VASM wrong opcode generated?

[fuzzylogic]

Hello guys,

I am playing around with VASM and came across something odd(?)

If I assemble

asl.b #1,d1

it will produce the opcode $D201 which is in my opinion add.b d1,d1 instead. If I take a look into the M68k manual I would expect opcode $E301 for asl.b #1,d1

Am I doing something wrong or is this a bug? I am using vasm 1.9f.

[ross]

Some optimizations are active by default (see manual).

To disable all: -no-opt on command line or opt o- in source.
You can disable and enable the various optimizations individually with other commands (always see manual).

[fuzzylogic]

WOW that was fast

Thanks ross , this fixed the issue.

[jotd]

vasm does that for a reason: for a 1 or 2 bit shift, it's faster to add the register with itself.

That said, I'm not sure that add.b d1,d1 is strictly equivalent to asl.b #1,d1 as there's the arithmetic part that isn't there in lsl.

[ross]

Originally Posted by jotd

vasm does that for a reason: for a 1 or 2 bit shift, it's faster to add the register with itself.

That said, I'm not sure that add.b d1,d1 is strictly equivalent to asl.b #1,d1 as there's the arithmetic part that isn't there in lsl.

Is the arithmetic part (V flag) of lsl that is not the same as add.
Therefore, in this case, the optimization is legitimate

But you are right about being careful with this optimization.

[roondar]

On a kinda-but-not-fully related question:
Can you disable these optimisations using opt -o for only part of a single file?
If so, how do you enable the default stuff again, opt +o?

[ross]

Sure!
You can insert opt in the source only for the part of the code on which you want to enable or disable that optimization.
(but the code may become less portable..)

[phx]

Originally Posted by ross

But you are right about being careful with this optimization.

By default vasm should only enable optimizations which are absolutely safe and without any side effect. Otherwise it is a bug (please tell me)!

[Don_Adan]

Yes, default asembler optimisations its always bug for me. I remember ROM original module code, perhaps from wb.find. When jmp was replaced with bra.b (if I remember right). And it was bug.
Anyway for this case i dont see reason for use asl command, for me is totally useles. Lsl is much logical for me.

[ross]

Originally Posted by phx

By default vasm should only enable optimizations which are absolutely safe and without any side effect. Otherwise it is a bug (please tell me)!

Yeah, I was referring to an explicit lsl optimization
If vasm optimize by default asl there is a good reason