These languages provide modules which are inherently concurrent and can handle asynchronous messages pretty neat (through ports). Keeping aside the fact that they cannot spawn module instances at runtime, do they qualify as actor based programming languages?
Thanks
EDIT:
What I'm really looking for is how well the language semantics can be used to "model" actors, rather than how a simulator would handle the code (of course, they are all event-driven underneath; and further down, we end up with transistors :-) ).
So if I create a bunch of Ip4Routers like this,
module Ip4Router (
inout interface_t intrf1, // Port that can atomically send/rcv packets
inout interface_t intrf2, ...
);
always @(intrf1) begin // Activity on intrf1
if (intrf1.valid & intrf1.ipaddr != myaddr && ...) begin
intrf2.valid <= .. // atomically bang data on intrf2 + other stuff
end
end
//...
endmodule
module test;
interface_t intrf1[1001];
Ip4Router r1(intrf1[0], intrf1[1])
Ip4Router r2...; // Create and wire up 1000 routers in some topology...
endmodule
would the routers qualify to be (miniature) actors?
I remember when I was first starting out I did some basic design work. I was always trying to fix problems in my design by fixing the code and then my mentor told me "you're not supposed to program here, you're supposed to describe what hardware is going to get synthesized from this code".
Verilog and VHDL are hardware description languages (HDLs) and when using them you think differently than you would when you program in some other language. I would avoid calling them programming languages (though they do also have constructs than aren't meant to be synthesized, but that's for the purpose of creating testbenches).
While these HDL's are not inherently designed as actor oriented languages, there has been multiple efforts to model abstract asynchronous message-passing using them. There are various examples of modeling Communicating Sequential Processes (CSP) in Verilog/SystemVerilog/VHDL/SystemC. Although they are meant to be used to design hardware and test environments, with a little bit of effort they can be used to "mimic" such higher level modeling language. The obvious reason for doing this is to take advantage of the existing powerful compilers and simulation tools for these language and not creating yet another new language.
For example, "SystemVerilogCSP" models CSP-like communication channels and abstract message passing using SystemVerilog's interfaces. While the communication channels are implemented using handshake protocols, from the user's perspective of this package, they are just atomic and blocking actions. See Figure 2 in this paper, in which two concurrent modules communicate through blocking CSP-like communication actions.
Moreover, Verilog and SystemVerilog can spawn parallel threads using fork-join constructs.
No. The execution semantics of these languages are event driven. They use a stratified even queue which makes execution on multi-core or other parallel architectures extremely difficult. An event is a change on a signal or some abstract message, and all event go on to a queue to be distributed (or fanned out) to all processes waiting on the event. Only one process is executed at a time.
Although you can synthesize these HDL descriptions into another form that might be actor oriented, the actual source cannot is not.