On a sunny day (Fri, 6 Mar 2020 15:25:38 -0000 (UTC)) it happened Martin
Gregorie  wrote in :

>On Fri, 06 Mar 2020 13:52:06 +0000, Jan Panteltje wrote:
>
>> 9 minutes, 540 seconds, makes little difference, Pi4_speed / P2_speed =
>> 540 / 17 = more than 31 times faster.
>>
>>
>Does the Pi 4 show the same difference in speed between successive
>identical compiler runs without a reboot or disk switch in between?

No, almost the same to a fraction of a second, sometimes even a second longer
then again a second earlier.

It is not a disk IO bottle-neck, but mainly processor power used to parse the C
code.
As to the sources I used:
graphcs.c     5684 lines of code (with some empty lines)
navigation.c   852 lines
xgpspc.c      5084
ais.c         4020
x11.c         2207
etc
processor cycles mainly go into parsing the C code.

With Pi4 compiling is fun,
4 seconds for graphics.c only...

Speed for my Hitachi 3.2 TB USB disk connected to the Pi:
raspi95: # hdparm -t /dev/sda2
/dev/sda2:
 Timing buffered disk reads: 382 MB in  3.01 seconds = 127.01 MB/sec

raspi95: # hdparm -T /dev/sda2
Timing cached reads:   1660 MB in  2.00 seconds = 831.84 MB/sec

No bottleneck there I think :-)

(apt-get install hdparm)



>On this T440  I see the same speed-up for subsequent compilations
>compared with the first following a reboot or switch from other workloads.
>Its not specific to C either: Open Java 1.8 does exactly the same, though
>I haven't yet attempted to measure the speed difference for ant-
>controlled Java compoilation.
>
>The R61i it replaced, which, BTW has 3GB of RAM and a Core Duo chip
>clocked at 1.6 GHz, used to show the same speed difference.

OK I have the 4GB version of the Pi4

I formatted the 3.2 GB disk with ext4.
From 'mount':
/dev/sda2 on /mnt/sda2 type ext4 (rw,relatime)


>I bought the T440 when the R61i disk failed and I discovered that its
>disk interfacing electronics could not handle a disk bigger than 250 GB,
>and at the time all disks under 320 GB were no longer available. I
>subsequently revived the machine by fitting a 128GB Sandisk SSD and I've
>just timed that using the ecaxt same workload.
>
>The 1 st compilation after booting it took 2.540 seconds and the second
>one took 2.217 secs , so the cache has a tiny effect, but thats largely
>masked by the much greater speed of the SSD.
>
>The T440 has a 500 GB Toshiba HDD installed, which threw its first hard
>error last weekend (at around 21,000 hours and no problems since) so I
>bought a WD Black drive to replace it: I'll be very interested to see
>what effect its large (32MB) cache has on overall system performance when
>the Tosh finally dies and gets replaced.

I now have 3 USB harddisks, 1TB unknow make on the TV, 1 TB Seagate normally
off for backups,
and the 3.2 TB Hitachi on the RP4.

So far no problems,

Only harddisk I ever killed was when I dropped one from the bookshelf 20 years
or so ago..
For the rest never have any problems, several on for 15 years or more 24/7.

--- SoupGate-Win32 v1.05