Re: [Qemu-devel] [PATCH 5/7] tcg-i386: Implement deposit operation.

[Edgar E. Iglesias]

On Tue, Jan 25, 2011 at 02:07:15PM -0800, Richard Henderson wrote:
> On 01/25/2011 08:48 AM, Edgar E. Iglesias wrote:
> > OK, I see. Maybe we should try to emit an insn sequence more similar
> > to what tcg was emitting (for the non 8 & 16-bit deposits)?
> > That ought too at least give similar results as before for those and
> > give us a speedup for the byte and word moves.
> 
> Please try this replacement version for tcg/i386/*.
> 
> I was able to run your cris testsuite, and stuff looked ok there.
> But for some reason the microblaze kernel would not boot.  It seems
> that the kernel commandline isn't in place properly and it isn't 
> finding the disk image.

Yes, you need to build qemu with libfdt for that image to boot.
IIRC, libfdt comes with the dtc package on some dists. In the future
I'll see if can upload an image that boots without the need of
devicetree manipulation in qemu.

I tried your new patch and got similar results as before. Maaybe slightly
faster but within the noise.

I looked a little bit more at it and realized that I'm probably not doing
a fair comparition with microblaze. The write_carry sequence actually
bit deposits a bit into two locations with a sequence of tcg ops that
is not much longer than the one to deposit a single bit. So I'm basically
comparing the cost of a single tcg deposit sequence with the cost of two
tcg deposit backend ops. I should probably just accept that the new
deposit op is not worth using for that particular case.

It would be nice if somebody else also tested this patch. Before we
agree on applying it.

One note, the tcg_scratch_alloc hunk from the previous version was
missing on this one.

Thanks