[Qemu-devel] [PATCH] * include/fpu/softfloat.h (floatx80_invalid_encoding): Handle m68k specific infinity pattern.

[Pierre Muller]

Hello,

  I submitted September 12. a bug report about wrong handling of
infinity values for m68k emulator.

https://bugs.launchpad.net/qemu/+bug/1843651

  The analysis I made in the bug report is wrong, because
I did not know that m68k FPU really uses 80-bit representations internally.
  It thus seems now to me that the m68k specific floatx80_infinity_low
corresponds to a different internal encoding of infinity on m68k floating
point units, different from the one of the usual x87 floating point unit.
  The main problem still remains that the function floatx80_invalid_encoding
does not properly handle those m68k infinity patterns.

  The patch below seems to fix the reported problem on current git.

Pierre Muller
Member of core development team of Free Pascal compiler


>From e053d3d07b1951faf0b4637ce1ec4194b8d952e5 Mon Sep 17 00:00:00 2001
From: Pierre Muller <address@hidden>
Date: Thu, 12 Sep 2019 08:10:48 +0000
Subject: [PATCH]        * include/fpu/softfloat.h (floatx80_invalid_encoding):
 Handle         m68k specific infinity pattern.

---
 include/fpu/softfloat.h | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h
index ecb8ba0114..538c35767e 100644
--- a/include/fpu/softfloat.h
+++ b/include/fpu/softfloat.h
@@ -685,10 +685,17 @@ static inline int floatx80_is_any_nan(floatx80 a)
 | pseudo-infinities and un-normal numbers. It does not include
 | pseudo-denormals, which must still be correctly handled as inputs even
 | if they are never generated as outputs.
+| As m68k uses a different pattern for infinity, thus an additional test
+| for valid infinity value must be added for m68k CPU.
 *----------------------------------------------------------------------------*/
 static inline bool floatx80_invalid_encoding(floatx80 a)
 {
+#if defined (TARGET_M68K)
+    return ((a.low & (1ULL << 63)) == 0 && (a.high & 0x7FFF) != 0)
+           && (! floatx80_is_infinity(a));
+#else
     return (a.low & (1ULL << 63)) == 0 && (a.high & 0x7FFF) != 0;
+#endif
 }

 #define floatx80_zero make_floatx80(0x0000, 0x0000000000000000LL)
--
2.20.1