It probably works now because the size of a logical_address and
physical_address is 64 bits, hence using them as arguments to printf()
is probably adding 64 bits into its argument stream. %x only consumes 32
bits, so when it is used twice it first consumes the lower-half of the
logical address, then the upper (all-zero) half is consumed by the %x to
print the "physical address".
%llx will consume 64.
Its a bit hairy to use printf() for certain combinations of data type
and format string.
vararg.h-style coding is some of the most subtle C-code around... hence
we suggest performing traces in the C++-section of ruby, such as
SimicsProcessor.C, where it is possible to use cout instead.
Regards,
Dan
Dave Z. wrote:
Thanks for your quick reply, Phil!
I've changed the printf statement in ruby_operate, and everything
shows up fine.
Instead of
printf("logical address %x physical address %x\n", mem_trans->s.logical_address, mem_trans->s.physical_address);
I do the following:
printf("logical address %x\n", mem_trans->s.logical_address);
printf("physical address %x\n", mem_trans->s.physical_address);
I don't understand why, but it works.
Dave
----- Original Message ----
From: Philip Garcia <pcgarcia@xxxxxxxx>
To: Gems Users <gems-users@xxxxxxxxxxx>
Sent: Tuesday, January 23, 2007 11:38:48 AM
Subject: Re: [Gems-users] physical_address 0
Are you sure your memory map routine is working properly? It could be
an invalid return, but I don't really know that much about mmap under
solaris. Does trying to access the ram cause the user app to seg
fault (regardless of what the read/write does)? Also, I'm assuming
the kernel is set to allocate the full 8KB? Try requesting a full
8KB, it really doesn't make things better or memory efficient to only
request 1KB, as the full 8KB is needed and has to be allocated
somewhere (the kernel can't control permissions on a finer grain than
the page size). See if that helps at all, and if not I'd try having
the kernel driver memmap some memory known to be valid, and/or have it
try to directly read and write to the i/o memory (such as a specific
read or write that happens on an mmap call for debugging purposes).
Phil
On Jan 23, 2007, at 11:36 AM, Dave Z. wrote:
This is under Solaris 10. The machine I'm using is sarek (simics
2.2.19). s.physical_address is always 0 in ruby_operate,
SimicsDriver, and SimicsProcessor when trying to access the memory
mapped area.
The memory address space is 4G in Ruby and Simics. The kernel memory
I'm allocating is one page (8K), but I'm mmapping only 1K.
All the accesses to the memory mapped region (no matter what the
offset is) point to 0. I've asked on the simics forum, but they
couldn't reproduce the same problem. I don't understand... Thanks for
your help.
Dave
----- Original Message ----
From: Philip Garcia <pcgarcia@xxxxxxxx <mailto:pcgarcia@xxxxxxxx>>
To: Gems Users <gems-users@xxxxxxxxxxx <mailto:gems-users@xxxxxxxxxxx>>
Sent: Saturday, January 20, 2007 10:12:07 AM
Subject: Re: [Gems-users] physical_address 0
Is this under solaris or linux? Also, is this followed by a trap to
the tlb miss handler? How big is the memory address space?
Actually, if the mmap for address 0xfffa000 pointed to 0x0 then
accessing 0xfffa00a should goto address 0xa. If this isn't the case,
something isn't working right (although I've never used the memop
operator myself. I'd ask on the simics.net forum if anyone has seen
this problem, might be some weird simics thing.
phil
On Jan 20, 2007, at 11:18 AM, Dave Z. wrote:
Yes, Simics is actually passing Ruby a physical address of 0. I have
some kernel memory mmap'ed into my application. "ptr" points to the
mmap'ed memory space, and I know the logical address of ptr. When I
do ptr[offset] = value or value = ptr[offset], the physical
addresses for the memory requests are always 0.
Thanks,
Dave
----- Original Message ----
From: Dan Gibson <degibson@xxxxxxxx <mailto:degibson@xxxxxxxx>>
To: Gems Users <gems-users@xxxxxxxxxxx <mailto:gems-users@xxxxxxxxxxx>>
Sent: Friday, January 19, 2007 2:35:14 PM
Subject: Re: [Gems-users] physical_address 0
I thought you were calling SIM_logical_to_physical to determine the
physical address. Do you mean that Simics is actually passing Ruby a
physical address of zero? If so, what is the ASI, and what type of
access (eg store, atomic, etc.) is it?
Regards,
Dan
Dave Z. wrote:
But there is no SIM_logical_to_physical in SimicsProcessor::makeRequest. If I understand correctly, the order of function calls is ruby_operate (ruby.c), _possible_cache_miss (commands.C), makeRequest (SimicsDriver.C), makeRequest (SimicsProcessor.C). The physical address is always 0 for certain logical
addresses.
Thanks,
Dave
----- Original Message ----
From: Dan Gibson <degibson@xxxxxxxx>
To: Gems Users <gems-users@xxxxxxxxxxx>
Sent: Thursday, January 18, 2007 8:16:58 PM
Subject: Re: [Gems-users] physical_address 0
You might want to check
that SIM_logical_to_physical isn't raising any
exceptions. Have a look at SIM_last_exception().
Regards,
Dan
Dave Z. wrote:
Hi,
I have a quick question about mem_trans object in SimicsProcessor::makeRequest. Is it something expected to have mem_trans->s.physical_address=0? I've printed the physical addresses of some logical addresses and they are all 0. Here is what I do:
if
(mem_trans->s.logical_address == XXX)
printf("logical address %x physical address %x\n", mem_trans->s.logical_address, mem_trans->s.physical_address);
Thanks,
Dave
____________________________________________________________________________________
Now that's room service! Choose from over 150,000 hotels
in 45,000 destinations on Yahoo! Travel to find your fit.
http://farechase.yahoo.com/promo-generic-14795097
_______________________________________________
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.
____________________________________________________________________________________
Don't pick lemons.
See all the new 2007 cars at Yahoo! Autos.
http://autos.yahoo.com/new_cars.html
_______________________________________________
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 <mailto: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/";
<https://lists.cs.wisc.edu/archive/gems-users/%22>; to your search.
------------------------------------------------------------------------
Any questions? Get answers on any topic at Yahoo! Answers
<http://answers.yahoo.com/;_ylc=X3oDMTFvbGNhMGE3BF9TAzM5NjU0NTEwOARfcwMzOTY1NDUxMDMEc2VjA21haWxfdGFnbGluZQRzbGsDbWFpbF90YWcx>.
Try it now.
_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx <mailto: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 <mailto: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/";
<https://lists.cs.wisc.edu/archive/gems-users/%22>; to your search.
------------------------------------------------------------------------
Get your own web address.
<http://us.rd.yahoo.com/evt=49678/*http://smallbusiness.yahoo.com/domains/?p=BESTDEAL>
Have a HUGE year through Yahoo! Small Business.
<%20http://us.rd.yahoo.com/evt=49678/*http://smallbusiness.yahoo.com/domains/?p=BESTDEAL>
_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx <mailto: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/";
<https://lists.cs.wisc.edu/archive/gems-users/%22>; to your search.
------------------------------------------------------------------------
Cheap Talk? Check out
<http://us.rd.yahoo.com/mail_us/taglines/postman8/*http://us.rd.yahoo.com/evt=39663/*http://voice.yahoo.com>
Yahoo! Messenger's low PC-to-Phone call rates.
------------------------------------------------------------------------
_______________________________________________
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.
--
http://www.cs.wisc.edu/~gibson [esc]:wq!
|
![[Computer Systems Lab]](http://www.cs.wisc.edu/pics/csl_logo.gif){kind=logo}