[Qemu-devel] Arm host float endian fix

[Qemu-devel] Arm host float endian fix

[Paul Brook]

The newer arm VFP floating point units store double values with natural 
word-ordering, unlike the old FPA unit which always used big-endian.

Patch below updates host conditionals to reflect this.

Paul

Index: cpu-all.h
===================================================================
RCS file: /cvsroot/qemu/qemu/cpu-all.h,v
retrieving revision 1.39
diff -u -p -r1.39 cpu-all.h
--- cpu-all.h 28 Jan 2005 22:37:22 -0000 1.39
+++ cpu-all.h 6 Feb 2005 00:57:27 -0000
@@ -114,10 +114,11 @@ static inline void tswap64s(uint64_t *s)
 #define tswapls(s) tswap64s((uint64_t *)(s))
 #endif
 
-/* NOTE: arm is horrible as double 32 bit words are stored in big endian ! */
+/* NOTE: arm FPA is horrible as double 32 bit words are stored in big endian!
+ */
 typedef union {
     double d;
-#if !defined(WORDS_BIGENDIAN) && !defined(__arm__)
+#if !defined(WORDS_BIGENDIAN) && !defined(__arm__) && !defined (__VFP_FP__)
     struct {
         uint32_t lower;
         uint32_t upper;

Re: [Qemu-devel] Arm host float endian fix

[Lennert Buytenhek]

On Sun, Feb 06, 2005 at 01:03:29AM +0000, Paul Brook wrote:

> The newer arm VFP floating point units store double values with natural 
> word-ordering, unlike the old FPA unit which always used big-endian.

While we're on the subject, something similar to the patch in this
email is also still needed for proper floating point emulation if the
host and (ARM) target are not both little-endian.

	http://lists.gnu.org/archive/html/qemu-devel/2004-12/msg00339.html


--L