[Arc][Sim] Lower Sim DPI func to func.func and support dpi call in Arc

[uenoku]

This PR implements initial support for lowering Sim DPI operations to Arc.

• sim::LowerDPIFuncPass implements lowering from sim.dpi.func to func.func that respects C-level ABI.
• acc::LowerStatePass is modified to allocate states and call functions for sim.dpi.call op.

In the follow-up I'll rename sim.dpi.call to sim.func.call since now sim.dpi.call accepts both sim.dpi.func and func.func.

[sequencer]

I think in the future. We may require the ClockGen intrinsic for triggering clock and reset. This is the only Verilog simulation blackbox that t1 use in the Verilator/VCS. It should be lowered to both Verilog and ArcOp.

[maerhart]

LGTM! Really cool that arcilator can deal with DPI calls now! 🎉

[fabianschuiki]

This is really really cool! Thanks for also adding an integration test with Arcilator's JIT to actually give this a practical check 🙇 🙌 🥳

[fabianschuiki]

Just out of curiosity: since the sim.func.dpi.call function seems like it accepts both sim.func.dpi and func.func callables, would this code also work with a second sim.func.dpi.call that calls the @adder_func directly?

[uenoku]

Directly calling @adder_func doesn't work since it's output value is passed by reference in argument. If @adder_func was a normal dataflow-ish func.func such as func.func @adder_func_ok(%arg0: i32, %arg1: i32) -> i32, we can call it directly.

[fabianschuiki]

Ooooh you're right, sorry, I overlooked that difference in the operand layout. Makes sense!