Simulator Experience at Heritage Hsimen Just over a year I started this project, Heritage / 1, and after many bumps I have managed to accommodate a stable architecture based on design principles that ultimately dictate exactly what I wanted to build and how.
My foray into the world of simulation, however, is much more recent, only a few weeks, I had read before about the issue in draft colleagues and I had never given due attention, specifically for two reasons: ( 1) I want to make a machine bodily non-virtual and (2) always seemed to me to write a simulator could be as complex, time consuming and prone to errors as the same as circuit simulation ... a waste of time in short-I said then.
But something happened: the design of the hardware came to be (almost) over, and consequently surprised me with a deep understanding of how Heritage / 1 works (... or work). Then came the desire to start writing software for it, including your operating system, which I call H1OS and is (ra)-oriented Batch Processing using magnetic tape as a storage medium, instead of disks. The problem is that the machine not yet exist and, worse, much still, a year-so maybe it is.
Meanwhile, the ideas on how to build an operating system and the desire to do so have been increasing and that is why I agreed to build this "Heritage / 1 virtual" or simulator called Hsimen.
Writing software is much faster to build circuits and is so in less than two weeks of gestation, Hsimen is now able to execute all the instructions (about 100) of Heritage / 1. A nice experience is that testing with programs Hsimen language test machine Heritage / 1, I found failures in such programs in the simulator code itself.
is not over yet. Lack simulate hardware interrupts, after which I will be able to simulate peripherals. I estimate that in two or three weeks, Hsimen v 1.0 will enter the production phase.
But even before birth, this little program is generating experiences. My idea for him was not always simulate the circuit at the gate, but only the behavior of the machine at the structural level. Had conceived, for example, the idea of \u200b\u200breading the set of instructions from a file in an array, so the first step to decode an instruction would found in the arrangement.
is a solution "program", contrary to my way of doing things because I tend to think more like an electrical engineer who as a programmer. The first trip I had in the large number of instructions to be processed and the second, the realization that the concept is out of "my style."
What I did then was to simulate the way the circuits Heritage / 1 decode the instructions. These are the operation code (located in the instruction register IR) as a "bit-map." The four most significant bits are the "class" depending on which two things happen: (1) is or is not an Operand Fetch Cycle, and (2) determine which circuit will address decode and execute the statement.
circuital
This level is quite simple matter of wiring is suitably different (16) outputs of register IR. In the case of the simulator is even easier: I just had to split the instruction code in different fields and make decisions based on values.
This kind of simulation software deployed closer to the circuit (even at the block level, not gates) has brought me pleasant surprises, such as inconsistencies in the design capture the set of instructions and (of course) correct immediately. It also has allowed me best study of how to handle flags.
The overall experience has been this: the simulator, which originally conceived as a tool for debugging software, has also served as a "model" or "virtual prototype" of the machine, through which I can finalize design details and still find fault with the existing design.
Hsimen is written in Borland Turbo Pascal 3.0 using an old Pentium machine running native MS-DOS 6.22. The machine does not use "mouse" ... I know that some will be surprised by this.
The interface is extremely rudimentary, made with pure ASCII characters in one color (green on black background). Could actually do something more "fancy" but the original flavor reminds me of my old days ... nostalgia is better, right?
The handling is very similar to Heritage / 1. I have not implemented (nor do I do now) typical facilities "debuggers" and "watch", "break points" etc. It is a matter of concept: Hsimen not a "debug" but a simulator, a virtual version of Heritage / 1, and the purpose is that he feels like that.
this simulator follow an assembler (HASM) after which they will be able to start writing the operating environment of Heritage / 1 ... before it gets built.
