Thanks for informing me, Giacomo.
This is likely due to the fact that the places that call setWhenReady ignore 
the previous whenReady cycle and overwrite it. I'll do a quick patchset to 
revert the first assert and workaround this problem. I imagine the tickInserted 
assertion was ok, because the block is always first inserted, then whenReady is 
set (I am trying to run the tests, but fs takes forever on my machine). Please 
let me know if it is still broken with the tickInserted assert, because then 
that means something is really bad (I'll check here too).

Nikos/Jason, this is likely due to an incomplete definition of whenReady. If we 
need to care about previous whenReady when overwriting a valid block, then we 
must add it to the calculations (i.e., a write must wait for previous writes to 
end). However (and I think this is a better approach), if whenReady is the tick 
at which the write contents (payload) is available to be accessed, then the 
first assertion does not make sense, whenReady should be left as is (i.e., 
including the fillLatency), but on the part that uses whenReady, if the access 
must be delayed due to the block not being present, the fillLatency should be 
deducted, and the forwardLatency included (i.e., the non-write access is being 
serviced by the write packet).

Speaking of FS, to whoever runs x86 fs simulations (Gabe, possibly?), here at 
the office we always have to apply 2 local patches to be able to run with our 
own image and kernel: 
https://gem5-review.googlesource.com/c/public/gem5/+/11850 and another one that 
I haven't submitted because I don't know if it's safe for every approach, but 
that removes the variable disk2 from makeX86System, which forces a swap that is 
not present on my image (linux-bigswap2.img). Then I run (build/X86/gem5.opt 
configs/example/fs.py --disk-image=img_path --kernel=kernel_path). Could you 
take a look at it and see if I am doing something wrong, or it makes sense?

Regards,Daniel



    Em quinta-feira, 6 de dezembro de 2018 18:00:11 GMT+1, Giacomo Travaglini 
<[email protected]> escreveu:  
 
  <!--#yiv1942089052 P {margin-top:0;margin-bottom:0;}-->

Hi Daniel,
A group of quick regressions are broken after the following patch (that has 
been recently merged)(They fail the assertion that has been introduced by the 
patch)
https://gem5-review.googlesource.com/c/public/gem5/+/14715

quick/fs/10.linux-boot/alpha/linux/tsunami-simple-atomic: 
FAILEDquick/fs/10.linux-boot/alpha/linux/tsunami-simple-atomic-dual: 
FAILEDquick/fs/10.linux-boot/arm/linux/realview-switcheroo-atomic: 
FAILEDquick/fs/10.linux-boot/arm/linux/realview-simple-atomic: 
FAILEDquick/fs/10.linux-boot/arm/linux/realview-simple-atomic-checkpoint: 
FAILEDquick/fs/10.linux-boot/arm/linux/realview-simple-atomic-dual: FAILED
Could you please have a look at it?
Thanks
Giacomo



IMPORTANT NOTICE: The contents of this email and any attachments are 
confidential and may also be privileged. If you are not the intended recipient, 
please notify the sender immediately and do not disclose the contents to any 
other person, use it for any purpose, or store or copy the information in any 
medium. Thank you.  
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to