Add ability to reconstruct the instruction from mapped json file

[Kuree]

Once the application is mapped, there is no way to reconstruct the peak instruction from the mapped json file.

It would be great to add the ability to reconstruct the instruction since consumers other than the the one that generated the json can infer the application data. However, if this kind of behavior is intentional for some reasons, feel free to close the issue. I think I can do some hacks with the debug string.

[rdaly525]

The flow that we agreed upon was that the instructions would remain in memory and would not be given to PnR. PnR should not care about the instruction. The debug string is for debug purposes and should not be reconstructed. There is an in-memory map from PE names to peak instruction objects that is passed to garnet.

Perhaps if we want in the future, we can have a direct instruction input to the PE where each PE instance will have a constant value input representing the instruction, but again this should not matter to PnR.

[Kuree]

I'm not referring to PnR. I'm working on a functional model that takes mapped JSON file and PnR result to simulate the CGRA. If you think everything should be in memory, I think I can work on a memory dump from the mapper so that it can be reloaded into the model compiler.

…

On Sun, Apr 14, 2019, 00:50 Ross Daly ***@***.***> wrote: The flow that we agreed upon was that the instructions would remain in memory and would not be given to PnR. PnR should not care about the instruction. The debug string is for debug purposes and should not be reconstructed. There is an in-memory map from PE names to peak instruction objects that is passed to garnet. Perhaps if we want in the future, we can have a direct instruction input to the PE where each PE instance will have a constant value input representing the instruction, but again this should not matter to PnR. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#14 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AF0QzXjlK5EohAh6Z_GXqD9uscRoTJsFks5vgt3LgaJpZM4cuZgt> .

[rdaly525]

If we want everything in the mapped json, we should go the other route I proposed which was to basically include the assembled instruction as a constant input to the PE. Then you can disassemble it to recreate the peak instruction.

[rdaly525]

PnR will need to know that it should ignore the instruction constant though.

[Kuree]

Build on top of your proposal, can we include some "magic" constant in the mapped json? This constant can be read and feed into peak to construct instruction for the simulator. As a result, neither does the simulator nor PnR know what's the functionality of a particular PE instance.

[rdaly525]

I think what you are asking for is a disassembler. A disassembler should be able to take in an assembled instruction and return the peak instruction associated with it. Is this correct or are you asking for something else?

[Kuree]

I believe so. Is there any support for disassembler in peak/mapper now?

[rdaly525]

I know that @cdonovick created a simple disassembler. Not sure exactly what the capabilities are.