Simplify sign-extension of immediates during instruction decoding

[cnlohr]

Simplify sign-extension of immediates during instruction decoding

[cnlohr]

Woah... I never thought about that! That's very cool. I won't be able to get to this in the next day or so - I will check into it soon after.

[cnlohr]

For completeness sake, this is the correct phrasing for the second one:

uint32_t immm4 = ((ir & 0xf00)>>7) | ((ir & 0x7e000000)>>20) | ((ir & 0x80) << 4) | ((uint32_t)((int32_t)ir >> 19) & 0xfffff000);

I am at a bit of an impass. I like the code the way you put it better, but doing optimizations takes the code from 15036 compiled to the same size for the second optimization (And appears to be the same code)... and if I do the first optimization, the code 15148 bytes!! And diffing the .lst's, I can't even figure out why. The compiler just like loses its mind when it goes through a signed typecast.

The first seems like it should be such an obvious win, but the compiler's fighting me and it's got hands.

[cnlohr]

#44