[libre-riscv-dev] [llvm-dev] RFC: First-class Matrix type

[Luke Kenneth Casson Leighton]

ok i think i got the picture, now.  a second level of redirection
(with additional CSRs) that specify that instead of a linear for-loop
around VL, the element offset is put through a 2D "remap" function,
based on (extra) CSRs for that register.

that should actually be quite straightforward to add... and would be
good to document... however implementation-wise can we get everything
else in place first? :)  the element-width thing really needs to be
resolved, first.

l.