Dev Tracker - Discussion

It wouldn't surprise me if they used asm stubs for arithmetic just out of habit. Most of them in the late 90s would've learned in environments where that sort of thing mattered and would've trusted the compiler to manage it for them a lot less. The compiler might've handled it for them fine but they still did it out of habit, like how a lot of C developers use macros when inline functions would be just as or more suitable (guilty) for no particular reason. Sometimes developers just do things out of habit and being used to seeing it, but they might also have done it out of a desire to avoid any weird compiler behavior. ***was wild in the late 90s.

Bit shifts are still used by allocators or anything doing aligned arithmetic on pointers. I'd define that as pretty rare but mostly because most modern C code is hot garbage 99% of the time unless it came from Microsoft, Google, or Apple where all the old crusty nerds work who complain about how some functions are slower than others because they have one added instruction.

Haha yeah back then lots of stuff like this was common. The PSX ran on a 33Mhz CPU with 5KB of cache, 2MB of "system" RAM and 1MB of "video" RAM. To get a game like FFVII to run on that with those kinds of graphics required squeezing every ounce of performance via tight code optimizations. The PS2, which FFXI was coded on, and didn't have any Microsoft Visual Studio compiler for, had a 299Mhz 64-bit MIPS CPU, 16KB of instruction cache, 8KB of data cache, 32MB of "system" memory, 4MB of "video" memory and 2MB of "sound" memory.

Every time we dig into any of this games math we always stumble into where percentage numbers are really integers over 1024. Haste isn't 15% (0.15) but rather 150/1024. WS waist / belts are not +10% WS damage but really +100/1024 each. That wouldn't make sense unless the WS's fTP values were themselves coded as 1024, making it a simple addition that costs 1 cycle to do. These sorts of low level optimizations would be the bread and butter of console game developers.

Like I'd bet that Fimbulvetr's fTP at 3K isn't 9.9 but rather 10138, which would actually account for some of the discrepancies that pop up occasionally. 9.9 doesn't convert cleanly into an integer over 1024, instead turning into 10137.6. 10137 turns into 9.8994, while 10138 becomes 9.9003, and even 10140 becomes 9.9023. The 60% STR is probably 614, which is 0.5996 or 615 at 0.6005.

For percentages this is really important as divides are crazy expensive, like orders of magnitude more. While a multiply can take 2~4 cycles, a division takes 32 to 89 cycles. All multiplication by a percentage becomes a division but you can avoid that by turning it into a pure multiplication.

A: 450 STR
B: 60% STR WSC (0.60)
450 * 0.60 = 270 (this requires a DIV instruction)

Or with shifting
A 450 STR
B: 60 (shift two places)

450 * 60 = 27000, shift two places to 270.

Those theories always sound sensible, but damage calculations take place on the server where SE is far less resource constrained.

The guys who wrote it almost certainly also wrote the client. Old coding habits die hard and I'm fairly sure this wasn't even built on x86. Japan early 2000's was heavily dominated by Fujitsu SPARC64 and HP PA-RISC.

It's likely long since been moved to another platform, but I doubt they refactored these kinds of low level code optimizations.

And it's a lot more then a theory, what I just described is the exact coding patterns you see in older console games. Lots of using fixed binary digit values to reduce the number of expensive multiplys and crazy expensive divides.

Sure, but sometimes they mix in IEEE standard float math because the server isn't actually resource constrained

Oh never said they didn't, if you have a good FPU then both multiplys and divides can be done in half the time while preserving accuracy. Of course a FPU divide is still something like 10~36 cycles so still expensive and if you don't care about decimal place accuracy, you can cut it out entirely. Of course that FPU might be busy doing something else, like assisting in graphics rendering (this is what the PS2 did).

The OG coders who wrote the server engine almost certainly were console developers and it shows because of some of the weird values that come out when we test this games mechanics. Devs who joined much later, after the OG dudes moved on, likely didn't bring the same coding techniques with them but also didn't rewrite the old functions to match modernish conventions.

Realistically speaking none of these optimizations are required, but likely a result of coding habits. Think of your own habits you learned earlier in your career and how those habits shape how you approach projects. For a bunch of dudes who spent their entire career programming in the quasi assembly language that consoles used, this would of seemed "normal".