[Qemu-devel] [PATCH] Sparc64 host support update: i386-user builds

[Qemu-devel] [PATCH] Sparc64 host support update: i386-user builds

[Blue Swirl]

[-- Attachment #1: Type: text/plain, Size: 610 bytes --]

I got the i386-user to build by moving the troublesome SSE ops to helper 
(only for Sparc32/64, no effect at least to x86 host). The emulator crashes 
when executing the first TB, code generation and register use could be 
suspected. The patch makes ops_sse.h more confusing to read. Any comments?

The remaining target specific problem with PPC FPU ops could be solved by 
changing PPC to soft float. Would that be OK?

_________________________________________________________________
FREE pop-up blocking with the new MSN Toolbar - get it now! 
http://toolbar.msn.click-url.com/go/onm00200415ave/direct/01/

[-- Attachment #2: sparc64_host.diff.bz2 --]
[-- Type: application/octet-stream, Size: 6942 bytes --]

Re: [Qemu-devel] [PATCH] Sparc64 host support update: i386-user builds

[Paul Brook]

> I got the i386-user to build by moving the troublesome SSE ops to helper
> (only for Sparc32/64, no effect at least to x86 host). The emulator crashes
> when executing the first TB, code generation and register use could be
> suspected. The patch makes ops_sse.h more confusing to read. Any comments?

Requiring that ops don't use a stack frame doesn't seem like a realistic 
constraint.

> The remaining target specific problem with PPC FPU ops could be solved by
> changing PPC to soft float. Would that be OK?

The use of soft-float should be a host decision, not a target decision.

All targets should be converted to use the softfloat.h interface. Hosts can 
then choose to use with the softfloat library or the native implementation 
depending on the host FP capabilities.

Paul

Re: [Qemu-devel] [PATCH] Sparc64 host support update: i386-user builds

[Blue Swirl]

>Requiring that ops don't use a stack frame doesn't seem like a realistic
>constraint.

It can be done, see my latest patch. Of course there's no point doing it if 
there is no real need.

> > The remaining target specific problem with PPC FPU ops could be solved 
>by
> > changing PPC to soft float. Would that be OK?
>
>The use of soft-float should be a host decision, not a target decision.
>
>All targets should be converted to use the softfloat.h interface. Hosts can
>then choose to use with the softfloat library or the native implementation
>depending on the host FP capabilities.

Good, I thought using softfloat.h interface implied softfloat library use.

I'll make an updated Sparc64 patch without any sse mess, and separate the 
softfloat.h interface changes.

_________________________________________________________________
FREE pop-up blocking with the new MSN Toolbar - get it now! 
http://toolbar.msn.click-url.com/go/onm00200415ave/direct/01/

Re: [Qemu-devel] [PATCH] Sparc64 host support update: i386-user builds

[Blue Swirl]

[-- Attachment #1: Type: text/plain, Size: 489 bytes --]

I made a simpler version of the Sparc64 patch which doesn't attempt to avoid 
the stack use. All targets build on Sparc64 host. I copied the relevant 
parts of the fixes to Sparc32 also.

The PPC patch is also updated, but I have no interest in maintaining this 
patch in the future.

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE! 
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/

[-- Attachment #2: sparc64_host.diff.bz2 --]
[-- Type: application/octet-stream, Size: 4324 bytes --]

[-- Attachment #3: ppc_softfloat.diff.bz2 --]
[-- Type: application/octet-stream, Size: 1861 bytes --]