On Donnerstag, 23. Januar 2020 12:13:51 CET Greg Kurz wrote:
> Honestly it doesn't seem to change anything significant for me.
> Mean time calculated over 100 runs:
>
> Without this patch:
>
> [greg@bahia qemu-9p]$ (cd .mbuild-$(stg branch)/obj ; export
> QTEST_QEMU_BINARY='x86_64-softmmu/qemu-system-x86_64'; make all
> tests/qtest/qos-test && for i in {1..100}; do tests/qtest/qos-test -p
> $(tests/qtest/qos-test -l | grep readdir/basic); done) |& awk '/IMPORTANT/
> { print $10 }' | sed -e 's/s//' -e 's/^/n+=1;x+=/;$ascale=6;x/n' | bc
> .055654
>
> With this patch:
>
> [greg@bahia qemu-9p]$ (cd .mbuild-$(stg branch)/obj ; export
> QTEST_QEMU_BINARY='x86_64-softmmu/qemu-system-x86_64'; make all
> tests/qtest/qos-test && for i in {1..100}; do tests/qtest/qos-test -p
> $(tests/qtest/qos-test -l | grep readdir/basic); done) |& awk '/IMPORTANT/
> { print $10 }' | sed -e 's/s//' -e 's/^/n+=1;x+=/;$ascale=6;x/n' | bc
> .058786
:))) Mhmm, that's because you have run this test WITHOUT the actual readdir
optimization patch. In this scenario the readdir latency issue is so bad, that
the driver's n-square issue does not even matter, so same here:
Unoptimized readdir, no n-squre correction hack:
Time client spent for waiting for reply from server: 0.082831s [MOST
IMPORTANT]
Unoptimized readdir, with n-squre correction hack:
Time client spent for waiting for reply from server: 0.082539s [MOST
IMPORTANT]
BUT, now look at this *with* readdir optimization applied:
Optimized readdir, no n-square correction hack:
Time 9p server spent on synth_readdir() I/O only (synth driver): 0.021304s
Time 9p server spent on entire T_readdir request: 0.022033s [IMPORTANT]
Time client spent for waiting for reply from server: 0.022408s [MOST
IMPORTANT]
Optimized readdir, with n-square correction hack:
Time 9p server spent on synth_readdir() I/O only (synth driver): 0.001576s
Time 9p server spent on entire T_readdir request: 0.002244s [IMPORTANT]
Time client spent for waiting for reply from server: 0.002566s [MOST
IMPORTANT]
Got it? :)
I had good reasons for printing out the time spent on raw driver only.
Best regards,
Christian Schoenebeck
