i#3544 RV64: Added an encoder and some fixes and improvments to the decoder

[ksco]

This patch added the following changes:

1. Added an encoder
2. Fixed various disassemble issues
3. Optimized disassembled immediate format
4. Added unit tests for the encoder and decoder

Issue: #3544

[ksco]

I wrote a simple test case in the above commits to check if things work ok, now I can use ctest -L RISCV64 to run the api.ir test, but the test itself doesn't work (Segfaulted due to uninitialized our_environ variable used by standalone_init).

I noticed that the RISC-V version of api.ir binary doesn't contain the _init symbol, so some initialization routines are not being called before main.

$ # RISC-V
$ riscv64-linux-gnu-objdump -d ./suite/tests/bin/api.ir | grep "<_init>"
$
$ # x86_64
$ objdump -d ./suite/tests/bin/api.ir | grep "<_init>"
0000000000003000 <_init>:
$

I'm guessing it's a CMake mistake, but I'm not familiar with CMake, do you know what's possibly wrong? @abhinav92003 @derekbruening, thanks.

[derekbruening]

I wrote a simple test case in the above commits to check if things work ok, now I can use ctest -L RISCV64 to run the api.ir test, but the test itself doesn't work (Segfaulted due to uninitialized our_environ variable used by standalone_init).

I noticed that the RISC-V version of api.ir binary doesn't contain the _init symbol, so some initialization routines are not being called before main.

$ # RISC-V
$ riscv64-linux-gnu-objdump -d ./suite/tests/bin/api.ir | grep "<_init>"
$
$ # x86_64
$ objdump -d ./suite/tests/bin/api.ir | grep "<_init>"
0000000000003000 <_init>:
$

I'm guessing it's a CMake mistake, but I'm not familiar with CMake, do you know what's possibly wrong? @abhinav92003 @derekbruening, thanks.

If there's no init what is the entry point then? Is there no libc setup code? Missing cross toolchain pieces? I would build a simple C app with a main directly and if it works I would examine its entry point and startup code and then start diffing the build commands between it and api.ir.

[ksco]

Thank you! According to what you said, I noticed RISC-V gcc does not generate _init/_fini at all, I don't why, maybe they're too old and have been replaced by init_array/fini_array? Anyway, after the above commit, it seems working.

[ksco]

I've added unit tests for the encoder and decoder and also enabled them in CI. Please review again, thanks!

[derekbruening]

Thank you for adding tests! Just a few comments/suggestions.

[derekbruening]

Thank you for making the requested changes, it looks good.

Alrealdy reviewed.

[ksco]

@derekbruening Hi, I successfully ran a simple hello world on RISC-V, but the size of the patch is relatively large. Is it better to split it into small PRs and submit them separately?

This is probably not the right place to ask this kind of question, please let me know if there are other recommended ways.

[derekbruening]

@derekbruening Hi, I successfully ran a simple hello world on RISC-V, but the size of the patch is relatively large. Is it better to split it into small PRs and submit them separately?

This is probably not the right place to ask this kind of question, please let me know if there are other recommended ways.

Smaller is always better: easier and faster to review; easier to isolate and revert if something goes wrong.