Aurelien Jarno wrote:
> On Wed, May 16, 2007 at 02:00:14AM +0200, Aurelien Jarno wrote:
> > On Tue, May 15, 2007 at 10:07:00PM +0100, Paul Brook wrote:
> > > > The patch below fixes the emulation of dmult and dmultu by doing a real
> > > > 64x64 -> 128 multiplication.
> > >
> > > Can we share the
On Tue, May 15, 2007 at 07:26:22PM -0600, C.W. Betts wrote:
> Would this patch work with other 64-bit processors, or is it
> x86-64-only? I'm mainly thinking about 64-bit PowerPCs, but I don't
> know if PowerPCs uses the same method.
This patch works on all hosts CPU using the C version, but it
On Wed, May 16, 2007 at 02:00:14AM +0200, Aurelien Jarno wrote:
> On Tue, May 15, 2007 at 10:07:00PM +0100, Paul Brook wrote:
> > > The patch below fixes the emulation of dmult and dmultu by doing a real
> > > 64x64 -> 128 multiplication.
> >
> > Can we share the implementation between targets ple
C.W. Betts wrote:
> Would this patch work with other 64-bit processors, or is it
> x86-64-only? I'm mainly thinking about 64-bit PowerPCs, but I don't
> know if PowerPCs uses the same method.
You may have missed the x86-64 specific assembly code in there.
Thiemo
Would this patch work with other 64-bit processors, or is it
x86-64-only? I'm mainly thinking about 64-bit PowerPCs, but I don't
know if PowerPCs uses the same method.
PGP.sig
Description: This is a digitally signed message part
On Tue, May 15, 2007 at 10:07:00PM +0100, Paul Brook wrote:
> > The patch below fixes the emulation of dmult and dmultu by doing a real
> > 64x64 -> 128 multiplication.
>
> Can we share the implementation between targets please. We already have an
> implementation of mul64 in target-i386/helper.c
> The patch below fixes the emulation of dmult and dmultu by doing a real
> 64x64 -> 128 multiplication.
Can we share the implementation between targets please. We already have an
implementation of mul64 in target-i386/helper.c.
I've no real preference which implementation we use, I just don't w
Hi,
The patch below fixes the emulation of dmult and dmultu by doing a real
64x64 -> 128 multiplication.
On x86_64, it uses the mul/imul instruction, an equivalent C code (but
much slower) is provided for other architectures.
Cheers,
Aurelien
Index: target-mips/op_helper.c
