ARM: workaround for bug in GNU assembler
authorMans Rullgard <mans@mansr.com>
Tue, 5 Jul 2011 17:29:35 +0000 (18:29 +0100)
committerMans Rullgard <mans@mansr.com>
Tue, 5 Jul 2011 17:39:23 +0000 (18:39 +0100)
Some versions of the GNU assembler do not handle 64-bit
immediate operands containing arithmetic.  Writing the
value out in full works correctly.

Signed-off-by: Mans Rullgard <mans@mansr.com>
libavcodec/arm/fft_fixed_neon.S

index 2cfd817..0508088 100644 (file)
@@ -56,7 +56,7 @@
         vhsub.s16       \r0, \d0, \d1           @ t3, t4, t8, t7
         vhsub.s16       \r1, \d1, \d0
         vhadd.s16       \d0, \d0, \d1           @ t1, t2, t6, t5
-        vmov.i64        \d1, #0xffff<<32
+        vmov.i64        \d1, #0xffff00000000
         vbit            \r0, \r1, \d1
         vrev64.16       \r1, \r0                @ t7, t8, t4, t3
         vtrn.32         \r0, \r1                @ t3, t4, t7, t8