Hi Eliot, (Picking this up again after a while.) :-) Thank you for your detailed answer. I was able to get a lot of useful data points from these statistics. Is there a way to get what instruction was retired/committed and when (tick)?
Thanks, Priyanka. On Thu, Feb 23, 2023 at 6:31 PM Eliot Moss <m...@cs.umass.edu> wrote: > On 2/24/2023 11:31 AM, Priyanka Ankolekar via gem5-users wrote: > > Hi Eliot, > > I am trying to correlate my gem5 RISCV configuration with a RISCV RTL > design for a certain benchmark. > > I would like to plot /retired instructions versus ticks/ to make sure > that both (gem5 and RTL) are > > (mostly) performing the same operations. (I already have this plot from > RTL). > > > > Is there a way to do that in gem5? Or are there other profiling tools > that will help me get this data? > > I thought this was already answered. You can simply examine a dump. The > amount of time, in ticks, is given on the simTicks line. While the length > of > time of one tick can be changed, it is usually one picosecond (1/1000 of a > nanosecond) so that everything can be expressed as an integral number of > ticks. But you can check that by examining the simFreq line in the dump. > If > you such for committedInsts, then that tells you the number of committed > instructions for each cpu, which would match the number retired (except > maybe > for a small number that were committed but not retired at the precise > moment > the dump was taken). You can also search for cpi and ipc for each > processor, > and total, namely cycles per instruction and instruction per cycle. Note > that > *cycle* is a processor clock cycles, not a *tick*. The line with > cpu_clk_domain.clock will tell you how many ticks per clock there are. And > there is also numCycles that gives you the number of cycles directly. > > cpi is numCycles / committedInsts, and ipc is 1/cpi. > > Is this what you were looking for, or something different that I have > somehow > not understood? > > EM >
_______________________________________________ gem5-users mailing list -- gem5-users@gem5.org To unsubscribe send an email to gem5-users-le...@gem5.org
