[Mesa-dev] [PATCH 6/6] glsl: Generate IR for switch statements

[Dan McCabe]

On 06/20/2011 03:50 PM, Ian Romanick wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 06/17/2011 05:43 PM, Dan McCabe wrote:
>> Beware! Here be dragons!
>>
>>
> I think this will generate the wrong code for:
>
> 	for (i = 0; i<  10; i++) {
> 		switch (i) {
> 			default: continue;
> 		}
> 	}
>
> It seems like that will generate some approximation of:
>
> 	for (i = 0; i<  10; i++) {
> 		do {
> 			continue;
> 			break;
> 		} while (true);
> 	}
>
> Right?  This is why I had originally tracked loops and switch-statements
> in a single stack.  In this situation, you'd want to set a "do a
> continue" flag in the switch-statement and emit a break (from the
> switch).  Something like:
>
>
> 	for (i = 0; i<  10; i++) {
> 		bool do_cont = false;
>
> 		do {
> 			do_cont = true;
> 			break;
> 			break;
> 		} while (true);
>
> 		if (do_cont)
> 			continue;
> 	}
>
>
Yikes! Looks like you tripped over one of those dragons :(.

Using a do_cont variable (or similar device) feels kludgey to me. I've 
been mulling this over for the last couple of days while I did other 
things and I think the right way to do this might be to get rid of the 
use of "loop" altogether.

But the devil is in the details, which I haven't worked out yet. Going 
back to the unified loop/switch stack might be needed, though.

cheers, danm