The OpenD Programming Language

_mm_cmpestra

So SSE4.2 has a lot of hard-to-understand instructions. Here is another explanation.

Alternative explanation of imm8

imm8 is an 8-bit immediate operand specifying whether the characters are bytes or words and the type of comparison to do.

Bits [1:0]: Determine source data format. 00: 16 unsigned bytes 01: 8 unsigned words 10: 16 signed bytes 11: 8 signed words

Bits [3:2]: Determine comparison type and aggregation method. 00: Subset: Each character in B is compared for equality with all the characters in A. 01: Ranges: Each character in B is compared to A pairs. The comparison basis is greater than or equal for even-indexed elements in A, and less than or equal for odd-indexed elements in A. 10: Match: Compare each pair of corresponding characters in A and B for equality. 11: Substring: Search B for substring matches of A.

Bits [5:4]: Determine whether to do a one's complement on the bit mask of the comparison results. \n 00: No effect. \n 01: Negate the bit mask. \n 10: No effect. \n 11: Negate the bit mask only for bits with an index less than or equal to the size of \a A or \a B.

Compare packed strings in a and b with lengths la and lb using the control in imm8, and returns 1 if b "does not contain a null character" and the resulting mask was zero, and 0 otherwise. Warning: actually it seems the instruction does accept \0 in input, just the length must be >= count. It's not clear for what purpose.

nothrow @nogc @trusted
int
_mm_cmpestra
(
int imm8
)
(
__m128i a
,
int la
,
__m128i b
,
int lb
)

Meta