ERROR: no valid data array organizations found

[Hamid]

Hi All,

I'm generating some design configurations and for some of them I got this error after Invoking CACTI for cache power and area estimates:

ERROR: no valid tag organizations found

I set the below parameters inside gem5.cfg:

cache_size = 2kB

cache_line_sz = 64

cache_assoc = 8
cache_bandwidth = 8
...

and corresponds to the gem5.cfg, I make the cacti cache config file as:

-size (bytes) 2048

-associativity 8

-read-write port 8

-cache type "cache"

-block size (bytes) 64
...

Apparently, the problem is with "block size (bytes)" in cacti config file. If I reduce it from 64 bytes to 8 (which results in underestimating the power and area of the accelerator), gem5-Aladdin will generate the results for me.

Right now, I cannot search the whole space and for some combinations, I cannot get the results from gem5-Aladdin.

I appreciate any help on the problem.

Thanks,

Hamid

[Sam Xi]

You can't just generate any arbitrary cache configuration. CACTI has restrictions on the minimum sizes of various blocks in the cache. You shouldn't expect to be able to search the entire space in this regard.

--
You received this message because you are subscribed to the Google Groups "gem5-Aladdin users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gem5-aladdin-us...@googlegroups.com.
To post to this group, send email to gem5-alad...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gem5-aladdin-users/8406fec6-064e-4831-a9ca-85c2c5c78a74%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Thanks,

Sam Xi

Harvard University

Computer Science, Ph.D. Candidate

http://www.samxi.org

[Mark Blanco]

Hi Sam,

Resurrecting this thread with a similar issue. I am performing parameter sweeps using a semi-custom script. The parameters I am sweeping for the accelerator (parameters which end up in the gem5 config) do include cache sizes and tlb sizes for non-scratchpad design runs. Initially I was using the common cacti config files in the integration-tests directory; recently however used some of your cacti config-writing code from the xenon sub-directory to write cacti configs that reflect at the minimum the overall cache size. On some design runs, the simulation fails with the error posted above.

My question on this is what values should remain unchanged in the cacti configs, and what values are safe to vary? I also noticed that the gem5 cfg contains paths for cache, tlb, load queue, and store queue configs but that your parameter sweeper only generates cache, tlb, and a single queue configuration. Why does it work this way? If it would be helpful I can share the scripts I've written.

Thanks,
Mark

To unsubscribe from this group and stop receiving emails from it, send an email to gem5-aladdin-users+unsub...@googlegroups.com.

To post to this group, send email to gem5-alad...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gem5-aladdin-users/8406fec6-064e-4831-a9ca-85c2c5c78a74%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Sam Xi]

Xenon doesn't accommodate for tool specific constraints, so there's no way for it to know what CACTI can or can't accept. Generally, you just need to be careful with small cache sizes and/or highly associative structures. You can vary most things but you'll find that a few don't work. Not much that Xenon can do about that.

I think I just assumed that the load/store queues used the same configuration. That's not really something that I commonly sweep. You can easily change that in sweeps/config_writer/gem5_config_writer.py (writeAllCactiConfigs).

To unsubscribe from this group and stop receiving emails from it, send an email to gem5-aladdin-us...@googlegroups.com.

To post to this group, send email to gem5-alad...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gem5-aladdin-users/8406fec6-064e-4831-a9ca-85c2c5c78a74%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Thanks,

Sam Xi

Harvard University

Computer Science, Ph.D. Candidate

http://www.samxi.org

--

You received this message because you are subscribed to the Google Groups "gem5-Aladdin users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to gem5-aladdin-us...@googlegroups.com.

To post to this group, send email to gem5-alad...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gem5-aladdin-users/dbcb947e-c691-4df8-a638-d733c95ae2d6%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.