On Sun, Apr 29, 2012 at 9:17 PM, R. Diez <[email protected]> wrote: > > >>> Yes, basically it should just be a register copy. >> [...] >> OK, I've confirmed this is a bug. > > Why don't we get rid of instructions l.extws and l.extwz for 32-bit > implementations of the processor? They've never worked according to the spec, > that means they've never matched or1ksim's behaviour. And they're just a > waste of FPGA flip-flops, as instruction "l.ori Rdest, Rsource, 0" will do > the same.
Sure, but what do you mean by "get rid of"? We still need to support it because it's a valid instruction in both 32- and 64-bit implementations. My change appears to fix the problem and essentially turns it into a register copy. Julius _______________________________________________ OpenRISC mailing list [email protected] http://lists.openrisc.net/listinfo/openrisc
