Re: [Gems-users] ruby for x86: to install ruby for multiple memory objects or just one

[Derek Hower <drh5@xxxxxxxxxxx>]

(Response from Dan Gibson (on vacation) via proxy Derek Hower)

Since the x86-Ruby patch was intended for GEMS 1.1, I have personally 
never tried to use Ruby on an x86 target with Simics 3.x. However, you 
are correct in that Ruby tries to install itself on /all physical memory 
objects/. For the x86 targets I tried (Dredd and Enterprise of Simics 
2.x, if memory serves), these memories were phys_mem[0..NCPUS-1].
Your patch should aim to install Ruby on physical memory objects in 
general. I would not be very worried about missing objects -- the 
various targets are often quite different.
If you would be interested in sharing your patch with other gems users, 
please let me know! There is a lot of interest in x86-based targets with 
GEMS. (please send email directly to gibson@xxxxxxxxxxx).
Regards,
Dan


soohong p kim wrote:
> I've been trying to work on ruby-x86-patch for gems1.4 (based on x86-patch
> for gems1.1).  My end goal is to use simics 3.x and gems 1.4 to run
> x86-target with ruby.
>
> x86-patch seemed to install ruby on all phys_mem objects ("phys_mem%i")
> instead of just one, i.e. phys_mem0 or phys_mem, resulting in the following
> in ruby.c for ruby-x86-patch for gems1.1:
>
>   for(i=0;i<nProcs;i++) {
>     sprintf(memname,"phys_mem%i",i);
>     memory = SIM_get_object(memname);
> ...
>
> Is memory-per-processor assumption true for any x86-based targets?  My
> 4-processor version of Tango seemed to have just one memory object
> (phys_mem0).  And all cpu[0123]_mem objects seem to point to phys_mem0.  And
> the above nProc loop seem to cause "load-module ruby" to fail.  
>
> Any help will be greatly appreciated.
>
> Soohong
>
>
> P.S.  memory-related objects in my checkpoint file...
>
> ...
> ...
> OBJECT cpu0_mem TYPE memory-space {
> 	queue: cpu0
> 	map: ((0xfee00000, cpu0_apic, 0, 0, 0x4000, NIL, 0, 0, 0))
> 	default_target: (phys_mem0, 0, 0, phys_mem0)
> }
> OBJECT cpu1_mem TYPE memory-space {
> 	queue: cpu0
> 	map: ((0xfee00000, cpu1_apic, 0, 0, 0x4000, NIL, 0, 0, 0))
> 	default_target: (phys_mem0, 0, 0, phys_mem0)
> }
> OBJECT cpu2_mem TYPE memory-space {
> 	queue: cpu0
> 	map: ((0xfee00000, cpu2_apic, 0, 0, 0x4000, NIL, 0, 0, 0))
> 	default_target: (phys_mem0, 0, 0, phys_mem0)
> }
> OBJECT cpu3_mem TYPE memory-space {
> 	queue: cpu0
> 	map: ((0xfee00000, cpu3_apic, 0, 0, 0x4000, NIL, 0, 0, 0))
> 	default_target: (phys_mem0, 0, 0, phys_mem0)
> }
> ...
> ...
> OBJECT phys_mem0 TYPE memory-space {
> 	queue: cpu0
> 	map: ((0, dram0, 0, 0, 0xa0000, dram0, 0, 0, 0),
>               (0x100000, ram0, 0, 0x100000, 0xff00000, NIL, 0, 0, 0))
> 	default_target: (pci_mem0, 0, 0, pci_mem0)
>
> _______________________________________________
> 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.