On 19 Jan 2002, Falk Hueffner wrote: > > > This works if I pass -Wa,-mpca56. But it will *not* work if I also use > > > -mcpu=ev56 for gcc, since gcc then emits .arch ev56 into the .s, and > > > as complains ev56 can't handle unpkbw. What should I do? Convince gcc > > > not to emit .arch? Convince as to prefer the command line option? Some > > > other suggestions? > > > > Basically, you can't use any of those MVI extensions on any processor that > > doesn't support it (>= PCA56, IIRC). That code should probably have a > > conditional around it that depends on an amask check of the processor's > > support for MVI. > > Yes, I know, and it is protected by amask. I just want it to > compile/assemble, but the assembler refuses to do so if it finds .arch > ev56.
So you want ev56 scheduling with pca56 instruction support to get this to assemble? IIRC, the scheduling code between the two processors doesn't differ...only it's instruction support. '-mcpu=pca56 -mno-bwx' may allow it to compile and assemble while disabling generation of code to use the BWX instruction sets (which aren't present on ev56, but are on pca56). C
