OSNews

yeah... academic groups produce new chip ideas now and then. See Sun's Niagra, for example. It was developed out by Sun, but the initial version was done by a Professor and his grad students.

Intel's Terrascale was also made by a small group. Look up Cavium's OCTEON as well.

Smaller experimental chips are more likely to need compiler support to test out their design ideas before going into mass production. If the researcher has a backend IR with some non-chip-specific optimizations, she has a great basis to work with for designing a specialized processor.

I think the cost of writing compiler backends would be insignificant compared to the cost of developing new chips. Do you have reasons to believe the contrary?

It depends on what you're trying to do.

If you're trying to create a clockless chip, I'm sure the chip development is far more expensive than the compiler development.

But look at the Itanium. It's a chip with a lot of theoretical performance. Getting that improvement requires vastly more work from compilers than on other architectures. No one has really managed to make the chip shine yet.