[PATCH v4 15/21] PM / devfreq: rockchip-dfi: Add perf support

[Sascha Hauer]

On Wed, May 10, 2023 at 09:56:34PM +0200, Sascha Hauer wrote:
> On Tue, May 09, 2023 at 09:04:58PM +0100, Robin Murphy wrote:
> > On 2023-05-05 12:38, Sascha Hauer wrote:
> > > The DFI is a unit which is suitable for measuring DDR utilization, but
> > > so far it could only be used as an event driver for the DDR frequency
> > > scaling driver. This adds perf support to the DFI driver.
> > > 
> > > Usage with the 'perf' tool can look like:
> > > 
> > > perf stat -a -e rockchip_ddr/cycles/,\
> > > 		rockchip_ddr/read-bytes/,\
> > > 		rockchip_ddr/write-bytes/,\
> > > 		rockchip_ddr/bytes/ sleep 1
> > > 
> > >   Performance counter stats for 'system wide':
> > > 
> > >          1582524826      rockchip_ddr/cycles/
> > >             1802.25 MB   rockchip_ddr/read-bytes/
> > >             1793.72 MB   rockchip_ddr/write-bytes/
> > >             3595.90 MB   rockchip_ddr/bytes/
> > > 
> > >         1.014369709 seconds time elapsed
> > > 
> > > perf support has been tested on a RK3568 and a RK3399, the latter with
> > > dual channel DDR.
> > 
> > Might it be useful to offer the option of monitoring each channel
> > individually?
> 
> I gave it a quick try. For my normal testing workload the values are
> spread quite homogeneously:
> 
>  Performance counter stats for 'system wide':
> 
>         1062183406      rockchip_ddr/cycles/
>            1517.57 MB   rockchip_ddr/read-bytes0/
>            1517.11 MB   rockchip_ddr/write-bytes0/
>            1519.78 MB   rockchip_ddr/read-bytes1/
>            1514.97 MB   rockchip_ddr/write-bytes1/
>            1518.69 MB   rockchip_ddr/read-bytes2/
>            1516.01 MB   rockchip_ddr/write-bytes2/
>            1519.37 MB   rockchip_ddr/read-bytes3/
>            1515.46 MB   rockchip_ddr/write-bytes3/
>           12137.84 MB   rockchip_ddr/bytes/
> 
>        1.005686209 seconds time elapsed
> 
> I found one workload that looks strange though:
> 
>  Performance counter stats for 'system wide':
> 
>         1076756945      rockchip_ddr/cycles/
>            2326.55 MB   rockchip_ddr/read-bytes0/
>              30.60 MB   rockchip_ddr/write-bytes0/
>               0.24 MB   rockchip_ddr/read-bytes1/
>               0.09 MB   rockchip_ddr/write-bytes1/
>               0.21 MB   rockchip_ddr/read-bytes2/
>               0.09 MB   rockchip_ddr/write-bytes2/
>               0.30 MB   rockchip_ddr/read-bytes3/
>               0.10 MB   rockchip_ddr/write-bytes3/
>            2357.87 MB   rockchip_ddr/bytes/
> 
> I am running this tool (likely not the exact version):
> https://github.com/mturquette/memtest/blob/master/mtest.c
> 
> I don't know yet what this tool is doing, but it somehow manages
> to use only a single channel.

Ok, I found it out. On my RK3588 board with four channels the channels
seem to be interleaved with 512 bytes. When I only read/write the first
512 bytes of a page then only channel 0 is used, channel 1 for the
second 512 bytes etc.

The mtest tool allocates a bunch of pages, but only ever reads and
writes the first word of each page.

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |