intel assembly TEST PF flag operation

问题

I'm was doing a manual operation with TEST (Parity flag operation), the problem it's that i can't get the right result, consider this:

ax = 256 = 0000 0001 0000 0000

so if i do:

test ah, 0x44

the PF flag operation should be:

0000 0000 = 0000 0001 & 0100 0100

PF = 0000 0000 XNOR 0000 0000

PF = 1111 1111?

I've followed the intel reference, according to this:

the question it's what i'm doing wrong?

回答1:

BitwiseXNOR is a horizontal XNOR of the bits, producing a single bit. Remember that PF is only 1 bit wide (a flag in EFLAGS), so it makes no sense to write PF=1111 1111.

---

It's the same parity calculation as always for instructions that "set flags according to the result", except that it's done on test's internal temporary result. (And as always, on the low 8 bits of it, regardless of operand size).

PF = 0 if the number of set bits is odd, PF = 1 if the number of set bits is even. So yes, the expression you posted in comments, ~(0^0^0^0^0^0^0^0) is correct.

See also:

• https://en.wikipedia.org/wiki/Parity_flag
• https://en.wikipedia.org/wiki/Parity_bit points out that parity is also the inverted (sum mod 2) of all the bits. (Because XOR is add-without-carry). Parity of an integer wider than 8 bits can be computed with popcnt eax, eax / not eax / and eax, 1. (Or use BMI andn eax, eax, ecx with ecx=1 set outside a loop to do the not-and part in one instruction.) Or just use the inverted parity value, where 1 = odd parity.

This answer has several examples of how PF is set from the horizontal-XOR of the bits in a result. In your case, the 8 bits are the AND result that test internally generates.

来源：https://stackoverflow.com/questions/47404539/intel-assembly-test-pf-flag-operation