Re: [Gems-users] Instruction count in Ruby

[Dan Gibson <degibson@xxxxxxxx>]

The total numbers of instruction and data references are reported by the Driver Stats. ("Insn requests" and "Data requests", respectively)

mehmetderin.harmanci@xxxxxxx wrote:

 Thanks Dan for the quick reply. Just one complementary question: Is it
 possible to find out the number of total memory accesses (instruction + data
 hits+misses )? Is it for example the numbers reported by the SimicsDriver
 Transaction results? I would say from those numbers the number of Fast Path
 accesses + number of Request misses is the total number of memory accesses
 (where fast path accesses are the number of hits).
 Would that statement be true?

 Sorry to bother you about those details,

 Derin Harmanci


Quoting Dan Gibson <degibson@xxxxxxxx>:

  

Ruby determines the number of executed instructions directly from
simics, via the SimicsDriver via SimicsProcessor via the Simics
interface (I believe the API call is SIMICS_get_instrucition_count).
Ruby's profiler adds 1 to the reported number so it is safe to use as a
divisor elsewhere (eg. no divide-by-zero). Hence, running 3 instructions
=> Ruby reports 4 instructions.

Regards,
Dan

mehmetderin.harmanci@xxxxxxx wrote:

    

Hello all,

I was trying to start a simulation and I have figured out something which
      

is
    

maybe obvious but, I think somewhat misleading (at least for me). The issue
      

is
    

the following (I 'll be explaining it on an example):

When I run 3 instructions on the cashew machine (Ultrasparc II processor)
      

Ruby
    

counts 4 instructions in its statistics. The difference comes from the fact
that
there is one ldx instruction among the instructions which of course causes
      

a
    

data access other than an instruction access and this data access is
      

counted as
    

an instruction whereas it is just a memory access.

With this experiment I come to the conclusion that the instruction count
reported by Ruby is just the number of memory accesses (sum of instruction
      

and
    

data accesses) and not the number of instructions run. Can somebody confirm
this observation?

Thanks for your attention,

Derin Harmanci.


_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding
      

"site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.
    

_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding
"site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.


    

_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.