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


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

Re: [patch] MIPS32 is a real ISA, and MIPS ISA/CPU name/constant cleanup.


cgd@sibyte.com (Chris G. Demetriou) writes:
> MIPS32 fake-cpu -> ISA, well, see my previous mail.
> 
> stubbed mips5 and mips64 bits, so i wouldn't have to churn stuff and
> wouldn't forget them later.  I promise i'll fill these out a little
> bit more in the very near future, but they shouldn't cause new
> problems as-is.
> 
> gas tc-mips CPU/ISA lookup table thang: so, the previous way of doing
> this was just ... -- obviously grown over time and in need of a good
> Cleaning.  New code much simpler, and more consistent.  Also, the
> md_begin() code to set ISA and then was just insane.  If you
> specified an ISA on the command line but not a CPU, it would end up
> marking the binary with the CPU + ISA that were the default for
> TARGET_CPU.  losing.  fixed.
> 
> ISA_* constants rather than plain numbers: same rationale for
> converting to CPU_* really, and also can if done carefully (which I
> did) can be used to do reasonable ISA instruction selection even in
> the new world where ISAs don't necessarily support supersets of
> previous ISAs instructions.
> 
> P4 -> I32: prep for addition of I64 and I5; this is an artifact of
> wanting to pipeline diffs, since i had to change the rest of the table
> in an already-outstanding patch and wanted this one to apply.  *sigh*
> If it's accepted I promise i'll clean it up ASAP.  8-)

FYI, i'm going to send in a revision of this diff later today.  i've
stripped out all of the not-yet-cooked stuff (i.e. the mips5 and
mips64 stubs I added), did some other cleanup, and made a "generic
MIPS32" processor be the default for MIPS32 (With the various MIPS32
4K names pointing to it).

There are a couple of reasons for that: (1) it's a complete spec, and
so the notion of an absoutely vanilla CPU really does make sense, and
afaik the 4K is basically "vanilla MIPS32 without FP;" and (2) i'll
need to do the same for mips64 (since the sb1 isn't a vanilla mips64
cpu 8-).


So, don't nobody jump up, approve, and apply that patch all of a
sudden, but do read it over.  8-)


chris

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