[llvm-dev] [cfe-dev] RFC: First-class Matrix type

[David Greene via llvm-dev]

Richard Smith via cfe-dev <cfe-dev at lists.llvm.org> writes:

> However, I do wonder: how is this different from, say, complex
> numbers, for which we don't have a native IR representation? (Maybe
> the answer is that we should have native IR support for complex
> numbers too.)

We've got at least one person here with decades of experience handling
complex operations in compilers who thinks LLVM IR should absolutely
have native support for complex, FWIW.  I don't have enough experience
with it myself to make a definitive statement but I can certainly see
advantages to it.

Most of us here also think that LLVM IR should absolutely have native
support for predication/masking.  :)

As for matrices, I again can see the advantages but have no practical
experience to draw upon.  But some people at Apple seem to think it's
advantageous and I'm interested in learning more.

                           -David