[Qemu-devel] [patch] CPU_DoubleU and soft-float

[Qemu-devel] [patch] CPU_DoubleU and soft-float

[Paul Brook]

The definition of CPU_DoubleU is incorrect on arm hosts when using
soft-float. In this case "float64" will have host integer word ordering, not
host hardware float ordering.

Paul

Index: cpu-all.h
===================================================================
RCS file: /cvsroot/qemu/qemu/cpu-all.h,v
retrieving revision 1.42
diff -u -p -r1.42 cpu-all.h
--- cpu-all.h 13 Mar 2005 18:50:23 -0000 1.42
+++ cpu-all.h 26 Mar 2005 20:48:25 -0000
@@ -120,7 +120,8 @@ static inline void tswap64s(uint64_t *s)
    endian ! */
 typedef union {
     float64 d;
-#if defined(WORDS_BIGENDIAN) || (defined(__arm__) && !defined(__VFP_FP__))
+#if defined(WORDS_BIGENDIAN) \
+    || (defined(__arm__) && !defined(__VFP_FP__) && !defined(CONFIG_SOFTFLOAT))
     struct {
         uint32_t upper;
         uint32_t lower;

Re: [Qemu-devel] [patch] CPU_DoubleU and soft-float

[Fabrice Bellard]

OK for the patch.

It would be better to force the use of soft float on ARM if the VFP is 
not present. Supporting weird ordering for doubles is a source of 
problems and soft float may be faster than the kernel based ARM fpu 
emulator anyway.

Fabrice.

Paul Brook wrote:
> The definition of CPU_DoubleU is incorrect on arm hosts when using
> soft-float. In this case "float64" will have host integer word ordering, not
> host hardware float ordering.
> 
> Paul
> 
> Index: cpu-all.h
> ===================================================================
> RCS file: /cvsroot/qemu/qemu/cpu-all.h,v
> retrieving revision 1.42
> diff -u -p -r1.42 cpu-all.h
> --- cpu-all.h 13 Mar 2005 18:50:23 -0000 1.42
> +++ cpu-all.h 26 Mar 2005 20:48:25 -0000
> @@ -120,7 +120,8 @@ static inline void tswap64s(uint64_t *s)
>     endian ! */
>  typedef union {
>      float64 d;
> -#if defined(WORDS_BIGENDIAN) || (defined(__arm__) && !defined(__VFP_FP__))
> +#if defined(WORDS_BIGENDIAN) \
> +    || (defined(__arm__) && !defined(__VFP_FP__) && !defined(CONFIG_SOFTFLOAT))
>      struct {
>          uint32_t upper;
>          uint32_t lower;
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
> 
> 

Re: [Qemu-devel] [patch] CPU_DoubleU and soft-float

[Paul Brook]

On Thursday 07 April 2005 20:47, Fabrice Bellard wrote:
> OK for the patch.
>
> It would be better to force the use of soft float on ARM if the VFP is
> not present. Supporting weird ordering for doubles is a source of
> problems and soft float may be faster than the kernel based ARM fpu
> emulator anyway.

Yes, however arm targets are the only ones that will currently build 
soft-float. The other targets still contain native floating point code.

Paul