SOP and FLAGS

[paiv]

Should SOP update FLAGS, or this is by design not touching FLAGS?

mtmc/src/main/java/mtmc/emulator/MonTanaMiniComputer.java

Lines 357 to 367 in 301477a

	case 0b0111 -> {
	// sop
	short aluOpCode = getBits(8, 4, instruction);
	switch (aluOpCode) {
	case 0b0000 -> {
	short currentTop = fetchWordFromMemory(getRegisterValue(stackReg));
	short nextDown = fetchWordFromMemory(getRegisterValue(stackReg) + WORD_SIZE);
	setRegisterValue(stackReg, getRegisterValue(stackReg) + WORD_SIZE);
	int value = nextDown + currentTop;
	writeWordToMemory(getRegisterValue(stackReg), value);
	}

[KyleDickersonComposer]

The way I think about it is that you reset the two ALU flags at the top of any ALU op. Then the ALU op sets them and you can do a test op to check for the flags as needed? Also, as far as I can tell there is no way to check overflow or nan or error in the spec? It seems like the test ops only set the test bit after comparing a thing in a register with another register or the imm test ops comparing the thing in the register to the thing in the fourth nibble. There doesn't seem to be a way to access the flags so that the user can do conditional things with them. It looks like there are four more slots in the test nibble. Maybe you could add four more ops into the test nibble that checks if a flag is set and sets test flag to true if so.

[jbanes]

no way to check nan

Without a floating point unit, there is no NaN to be checked

no way to check overflow

I believe this was an intentional decision. I went back and forth on it myself, until I realized that Sea (like most C compilers) won't support it. So no overflow checking.

Carson would have to speak to this further, but I believe the display for the overflow flag is mostly informational and possibly a carryover from early design work.

Should SOP update FLAGS, or this is by design not touching FLAGS?

Only test instructions currently update the flags. So I believe this is by design.

[paiv]

Only test instructions currently update the flags.

Here is the ALU (0001) branch:

mtmc/src/main/java/mtmc/emulator/MonTanaMiniComputer.java

Line 303 in ec5cadd

setFlagTestBit(result != 0);

From the spec:

mtmc/docs/MTMC_SPECIFICATION.md

Lines 134 to 135 in ec5cadd

	If the result of an ALU operation is zero the `test` bit of the `flag` register will be set to `0`, otherwise it will
	be set to `1`

Since SOP "applies the ALU operation", I would expect it updating the FLAGS too. It would be useful if not required in stack languages. In the current implementation this is missing.

Potentially, MISC operations INC (0010) and DEC (0011) should be reviewed too.

In any case this should be clarified in the spec.