Have people found that the gnu assembler (2.5.3) for the Alpha does
some pretty wierd things ?

For instance:

	.set noat
	[...]
	stl v0,(t1)

produces an error "insn expansion requires AT use, but `noat'
specified," but:

	.set noat
	[...]
	stl v0,0(t1)

does not. Why aren't they sematically equivalent?

Also, 

	jmp zero,0(pv)
                 ^

causes the assembler to dump core.

The OSF1 assembler does the right thing for all these examples.

--
Trevor Blackwell         tlb@eecs.harvard.edu          (617) 495-8912