Re: [gem5-dev] Review Request 3809: imported patch rr_arbiter_fix

2017-02-13 Thread Jason Lowe-Power

---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3809/#review9426
---

Ship it!


Ship It!

- Jason Lowe-Power


On Feb. 12, 2017, 4:36 a.m., Tushar Krishna wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/3809/
> ---
> 
> (Updated Feb. 12, 2017, 4:36 a.m.)
> 
> 
> Review request for Default, Brad Beckmann, Jieming Yin, Matthew Poremba, and 
> Onur Kayiran.
> 
> 
> Repository: gem5
> 
> 
> Description
> ---
> 
> ruby: fix round robin arbiter in garnet2.0
> The rr arbiter pointer in garnet was getting updated on every request,
> even if there is no grant. This was leading to a huge variance in wait
> time at a router at high injection rates.
> This patch corrects it to update upon a grant.
> 
> 
> Diffs
> -
> 
>   src/mem/ruby/network/garnet2.0/SwitchAllocator.cc 79c3f6a60392 
> 
> Diff: http://reviews.gem5.org/r/3809/diff/
> 
> 
> Testing
> ---
> 
> 
> Thanks,
> 
> Tushar Krishna
> 
>

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev


Re: [gem5-dev] Review Request 3809: imported patch rr_arbiter_fix

2017-02-11 Thread Tushar Krishna


> On Feb. 12, 2017, 4:50 a.m., Jieming Yin wrote:
> > We also observed long packet delay in heavy-traffic workloads. Just 
> > curious, how much does this fix improve?

I noticed the problem when the deadlock threshold that was added recently 
started getting triggered at very high injection rates (post-saturation) with 
the synthetic network traffic.
It seemed odd to me that a NI could not inject for 1 cycles, however 
congested the network may be, as round-robin should ensure that eventually 
every VC at every input port should get a chance to leave the router (in the 
absence of a true deadlock).

I didn't measure how the overall stats change now, but the deadlock counter 
does not trigger now even if I (try to) inject 1-flit every cycle showing that 
this fix works.


- Tushar


---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3809/#review9418
---


On Feb. 12, 2017, 4:36 a.m., Tushar Krishna wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/3809/
> ---
> 
> (Updated Feb. 12, 2017, 4:36 a.m.)
> 
> 
> Review request for Default, Brad Beckmann, Jieming Yin, Matthew Poremba, and 
> Onur Kayiran.
> 
> 
> Repository: gem5
> 
> 
> Description
> ---
> 
> ruby: fix round robin arbiter in garnet2.0
> The rr arbiter pointer in garnet was getting updated on every request,
> even if there is no grant. This was leading to a huge variance in wait
> time at a router at high injection rates.
> This patch corrects it to update upon a grant.
> 
> 
> Diffs
> -
> 
>   src/mem/ruby/network/garnet2.0/SwitchAllocator.cc 79c3f6a60392 
> 
> Diff: http://reviews.gem5.org/r/3809/diff/
> 
> 
> Testing
> ---
> 
> 
> Thanks,
> 
> Tushar Krishna
> 
>

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev


[gem5-dev] Review Request 3809: imported patch rr_arbiter_fix

2017-02-11 Thread Tushar Krishna

---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3809/
---

Review request for Default, Brad Beckmann, Jieming Yin, Matthew Poremba, and 
Onur Kayiran.


Repository: gem5


Description
---

ruby: fix round robin arbiter in garnet2.0
The rr arbiter pointer in garnet was getting updated on every request,
even if there is no grant. This was leading to a huge variance in wait
time at a router at high injection rates.
This patch corrects it to update upon a grant.


Diffs
-

  src/mem/ruby/network/garnet2.0/SwitchAllocator.cc 79c3f6a60392 

Diff: http://reviews.gem5.org/r/3809/diff/


Testing
---


Thanks,

Tushar Krishna

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev