00:50 anholt: coolcool. s390x incorrectly gives +0.0 for -1.0 * 0.0. love that for us.
00:54 airlied: gfxstrand, alyssa : know if anyone has thought about something like LLVM SCEV for NIR?
06:10 mareko: [A
08:42 digetx: alyssa: reviews and testing are welcomed for intel nctx