Discussion:
Feedback on benchmarking results of SDHCI and MMCCAM driver
(too old to reply)
Udit agarwal
2018-05-14 21:18:43 UTC
Permalink
Hi,
I have conducted several sd card benchmarking tests on FreeBSD 11.1, with
both SDHCI and the newly incorporated MMCCAM driver to compare their
performance under different conditions.
I have compiled the plots and raw data in this spreadsheet.
<https://docs.google.com/spreadsheets/d/1_lf9S136z0tJyni9W3t1__Rlrkal1fd6L7-vnlmYQpc/edit?usp=sharing>,

Please provide some feedback on the above results.

Regards,
Udit agarwal
Udit agarwal
2018-05-14 21:51:00 UTC
Permalink
Post by Udit agarwal
I have conducted several sd card benchmarking tests on FreeBSD 11.1, with
both SDHCI and the newly incorporated MMCCAM driver to compare their
performance under different conditions.
1) How reproduceable are the results? Have you performed the test
multiple times
to check for variance?
Yes, i have carried these tests multiple time and the results are pretty
consistent upto 0.1MBPS.
2) How did you prepare the SD card prior to each write test? Flash write
tests
are extremely sensitive to hidden move/rewrite and erase cycles performed
by the
flash translation layer.
Before each test, I freshly recreated the filesystem to remove any trace of
the previous tests. Moreover, I have mounted/unmounted the device to purge
any data in the cache.
3) Have you tried different SD cards from different manufacturers to see
if the
drivers happen to interact well or badly with the specific flash
implementation.
No, currently i havn't experimented with different sd cards. True, it would
be interesting to figure out the interaction of the driver with different
flash implementations. I'll perform these tests ASAP.

Regards,
Udit
--
Peter Jeremy
Russell Haley
2018-05-14 23:18:29 UTC
Permalink
Post by Udit agarwal
Post by Peter Jeremy
Post by Udit agarwal
I have conducted several sd card benchmarking tests on FreeBSD 11.1,
with
Post by Peter Jeremy
Post by Udit agarwal
both SDHCI and the newly incorporated MMCCAM driver to compare their
performance under different conditions.
1) How reproduceable are the results? Have you performed the test multiple times
to check for variance?
The images that Udit is using for his tests can be made available for
verification if that's desirable (we have a webserver available). His blogs
are very good as well.
Post by Udit agarwal
Yes, i have carried these tests multiple time and the results are pretty
consistent upto 0.1MBPS.
Post by Peter Jeremy
2) How did you prepare the SD card prior to each write test? Flash write tests
are extremely sensitive to hidden move/rewrite and erase cycles performed by the
flash translation layer.
Before each test, I freshly recreated the filesystem to remove any trace of
the previous tests. Moreover, I have mounted/unmounted the device to purge
any data in the cache.
Post by Peter Jeremy
3) Have you tried different SD cards from different manufacturers to see if the
drivers happen to interact well or badly with the specific flash implementation.
No, currently i havn't experimented with different sd cards. True, it would
be interesting to figure out the interaction of the driver with different
flash implementations. I'll perform these tests ASAP.
I'll just point out that Udits SGOC project has nothing to do with the
flash implementation of the cards. Udit, I would hesitate to go down the
road of multiple card tests unless you have a great deal of spare time.
There are far more pertinent questions that can be asked about the variance
between SDHCI and SDIO implementations. There is still the little detail of
porting to RTEMS too. :)
Post by Udit agarwal
Regards,
Udit
Post by Peter Jeremy
--
Peter Jeremy
_______________________________________________
https://lists.freebsd.org/mailman/listinfo/freebsd-arm
Russell Haley
2018-05-14 23:22:45 UTC
Permalink
Post by Russell Haley
Post by Udit agarwal
Post by Peter Jeremy
Post by Udit agarwal
I have conducted several sd card benchmarking tests on FreeBSD 11.1,
with
Post by Peter Jeremy
Post by Udit agarwal
both SDHCI and the newly incorporated MMCCAM driver to compare their
performance under different conditions.
1) How reproduceable are the results? Have you performed the test multiple times
to check for variance?
The images that Udit is using for his tests can be made available for
verification if that's desirable (we have a webserver available). His blogs
are very good as well.
Sorry, sent before adding the blog address: http://81.4.107.225/wordpress/
Post by Russell Haley
Post by Udit agarwal
Yes, i have carried these tests multiple time and the results are pretty
consistent upto 0.1MBPS.
Post by Peter Jeremy
2) How did you prepare the SD card prior to each write test? Flash
write
Post by Peter Jeremy
tests
are extremely sensitive to hidden move/rewrite and erase cycles
performed
Post by Peter Jeremy
by the
flash translation layer.
Before each test, I freshly recreated the filesystem to remove any trace of
the previous tests. Moreover, I have mounted/unmounted the device to purge
any data in the cache.
Post by Peter Jeremy
3) Have you tried different SD cards from different manufacturers to see if the
drivers happen to interact well or badly with the specific flash implementation.
No, currently i havn't experimented with different sd cards. True, it would
be interesting to figure out the interaction of the driver with different
flash implementations. I'll perform these tests ASAP.
I'll just point out that Udits SGOC project has nothing to do with the
flash implementation of the cards. Udit, I would hesitate to go down the
road of multiple card tests unless you have a great deal of spare time.
There are far more pertinent questions that can be asked about the variance
between SDHCI and SDIO implementations. There is still the little detail of
porting to RTEMS too. :)
Post by Udit agarwal
Regards,
Udit
Post by Peter Jeremy
--
Peter Jeremy
_______________________________________________
https://lists.freebsd.org/mailman/listinfo/freebsd-arm
Peter Jeremy
2018-05-14 21:41:35 UTC
Permalink
Post by Udit agarwal
I have conducted several sd card benchmarking tests on FreeBSD 11.1, with
both SDHCI and the newly incorporated MMCCAM driver to compare their
performance under different conditions.
Some questions:
1) How reproduceable are the results? Have you performed the test multiple times
to check for variance?
2) How did you prepare the SD card prior to each write test? Flash write tests
are extremely sensitive to hidden move/rewrite and erase cycles performed by the
flash translation layer.
3) Have you tried different SD cards from different manufacturers to see if the
drivers happen to interact well or badly with the specific flash implementation.
--
Peter Jeremy
Loading...