Re: [Qemu-devel] [PATCH] SH4: convert fmov/fadd to TCG

[Aurelien Jarno <aurelien@aurel32.net>]

On Sun, Aug 31, 2008 at 09:32:03PM +0200, Aurelien Jarno wrote:
> On Mon, Sep 01, 2008 at 02:28:59AM +0900, Shin-ichiro KAWASAKI wrote:
> > Thank your for comments, Blue Swirl!
> > The new patch is shown at the end of this mail.
> > Reviews are welcome, again.
> >
> 
> Thanks for you patch. It looks good, but it doesn't work here (on an
> amd64 host): while the kernel is booting correctly with qemu-system-sh4,
> running /usr/bin/cal with qemu-sh4 causes a segfault.
> 

I have finally found the problem: the gen_ld_frN_FT and gen_st_frN_FT
already have access to cpu_env, this value does not needed to be passed
again. Passing it as a TCGv variable caused it to be truncated.

At the end, looking at what other targets do, I have decided to convert
the SH4 floating point ops the "MIPS way", which has the advantage of
only using TCG temp instead of global FT0, FT1, DT0 and DT1 variables.

I have committed that to the SVN, so the SH4 target is now fully
converted to TCG. I have tested my changes using /usr/bin/awk with
sh4-linux-user doing FP computations.

In the next days/weeks, I'll try to convert the alpha target to TCG, as
I need some more experience with TCG before doing the PowerPC one. I
have seen that Tristan already send patches about that, I'll have a look
later.

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net