pon., 15 cze 2020 o 11:05 Maarten Brock <[email protected]>
napisał(a):
> This sounds like an excellent job for the peephole optimizer.
> Are there restrictions on the operand that should be checked?
>
> Maarten
>
I have created following peephole rule for my if statement:
replace {
mov a, __pa
and a, #0x10
cneqsn a, #0x00
goto %1
} by {
; peephole PA4 check optimization.
t1sn __pa, #4
goto %1
}
I didn't use a pattern variable as __pa because i want to have more
control, at least for now. The problem for me is that in C bit position is
specified as bitmash ( byte & 0x10) while tXsn takes bit number (#4) and I
don't know how to create a rule that converts bit number as bitmask to bit
number. I could create 8 rules foreach tXsn with different bit numbers :/
pon., 15 cze 2020 o 17:24 Philipp Klaus Krause <[email protected]> napisał(a):
> But I had thought about having __near as a subspace of __data, for the
> lower half of the __data address space.
>
> Philipp
>
I didn't know about that, lower half. So checking variables thru tXsn
might fail ? If i use following rules can they make problems ?
// if( !(byte & 0x80) ) PULLSDA(); //Pull SDA low if MSB is 0
replace {
mov a, _SendByte_PARM_1+0
and a, #0x80
ceqsn a, #0x00
goto %1
} by {
t0sn _SendByte_PARM_1+0, #7
goto %1
}
// if( (byte & 0x80) ) RELEASESDA(); //Release SDA if MSB is 1
replace {
mov a, _SendByte_PARM_1+0
and a, #0x80
cneqsn a, #0x00
goto %1
} by {
t1sn _SendByte_PARM_1+0, #7
goto %1
}
Slawek
PS. Another rule that i'm using
// I2CData = (I2CData << 1) | (SDA);
replace {
mov a, _I2CData+0
mov p, a
sl p
mov a, __pa
and a, #0x01
or a, p
mov _I2CData+0, a
} by {
; peephole SDA pulling optimization
set0 __flag, #1 //Clear carry flag, nedeed ?
t0sn __pa, #0
set1 __flag, #1
slc _I2CData+0
}
_______________________________________________
Sdcc-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sdcc-user
