qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
* [Qemu-devel] Attempt for reproducible behavior
@ 2009-02-26 12:44 Alex-Uni
  2009-02-26 16:43 ` Jan Marten Simons
  0 siblings, 1 reply; 2+ messages in thread
From: Alex-Uni @ 2009-02-26 12:44 UTC (permalink / raw)
  To: qemu-devel

Hello,

I am trying to modify Qemu, so it offers reproducible behavior. I'm 
using the icounts for the time measurement.
For example:
I have a simple code which measures the time before the start of a loop. 
The loop runs for an definite amount of rounds. In this case I let it 
count to 100000000 in a while-loop. Then the codes measures the time 
after the loop. Both times in milliseconds. The time difference is about 
5000 milliseconds.

Now, while the loop is still running, i make a snapshot. Lets say at 
2000 milliseconds.
After loading this snapshot it the loop should run for 3000 
milliseconds. But it doesnt, it's alway different.

I figured, that it depends on the value of next_tb and therefor on 
tcg_qemu_tb_exec(tc_ptr) in line 615 in the cpu-exec.c.

Now I have the problem, that I can't follow the code. I don't know why 
tcg_qemu_tb_exec(tc_ptr) produces different results
and where to look. I only found the define in the tcg.h.

Can somebody help me with this problem?
I'm working on the svn-version 5806.

Thank you
Alex
(Please CC: me when replying, I am not on the list)

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [Qemu-devel] Attempt for reproducible behavior
  2009-02-26 12:44 [Qemu-devel] Attempt for reproducible behavior Alex-Uni
@ 2009-02-26 16:43 ` Jan Marten Simons
  0 siblings, 0 replies; 2+ messages in thread
From: Jan Marten Simons @ 2009-02-26 16:43 UTC (permalink / raw)
  To: qemu-devel

Am Donnerstag, 26. Februar 2009 13:44:15 schrieb Alex-Uni:
> I am trying to modify Qemu, so it offers reproducible behavior. I'm
> using the icounts for the time measurement.

qemu is not cycle accurate. Its using code translation with some quite cleveer 
caching.

Perhapps Bochs is the better alternative to work on for your problem.

With regards,
Jan

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2009-02-26 16:43 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-02-26 12:44 [Qemu-devel] Attempt for reproducible behavior Alex-Uni
2009-02-26 16:43 ` Jan Marten Simons

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).