[Dwarf-discuss] Dwarf-discuss Digest, Vol 23, Issue 1

[Ramana Radhakrishnan]

I thought this was decided by a paragraph some place in the ABI which
dealt with DWARF register numbers so that all vendors would agree on
it. 


-Ramana

On Wed, 2006-10-04 at 16:21 -0400, Bishop, John E wrote:
> > I understand that DW_OP_breg5 is the value of a register, right? But
> > what is that register? (ebp? I am on Intel x86)
> 
> This is one of those things that each compiler gets to
> decide, if I understand correctly.  DWARF just says how
> to name a register, the compiler guys decide how to map
> those names to actual registers, and the debugger must
> follow suit.
> 
> I believe the Intel mapping is as follows, based on one
> experiment with IDB internals:
> 
>  Register number 0 is eax
>  Register number 1 is ecx
>  Register number 2 is edx
>  Register number 3 is ebx
>  Register number 4 is esp
>  Register number 5 is ebp
>  Register number 6 is esi
>  Register number 7 is edi
>  Register number 8 is eip
>  Register number 9 is eflags
>  Register number 10 is cs
>  Register number 11 is ss
>  Register number 12 is ds
>  Register number 13 is es
>  Register number 14 is fs
>  Register number 15 is gs
>  Register number 16 is orig_eax
>  Register number 17 is fctrl
>  Register number 18 is fstat
>  Register number 19 is ftag
>  Register number 20 is fiseg
>  Register number 21 is fioff
>  Register number 22 is foseg
>  Register number 23 is fooff
>  Register number 24 is fop
>  Register number 25 is f0
>  Register number 26 is f1
>  Register number 27 is f2
>  Register number 28 is f3
>  Register number 29 is f4
>  Register number 30 is f5
>  Register number 31 is f6
>  Register number 32 is f7
>  Register number 33 is xmm0
>  Register number 34 is xmm1
>  Register number 35 is xmm2
>  Register number 36 is xmm3
>  Register number 37 is xmm4
>  Register number 38 is xmm5
>  Register number 39 is xmm6
>  Register number 40 is xmm7
>  Register number 41 is mxcsr
>  Register number 42 is vfp
> 
> Note that these don't exactly reflect any of
> the structs the system returns to PTRACE calls
> or the like.  And they may be IDB-internal rather
> than the actual BREG mapping.  I'll try to 
> chase that down and post another reply.
> 
> 	-John
> 
> _______________________________________________
> Dwarf-discuss mailing list
> Dwarf-discuss at lists.freestandards.org
> http://lists.freestandards.org/mailman/listinfo/dwarf-discuss
-- 
Ramana Radhakrishnan <ramana.radhakrishnan at codito.com>
Codito Technologies Pvt. Ltd.