12.8 __qsax intrinsic
This intrinsic inserts a
QSAX instruction into the instruction stream generated by the compiler.
you to exchange the halfwords of one operand, then subtract the
high halfwords and add the low halfwords, saturating the results
to the 16-bit signed integer range -215 ≤ x ≤
215 - 1.
unsigned int __qsax(unsigned int
, unsigned int
holds the first operand for the addition in the
low halfword, and the first operand for the subtraction in the high
holds the second operand for the addition in the
high halfword, and the second operand for the subtraction in the
__qsax intrinsic returns:
The saturated addition of the low halfword of the
first operand and the high halfword of the second operand, in the
low halfword of the return value.
The saturated subtraction of the low halfword of
the second operand from the high halfword of the first operand,
in the high halfword of the return value.
The returned results are saturated to the 16-bit signed integer
range -215 ≤ x ≤
215 - 1.
unsigned int exchange_subtract_and_add(unsigned int val1, unsigned int val2)
unsigned int res;
res = __qsax(val1,val2); /* res[15:0] = val1[15:0] + val2[31:16]
res[31:16] = val1[31:16] - val2[15:0]
/* Alternative equivalent representation:
val2[15:0][31:16] = val2[31:16][15:0]
res[15:0] = val1[15:0] + val2[15:0]
res[31:16] = val[31:16] - val2[31:16]