> What worries me more is something like Nvidia's Denver architecture because that's actually a full abstraction above machine code.
What relieves me is that RISC-V is demonstrating that opensource hardware can be a reality. This will provide people with a concrete alternative to Intel and ARM chips that does away with (closed) microcode and shadowy marketing/security procedures.
Yes, we are far away from FPGA-ing our CPUs, but I see the years 2010 as the years 1980 for FLOSS code. In the '80s a bunch of people sent around tapes with copies of EMACS, cp and mkdir; 20 years later multibillion dollars infrastructures rely on open source.
Today we use Intel microprocessors fearing that their ME components will spy on us and that secure boot will lock us in. In some years we will just install our Debian-for-HW and be done with it. It is a liberating thought.
Yeah FPGAs sound nice but I seriously doubt that will ever happen where performance and power are major concerns.
I'm not sure that just having the hardware design makes much of a difference either as it'd be the equivalent of shipping a binary blob then making the source available without the compiler used or a way to decompile because the tools required are so expensive.
In CPU land have ARM which is probably as close as we can practically get to open source but IMO the real problem is still verification of the shipped product and without third parties auditing the design I'm not sure we could ever be sure source or not.
the real problem is still verification of the shipped product
In that vein, there is a lot about a modern FPGA that's a "binary blob". Do you fully trust Xilinx? Do you fully trust Altera?
Do you continue to trust Altera now that they are a subsidiary of Intel? If so, why don't you just trust Intel directly? Perhaps it's turtles all the way down? :)
Open source hw maybe the ejection seat from black boxes (tinfoil and real concerns) and proprietary facism, could solve much in the same way marginalized Unices. The issues are funding, talent wrangling and practical desktop lithography (FPGA is a stepping-stone).
I read an article* about ARM using imprint lithography to create a working Cortex M0 microcontroller. Feature size is equivalent to a 2 um process (which was achieved about 30 years ago in silicon), but it still seems pretty exciting - I could find use-cases for something like that in my own day-to-day hardware projects if hobbyists ever find a way to recreate these kind of machines for cheap (much like the DIY CNC machine & 3d printing scene has done)
What worries me more is something like Nvidia's Denver architecture because that's actually a full abstraction above machine code.