap
Hi,
* Mighty M <mightymb17@gmail.com> [210430 07:49]:
> I have a Samsung Galaxy Tab 2 which has an OMAP 4430 clocked at 1Ghz,
> with that there is quite some difference in performance when comparing
> the mainline and downstream kernels. To confirm my suspicions I ran
> UnixBench(https://github.com/kdlucas/byte-unixbench) on both of them,
> the results are as follows -
> 1) Downstream (running SailfishOS) - Kernel Source -
> https://github.com/CyanogenMod/android_kernel_samsung_espresso10, I
> get a result of 580.4 in dual-core and 292.9 in single core. I
> compiled UnixBench manually.
> 2) Mainline (running PostMarketOS) - Kernel Source -
> https://github.com/MightyM17/linux_openpvrsgx/tree/letux-pvrsrvkm-5.12-rc1,
> when using UnixBench package from PostMarketOS repo, I get a score of
> 352.7 dual core and 192.7 single core, and compiling it myself and
> then running gives the score 415.1 dual core and 209.9 single core.
>
> The downstream scores are quite high even beating Droid 4 which has
> CPU clocked at 1.2Ghz when Tab 2 has it at 1 GHz.
>
> Why is there so much difference in mainline and downstream and what
> can be done to improve the performance.
If you have a good test case, it would be nice to narrow down the
performance issue and fix it. My guess is that the following might
help with the mainline Linux kernel built with omap2plus_defconfig:
- Disable CONFIG_DEBUG in .config
- Disable CONFIG_ARCH_OMAP2 in .config to disable the build options
for ARMv6
If the above does not help, then let's look further in the L2
settings for example.
Regards,
Tony
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Performance difference in mainline and downstream
2021-05-03 10:19 ` Tony Lindgren
@ 2021-05-04 6:02 ` Mighty M
2021-05-04 12:50 ` Tony Lindgren
0 siblings, 1 reply; 6+ messages in thread
From: Mighty M @ 2021-05-04 6:02 UTC (permalink / raw)
To: Tony Lindgren; +Cc: linux-omap
On Mon, May 3, 2021 at 3:49 PM Tony Lindgren <tony@atomide.com> wrote:
>
> Hi,
>
> * Mighty M <mightymb17@gmail.com> [210430 07:49]:
> > I have a Samsung Galaxy Tab 2 which has an OMAP 4430 clocked at 1Ghz,
> > with that there is quite some difference in performance when comparing
> > the mainline and downstream kernels. To confirm my suspicions I ran
> > UnixBench(https://github.com/kdlucas/byte-unixbench) on both of them,
> > the results are as follows -
> > 1) Downstream (running SailfishOS) - Kernel Source -
> > https://github.com/CyanogenMod/android_kernel_samsung_espresso10, I
> > get a result of 580.4 in dual-core and 292.9 in single core. I
> > compiled UnixBench manually.
> > 2) Mainline (running PostMarketOS) - Kernel Source -
> > https://github.com/MightyM17/linux_openpvrsgx/tree/letux-pvrsrvkm-5.12-rc1,
> > when using UnixBench package from PostMarketOS repo, I get a score of
> > 352.7 dual core and 192.7 single core, and compiling it myself and
> > then running gives the score 415.1 dual core and 209.9 single core.
> >
> > The downstream scores are quite high even beating Droid 4 which has
> > CPU clocked at 1.2Ghz when Tab 2 has it at 1 GHz.
> >
> > Why is there so much difference in mainline and downstream and what
> > can be done to improve the performance.
>
> If you have a good test case, it would be nice to narrow down the
> performance issue and fix it. My guess is that the following might
> help with the mainline Linux kernel built with omap2plus_defconfig:
>
> - Disable CONFIG_DEBUG in .config
>
> - Disable CONFIG_ARCH_OMAP2 in .config to disable the build options
> for ARMv6
>
> If the above does not help, then let's look further in the L2
> settings for example.
>
> Regards,
>
> Tony
>
>
Hello, CONFIG_ARCH_OMAP2 is already disabled in my .config and I
couldnt find CONFIG_DEBUG.
https://pastebin.ubuntu.com/p/qKYpCPCk5Q/ here is my .config file.
Anyways it seems we need to look further now.
Thanks!
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Performance difference in mainline and downstream
2021-05-04 6:02 ` Mighty M
@ 2021-05-04 12:50 ` Tony Lindgren
2021-05-04 16:48 ` Dev Null
0 siblings, 1 reply; 6+ messages in thread
From: Tony Lindgren @ 2021-05-04 12:50 UTC (permalink / raw)
To: Mighty M; +Cc: linux-omap
* Mighty M <mightymb17@gmail.com> [210504 06:02]:
> Hello, CONFIG_ARCH_OMAP2 is already disabled in my .config and I
> couldnt find CONFIG_DEBUG.
Sorry I meant CONFIG_DEBUG_KERNEL :)
Try first dropping these lines from your .config:
CONFIG_EXPERT=y
CONFIG_DEBUG_KERNEL=y
Regards,
Tony
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Performance difference in mainline and downstream
2021-05-04 12:50 ` Tony Lindgren
@ 2021-05-04 16:48 ` Dev Null
2021-05-05 12:04 ` Mighty M
0 siblings, 1 reply; 6+ messages in thread
From: Dev Null @ 2021-05-04 16:48 UTC (permalink / raw)
To: Tony Lindgren; +Cc: Mighty M, linux-omap
armv7h in general also has some of the secuity mitigations around
speculative execution / branch prediction.
no clue what the cumulative inpact of these is on omap4 but the vendor
kernel very likely dosent have any
On Tue, 4 May 2021 15:50:37 +0300
Tony Lindgren <tony@atomide.com> wrote:
> * Mighty M <mightymb17@gmail.com> [210504 06:02]:
> > Hello, CONFIG_ARCH_OMAP2 is already disabled in my .config and I
> > couldnt find CONFIG_DEBUG.
>
> Sorry I meant CONFIG_DEBUG_KERNEL :)
>
> Try first dropping these lines from your .config:
>
> CONFIG_EXPERT=y
> CONFIG_DEBUG_KERNEL=y
>
> Regards,
>
> Tony
--
Dev Null <devnull@uvos.xyz>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Performance difference in mainline and downstream
2021-05-04 16:48 ` Dev Null
@ 2021-05-05 12:04 ` Mighty M
0 siblings, 0 replies; 6+ messages in thread
From: Mighty M @ 2021-05-05 12:04 UTC (permalink / raw)
To: Dev Null; +Cc: Tony Lindgren, linux-omap
On Tue, May 4, 2021 at 10:18 PM Dev Null <devnull@uvos.xyz> wrote:
>
> armv7h in general also has some of the secuity mitigations around
> speculative execution / branch prediction.
>
> no clue what the cumulative inpact of these is on omap4 but the vendor
> kernel very likely dosent have any
>
> On Tue, 4 May 2021 15:50:37 +0300
> Tony Lindgren <tony@atomide.com> wrote:
>
> > * Mighty M <mightymb17@gmail.com> [210504 06:02]:
> > > Hello, CONFIG_ARCH_OMAP2 is already disabled in my .config and I
> > > couldnt find CONFIG_DEBUG.
> >
> > Sorry I meant CONFIG_DEBUG_KERNEL :)
> >
> > Try first dropping these lines from your .config:
> >
> > CONFIG_EXPERT=y
> > CONFIG_DEBUG_KERNEL=y
> >
> > Regards,
> >
> > Tony
>
>
> --
> Dev Null <devnull@uvos.xyz>
Well that might be a proper reason, also its running on 1GB /system
partition which might be a factor as well
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-05-05 12:04 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-04-30 7:49 Performance difference in mainline and downstream Mighty M
2021-05-03 10:19 ` Tony Lindgren
2021-05-04 6:02 ` Mighty M
2021-05-04 12:50 ` Tony Lindgren
2021-05-04 16:48 ` Dev Null
2021-05-05 12:04 ` Mighty M
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).