[PATCH] AArch64 Neon scalar saturating accumulated of unsigned value instructions
Tim Northover
t.p.northover at gmail.com
Tue Oct 8 08:00:54 PDT 2013
Hi Chad,
+ def bb : NeonI_Scalar2SameMisc<u, 0b00, opcode,
+ (outs FPR8:$Rd), (ins FPR8:$Rn),
I think there should be a second input operand for all of these, tied
to the output. Otherwise what's being added? This affects the patterns
and the tests (and Clang) too.
A few tests in MC/Disassembler/AArch64 would be useful too. Not that I
think they're wrong now, but just in case someone does some weird
poking around in future.
Cheers.
Tim.
More information about the llvm-commits
mailing list