"move16 %a0@+,%a1@" and "fmovel (cpid=3) %a0@-,%fpcr" share the same opcode...
but QEMU executes move16 (and M68040 too).
You can try:
--------------8<--------------------------------------- move16.S
.data
src:
.long 0x01020304, 0x05060708, 0x090a0b0c, 0x0d0e0f00
dst:
.long 0, 0, 0, 0
.text
.globl _start
_start:
lea src,%a0
lea dst,%a1
.fopt id=3
fmovel -(%a0),%fpcr
move16 (%a0)+, (%a1)+
move.l #0,%d1
move.l #1, %d0
trap #0
--------------8<--------------------------------------- move16.S
m68k-linux-gnu-gcc -g -m68040 -nostartfiles -nodefaultlibs \
-nostdlib -o move16 move16.S
m68k-linux-gnu-objdump -d move16
move16: file format elf32-m68k
Disassembly of section .text:
800000b8 <_start>:
800000b8: 41f9 8000 20d4 lea 800020d4 <src>,%a0
800000be: 43f9 8000 20e4 lea 800020e4 <dst>,%a1
800000c4: f620 9000 move16 %a0@+,%a1@+
800000c8: f620 9000 move16 %a0@+,%a1@+
800000cc: 7200 moveq #0,%d1
800000ce: 7001 moveq #1,%d0
800000d0: 4e40 trap #0
qemu-m68k -d in_asm ./move16
----------------
IN:
0x800000b8: lea 0x800020d4,%a0
0x800000be: lea 0x800020e4,%a1
0x800000c4: fmovel (cpid=3) %a0@-,%fpcr
0x800000c8: fmovel (cpid=3) %a0@-,%fpcr
0x800000cc: moveq #0,%d1
0x800000ce: moveq #1,%d0
0x800000d0: trap #0
This patch backports the fix from binutils to only match FPU instructions
with coprocessor ID 1.
Laurent Vivier (1):
target/m68k: correctly disassemble move16
disas/m68k.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)
--
2.14.4
