[DEBUGGER] Assemble dialog box instruction proper disassembly and absolute addresses support #13468
Conversation
…he right disassembly instead of .li and the hex since the hex edition is available in "replace instruction" [DEBUGGER] Assemble dialog can now deal with absolute addresses computing automatically the offset. For example assembling b ->0x80359370 at the address 0x8035936c will result in the right relative jump. Add -> before any address you want to use as absolute to respect the disassembly syntax
|
@vyuuui Do you want to look at this? |
|
The reason I never did this was because disassembler output is often not valid assembly, both for this assembler and just in general. As an example, the disassembly dolphin gives for |
I did not think it was a big issue since you can delete the added part, and it is more useful than a .li but sure you have a good idea I can look into the disassembler code, add a boolean for verbosity. |
…recommendation.
1. Adjusted disassembly decorations when using assemble dialog.
- Right clicking and assembling "rlwinm r0, r0, 2, 0, 29 (3fffffff)" gets "rlwinm r0, r0, 2, 0, 29" in the dialog box.
- Right clicking and assembling "ps_add p6, p2+p4" gets "ps_add p6, p2, p4" in the dialog box.
2. Added support for pair single p1,...,p31 register for assembly of all ps instructions. "ps_add p6, p2, p4" works instead of "ps_add f6, f2, f4" required before
3. Added support for assembly of Graphics Quantization Registers for assembly. "psq_l p2, 0(r3), 0, qr0" can now be assembled for example
|
@vyuuui I made a pass to extend support as you suggested.
|
|
Wow thanks for doing this. Can we be sure the ps register order is always the same between the readable code and what the assembler expects? |
|
You mean for the ps_ instruction right? If yes, then yes. I tested as much as I could, the way I did it was by right clicking the ps instructions I would see, assemble them back and make sure the output had not changed. When I did not find them in the code, I assembled them, and made sure when disassembling the values were in the same order. Now I might have missed something, so open to any testing you recommend. |
| @@ -220,6 +222,8 @@ std::string_view TokenTypeToStr(TokenType tp) | |||
| return "GPR"; | |||
| case TokenType::FPR: | |||
| return "FPR"; | |||
| case TokenType::GQR: | |||
| return "QGR"; | |||
There was a problem hiding this comment.
Thanks for finding it. Just fixed it.
[DEBUGGER] Assemble dialog box instruction automatically filled with the right disassembly instead of .li and the hex since the hex edition is available in "replace instruction"
[DEBUGGER] Assemble dialog can now deal with absolute addresses computing automatically the offset. For example assembling b ->0x80359370 at the address 0x8035936c will result in the right relative jump. Add -> before any address you want to use as absolute to respect the disassembly syntax