Clarify Make targets?

[sampsyo]

The README currently says to run make cosim to run an example. But currently, it doesn't look like cosim is a valid target.

More broadly, for using this stuff as an example for other people trying to write CUDA-Lite code, it would be nice to know what targets are available—it looks like there are several with different intents and different origins:

• analysis: Defined in some example-specific Makefiles to run multiple versions. (The concept of "versions" might also need additional explanation—it looks like the idea for all of these examples is to possibly have multiple drop-in alternative implementations of the same kernel. That's fine for full-fledged performance tests, but it's a little confusing if you're just trying to get one thing up and running from scratch. Maybe it would make sense to have one example without "versions" for simplicity?)
• stats: Actually not entirely sure what this does? (And default is an alias for stats?)
• kernel/<version>/kernel.riscv: Build the device-side code for a given kernel version.
• *.cosim (where * is the name of the example, i.e., $HOST_TARGET): Build the host-side program.
• <version> or kernel/<version>/*.cosim.log (where the latter is $HOST_TARGET again): Actually run the simulation?

In general, it would be awesome to have documented the expected targets do these things separately:

• Build the host code (for simulation).
• Build the device code.
• Run the simulation.
• Plot the statistics into a cool graph.

…because it's often super convenient to do those individually rather than in one fell swoop. (And I don't think they need fancy .PHONY target names! Just knowing what does what can really help orient new users.)

[drichmond]

I think this is fixed now?