10:04:42hayleyHe used to be active on IRC, but I think he's now busy with PhD work. But if I get the time (I'm also busy with university) I can see about adding in a u8.32-string-ref, though that of course would only be usable from the SBCL release after I do that.
10:07:03kami_Is there any commit in the source tree which adds something like this?
10:08:29kami_Yes, just checked, sb-simd was committed in one chunk
10:09:07hayleyI have https://github.com/telekons/one-more-re-nightmare/blob/master/Code/SIMD/new-sbcl-x86-64.lisp#L236-L250 (see lines 70-80 for DEFINE-BORING-VOP) which you could use.
10:09:28hayley(Also note that it's in the SB-VM package)
10:11:27kami_Heh, was looking at one-more-re-nightmare two days ago. Thank you!
10:13:11kami_And I don't see any mention of vptest
12:24:04kami_when the first compilation occurs. Wait, I'll paste the stack
12:25:06hayleyMy understanding is VPTEST sets flags and not a register; DEFINE-BORING-VOP won't work there. I can't remember how to specify which flag the VOP "returns", but you'd need to specify the Carry flag.
12:25:38hayleyAnd so the instruction is (inst vptest a b)
12:30:39hayleyThe DEFINE-VOP form should have (:conditional :c) I think, and similarly the DEFKNOWN should return a BOOLEAN. A guess though, I don't have any conditional VOPs I wrote on me.
12:32:17kami_I'll check the vop sources. Thanks again