In the emulator

Reading the binary tells you what the code does. Some things only tell you what the code is worth once it runs, and that is what openMSX is for. Everything on this page is measured, and the raw tables are in medidas/.

The house rule

The scripts in tools/omsx_*.tcl only read memory. The ROM is never touched: no patches, no changed bytes, nothing forced to execute.

The one exception is declared and is a measurement, not a modification: tools/omsx_prueba_n.tcl stops the machine at 0x4220 —right after PARTIDA_NUEVA (0x4218) sets the round and the event to 1— and writes into 0xE016 the number of the event you want to see. 0xE016 is the game's own variable, and from there on the cartridge runs by itself, with its data and its code. What you save is having to qualify in the three previous events to reach the fourth.

The four events

With that, they were photographed one by one (medidas/pruebas.txt, captures in docs/imagenes/):

0xE016	label that comes up	record it paints
1	100 METER DASH	09 SEC 95
2	LONG JUMP	08 M 90
3	HAMMER THROW	83 M 98
4	400 METER DASH	43 SEC 86

All four records match byte for byte the table at 0x5174, which INIT copies to 0xE040: 00 09 95 / 00 08 90 / 00 83 98 / 00 43 86. So what is on screen and what the binary says are the same thing, checked from both ends.

The clock was worked out for 60 Hz

This is the measurement that most changes what you think of the game.

SUMA_AL_RELOJ (0x5695) adds to the BCD counter at 0xE0A9-0xE0AB once per frame, alternating 0x0167 and 0x0166. If the top byte is seconds and the middle one hundredths, that is 1.67 and 1.66 hundredths: an average of 1.665, which is one sixtieth of a second to four figures.

tools/omsx_reloj.tcl runs the 100 metres mashing the space bar and samples 0xE0A9-0xE0AB once per emulated second. On two machines:

machine	Hz	game clock per real second
Philips VG-8020 (PAL)	50	0.8325 s — runs 16.75 % short
C-BIOS MSX1 JP (NTSC)	60	0.9990 s — spot on

Both figures come from subtracting two consecutive samples in reloj_pal.txt and reloj_ntsc.txt, and they agree with the arithmetic: 50 × 1.665 = 83.25 and 60 × 1.665 = 99.9.

The consequence: on a European MSX the times the cartridge shows are not seconds. A 100 metres timed at 12.00 by the game took 14.4 real seconds. The qualifying marks at 0x5008 and the world records at 0x5174 are meant for the Japanese machine.

And mind reading it backwards. That is a period, not a speed: the game does not run slower in Europe —the athlete moves across the screen at whatever speed 50 Hz allows— what happens is that the stopwatch counts short. Those are two different claims and only one of them is measured.

The demo: who is really in charge

When nobody touches anything, the cartridge falls into a demo that plays itself. The pilot is INT_CON_MANDOS (0x4027) dropping into 0xE00B a value from ld a,r masked to two bits, once every four frames: random keypresses.

Reading the listing, the switch looks like bit 7 of 0xE029, because 0x5849 sets it and 0x4030 reads it. tools/omsx_demo.tcl samples both situations without touching anything, first letting the demo run and then starting a real game, and this comes out (medidas/demo.txt, abridged):

	0xE002	0xE029
demo	01	0x82
game	00	0x82

0xE029 holds the same value in both. The one in charge is 0xE002, which is the first thing the interrupt looks at: if it is zero it reads the real controls, and if not it makes the keypresses up. In a real game the made-up ones never reach 0xE00B.

It is a good example of why you have to measure: reading the code, the bit 7 hypothesis was reasonable, well built, and false.

Four lines, and what each one leaves in memory:

line	0xE01B
1PLAYER with JOYSTICK	1
2PLAYERS with JOYSTICK	2
1PLAYER with KEYBOARD	3
2PLAYERS with KEYBOARD	4

EMPIEZA_LA_PARTIDA (0x41BC) splits it up: bit 0 of 0xE01B gives the number of players (0xE010) and the value being below 3 gives the controller (0xE004). One variable with both answers inside it.

How to repeat it

openmsx -machine Philips_VG_8020 -cart hyperolympic1.rom \
        -script tools/omsx_reloj.tcl
PRUEBA=3 openmsx -machine Philips_VG_8020 -cart hyperolympic1.rom \
        -script tools/omsx_prueba_n.tcl