For example:
if (!(opcode&0xFF0)) /*No shift*/
{
addbyte(0x8A); addbyte(0x0D); addlong(pcpsr+3); /*MOVB
*pcpsr,%cl*/
generateload(opcode&0xF);
addbyte(0x80); addbyte(0xE1); addbyte(~0xC0); /*AND
$ZFLAG+NFLAG,%cl*/
return 1;
}
No shift but clears C flag.
Second example:
case 0x00: /*LSL*/
addbyte(0x8A); addbyte(0x0D); addlong(pcpsr+3); /*MOVB
*pcpsr,%cl*/
generateload(opcode&0xF);
if (temp)
{
addbyte(0x80); addbyte(0xE1); addbyte(~0xE0); /*AND
$ZFLAG+NFLAG+CFLAG,%cl*/
addbyte(0xC1); addbyte(0xE0); addbyte(temp); /*SHL
$temp,%eax*/
addbyte(0x73); addbyte(3); /*JNC nocarry*/
addbyte(0x80); addbyte(0xC9); addbyte(0x20); /*OR
$CFLAG,%cl*/
}
else
{
addbyte(0x80); addbyte(0xE1); addbyte(~0xC0); /*AND
$ZFLAG+NFLAG,%cl*/
}
return 1;
Here if temp is 0, there is no shift but it clears C flag.
If temp is non-0, current C flags is kept (it should be cleared) before the SHL
and set to 1 if required after the SHL (correct).
André
________________________________
From: [email protected] [mailto:[email protected]] On Behalf
Of Tom Walker
Sent: mardi 21 février 2012 21:04
To: [email protected]
Subject: Re: [Rpcemu] Code review
> I have also looked at the file codegen_x86.c.
>
>The function generateshiftflags() seems to be mishandling the C flag
>(at least if the comments match te actual opcodes used).
Can you point to specific examples of this? I've just had a look and every area
where this seems to be a problem is preceded with 'return 0;' - ie fall back to
interpreter code.
Tom
_______________________________________________
Rpcemu mailing list
[email protected]
http://www.riscos.info/cgi-bin/mailman/listinfo/rpcemu
