This is the mail archive of the gdb@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFC] Simplify SIMD registers


Martin M. Hunt writes:
 > On Fri, 26 Apr 2002, Elena Zannoni wrote:
 > 
 > > Now, I have concerns for platforms other than Altivec, that may have
 > > used a representation like the one I would like to abandon.
 > > I am talking about SSE regs on x86 which are v4sf type. Right?
 > 
 > SSE registers are v4sf only, but SSE2 registers (Pentium 4) are
 > v2df, v4sf, v16qi, v8hi, v4si, and v2di. I don't believe GDB supports 
 > them yet.
 > 

Ah ok, so those could be supported w/o structure wrappers. But the
debug format is only stabs on x86, and we don't set the flags for those
types to TYPE_VECTOR. If we did that we could avoid the structures.

 > 
 > > For this reason I didn't touch the existing built in types which have
 > > the structure wrapping, but I introduced new ones to be used by
 > > AltiVec registers.
 > > 
 > > Comments? 
 > 
 > I would like to see us treat all SIMD registers in the same manner, not 
 > just Altivec. Even though the packing and size will vary, the display 
 > syntax should be consistent.
 > 

Yes, I would like this too. 
Any opinions from the x86 aficionados? Eli? Daniel? Mark?

 > Martin

Elena


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]