Is there a function in cgen for parsing short integers?

Is there a function in cgen for parsing short integers?

[Mehak Mahajan]

Hi,

I am porting cgen to a 16 bit architecture. The immediates are 16 bits 
long. I am also parsing all hex numbers as signed numbers. Hence 0xFFFF 
will be -1.
When I parse -1, I get 0xFFFFFFFF, since cgen_parse_signed_integer 
accepts a long.
My problem is that since the max size of immediate can only be 16 bits, 
I want to flag an error if the user enters an immediate like 0xFFFFFFFF. 
However, since -1 is parsed as a long, there is no way for me to 
restrict the length of the immediates to 16 bits.

Is there an equivalent function for parsing short integers as well?

Cheers,
Mehak

Re: Is there a function in cgen for parsing short integers?

[Dave Brolley]

You need to define your own parse handler for this operand. Take a look 
at the operand 'hi16' in cgen/cpu/openrisc.cpu and its parse handler 
'parse_hi16' in cgen/cpu/openrisc.opc. At the end, this particular 
handler chooses to extend the sign from bit 16. You could issue your 
error/warning at this point in the case of overflow.

I hope this helps,
Dave

Mehak Mahajan wrote:
> Hi,
>
> I am porting cgen to a 16 bit architecture. The immediates are 16 bits 
> long. I am also parsing all hex numbers as signed numbers. Hence 
> 0xFFFF will be -1.
> When I parse -1, I get 0xFFFFFFFF, since cgen_parse_signed_integer 
> accepts a long.
> My problem is that since the max size of immediate can only be 16 
> bits, I want to flag an error if the user enters an immediate like 
> 0xFFFFFFFF. However, since -1 is parsed as a long, there is no way for 
> me to restrict the length of the immediates to 16 bits.
>
> Is there an equivalent function for parsing short integers as well?
>
> Cheers,
> Mehak

Re: Is there a function in cgen for parsing short integers?

[Dave Brolley]

Why would you want to issue a diagnostic for -1? It's not out of range 
for a 16 bit literal. I would think you would be more interested in 
cases like 65536 or -65535. As for the user specifying 0xffffffff, I'm 
wondering if cgen_parse_signed_integer will indicate an error since (in 
C at least) that's out range of a signed 32 bit integer.

Mehak Mahajan wrote:
> Hi,
>
> Thanks for your reply. :)
> I am doing precisely what you have decribed.
> However, I need to differentiate between -1 which is represented as a 
> 16 bit number (0xFFFF) and -1 which is represented as a 32 bit number 
> (0xFFFFFFFF).
> Is there a way to do that?
>
> Cheers,
> Mehak

Re: Is there a function in cgen for parsing short integers?

[Dave Brolley]

Why would you want to issue a diagnostic for -1? It's not out of range 
for a 16 bit literal. I would think you would be more interested in 
cases like 65536 or -65535. As for the user specifying 0xffffffff, I'm 
wondering if cgen_parse_signed_integer will indicate an error since (in 
C at least) that's out range of a signed 32 bit integer.

Mehak Mahajan wrote:
> Hi,
>
> Thanks for your reply. :)
> I am doing precisely what you have decribed.
> However, I need to differentiate between -1 which is represented as a 
> 16 bit number (0xFFFF) and -1 which is represented as a 32 bit number 
> (0xFFFFFFFF).
> Is there a way to do that?
>
> Cheers,
> Mehak

Is there a function in cgen for parsing short integers?

[Mehak Mahajan]

Hi,

Thanks for your reply. :)
I am doing precisely what you have decribed.
However, I need to differentiate between -1 which is represented as a 16 
bit number (0xFFFF) and -1 which is represented as a 32 bit number 
(0xFFFFFFFF).
Is there a way to do that?

Cheers,
Mehak