09174: CPU Hitachi HD6309: cycle counts of branch-instructions of hd6309 cpu in native-mode are incorrect

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
09174	Core	Minor	Always	16 days ago	15 days ago

Tester	Peer	View Status	Public	Platform	MAME (Self-compiled)
Assigned To	Peer	Resolution	Fixed	OS	Windows 10/11 (64-bit)
Status [?]	Resolved			Driver
Version	0.277	Fixed in Version	0.278GIT	Build	64-bit
		Fixed in Git Commit	35b7abd	Github Pull Request #

Summary	09174: CPU Hitachi HD6309: cycle counts of branch-instructions of hd6309 cpu in native-mode are incorrect
Description	The cycle counts of the relative-branch-instructions of the hd6309 CPU in native-mode are the same as in emulation-mode. The only exceptions are the BRA and LBRA instructions. All others have the same cycle counts. However, the code in devices/cpu/m6809/base6x09.lst gives all branch instructions in native-mode one cycle less than in emulation-mode.
Steps To Reproduce	Just check the code in devices/cpu/m6809/base6x09.lst and compare their timing to the cycles counts listed in "Motorola 6809 and Hitachi 6309 Programming Reference" by Darren Atkinson.
Additional Information	I have a real Vectrex console with a real Hitachi hd6309 CPU inside, running a customized 6309 BIOS in native-mode, with all cycle-dependent routines adapted to the 6309 native-mode timing. I have compiled a custom mame version, which uses the hd6309 CPU inside vextrec.cpp, and which loads my custom 6309 BIOS file. This shows the simulation mismatch to my real system. Fixing the cycle-counts in devices/cpu/m6809/base6x09.lst leads to my mame version correctly emulating the behavior of the real console.
Github Commit

Flags	Verified with Code
Regression Version
Affected Sets / Systems	CPU Hitachi HD6309

Attached Files	base6x09.zip (2,544 bytes) 15 days ago Uploaded by Peer
Attached Files	vectrex.zip (6,621 bytes) 15 days ago Uploaded by Peer custom 6309 native mode BIOS for the Vectrex

No.23289 hap Developer 16 days ago edited on: 16 days ago	Could you attach your modified base6x09.lst file? (package it in a .zip if the website doesn't accept the file type) And if you're willing to share it, the custom Vectrex BIOS too.
No.23290 Peer Tester 15 days ago edited on: 15 days ago	Sure, and sorry for not thinking of this right away. You will find my patch in line 493. Note that this patch fixes the cycle count for all short relative branches but BRA and LBRA, that is still to do. Also see my comment in line 507 regarding the long branches, those are also still to do. I will attach my custom Vectrex 6309 BIOS next.
No.23291 Peer Tester 15 days ago edited on: 15 days ago	The vectrex.zip is to be put into the "roms" folder of mame and contains the file "exec_rom.bin", which is my 6309 native-mode BIOS version. It will say so on the boot screen of the Vectrex, and it will switch the 6309 to native mode automatically. Note that mame will complain about the checksum. If you want to try one of the Vectrex cartridges, use e.g. Scramble. That game will run visibly faster in 6309 native-mode compared to the regular 6809 (the scrolling is faster, and the game thus a bit harder). Please let me know if you have further questions or if I can be of more help.
No.23292 hap Developer 15 days ago edited on: 15 days ago	Ok, it's fixed. The docs are a bit confusing, until I realized that unconditional long branch opcode is 1 byte less than the conditional ones. So: short branch: always 3 cycles u. long branch: 5 cycles on 6809, 4 cycles on native 6309 c. long branch: 5 cycles, +1 cycle if branch is taken on 6809
No.23293 Peer Tester 15 days ago edited on: 15 days ago	Yes, the docs are a bit confusing. Regarding the long branches, just to be sure, from the docs I concluce that LBRN is always 5 cycles, be it 6809, or 6309 emulation-mode, or 6309 native-mode. Does that match your interpretation and your fix?
No.23294 hap Developer 15 days ago	Yes, 'long branch never' is a 4-byte opcode and is always 5 cycles.
No.23295 Peer Tester 15 days ago	Cool, thanks!