SPRUI04F july 2015 – april 2023
Table 7-9 lists the built-in functions that can take vector arguments and return a vector result. Unless otherwise specified, the function is applied to each element in the vectors. That is, each element in the resulting vector is the result of applying the function to the corresponding elements in the source vectors. Unless otherwise specified, the vector types of all the arguments and the vector returned must be identical.
| Function | Description | Vector Types Supported |
|---|---|---|
| __hadd( x, y ) | Returns average using (x + y) >> 1. The intermediate sum does not modulo overflow. The vector types of x, y, and the vector returned must be identical. | short to short16 uchar to uchar16 |
| __rhadd( x, y ) | Returns average with rounding using (x + y + 1) >> 1. The intermediate sum does not modulo overflow. | short to short16 uchar to uchar16 |
| __max( x, y ) | Returns the greater value of x and y in each element of the returned vector. | short to short16 uchar to uchar16 |
| __min( x, y ) | Returns the lesser value of x and y in each element of the returned vector. | short to short16 uchar to uchar16 |
| __add_sat( x, y ) | Returns x + y and saturates the result. That is, if there is an overflow, the maximum value in range for the type is returned. | short to short16 ushort to ushort16 uchar to uchar16 int to int16 |
| __sub_sat( x, y ) | Returns x - y and saturates the result. That is, if there is an overflow, the maximum value in range for the type is returned. | short to short16 int to int16 |
| __abs_diff( x, y ) | Returns the absolute difference using | x - y | without overflow. | uchar to uchar16 |
| __abs( x ) | Returns the absolute value using | x | | short to short16 |
| __popcount( x ) | Returns the number of non-zero bits in x. | uchar to uchar16 |
| __mpy_ext( x, y ) | Extended precision multiplication. | short to short16 ushort to ushort16 char to char16 uchar to uchar16 |
| __mpy_fx_ext( x, y ) | Fixed point multiplication. | short to short16 |
| __conj_cmpy( x, y ) | Conjugate complex multiplication. | cfloat to cfloat8 |
| __cmpy_ext( x, y ) | Extended precision complex multiplication. | cshort to cshort8 |
| __cmpyr_fx( x, y ) | Fixed point complex multiplication with rounding. | cshort to cshort8 |
| __cmpy_fx( x, y ) | Fixed point complex multiplication. | cshort to cshort8 cint to cint8 |
| __conj_cmpy_fx( x, y ) | Fixed point conjugate complex multiplication. (C6600 only) | cint to cint8 |
| __crot90( x ) | Rotate by 90 degrees. (C6600 only) | cshort to cshort8 |
| __crot270( x ) | Rotate by 270 degrees. (C6600 only) | cshort to cshort8 |
| __dot_ext( x, y ) | Extended precision dot product using x • y. | short2 factors give an int result char4 factors give an int result uchar4 factors give a uint result short4 factors give an int result (C6600 only) short4 and ushort4 factors give an int result (C6600 only) |
| __dot_extll( x, y ) | Extended precision dot product using x • y. | short2 factors give a longlong result short4 factors give a longlong result (C6600 only) short4 and ushort4 factors give a longlong result (C6600 only) |
| __dot_fx( x, y ) | Fixed point dot product using x • y. | short2 • ushort2 = int |
| __gmpy( x, y ) | Galois field multiplication. | uchar4 |
| __ddot_ext( x, y ) | Extended precision 2-way dot product. | short2 • char4 = int4 short8 • short8 = int2 (C6600 only) short8 • ushort8 = int2 (C6600 only) |
| __mpy_fx( x, y ) | Fixed point multiplication. | short2 * int = int2 int4 * int4 = int4 (C6600 only) |
| __apply_sign( x, y ) | Uses the sign bit of x to determine whether to multiply the four 16-bit values in y by 1 or -1. Yields four signed 16-bit results. This function is an alias for the _dapys2() intrinsic. (C6600 only) | short4 |
| __cmpy_conj_ext( x, y ) | Extended precision complex multiplication conjugate. (C6600 only) | cshort2 * cshort2 = cint2 |
| __cmpy_conj_fx( x, y ) | Fixed point complex multiplication conjugate. (C6600 only) | cshort2 * cshort2 = cshort2 |
| __cmatmpy_ext( x, y ) | Extended precision complex matrix multiplication. (C6600 only) | cshort2 * cshort4 = cint2 |
| __conj_cmatmpy_ext( x, y ) | Extended precision complex matrix multiplication conjugate. (C6600 only) | cshort2 * cshort4 = cint2 |
| __cmatmpy_fx( x, y ) | Fixed point complex matrix multiplication. (C6600 only) | cshort2 * cshort4 = cshort2 |
| __conj_cmatmpy_fx( x, y ) | Fixed point complex matrix multiplication conjugate. (C6600 only) | cshort2 * cshort4 = cshort2 |
Prototypes for all the supported vector built-in functions are listed in the "c6x_vec.h" header file, which is located in the "include" sub-directory of your Code Generation Tools installation. Please see the "c6x_vec.h" for a complete list of the vector built-in functions.
The following example vbif_ex.c file uses the __add_sat() and __sub_sat() built-in functions with vectors:
#include <stdio.h>
#include <c6x_vec.h>
void print_short4(char *s, short4 v)
{
printf("%s", s);
printf(" <%d, %d, %d, %d>\n", v.x, v.y, v.z, v.w);
}
int main()
{
short4 va = (short4) (1, 2, 3, -32766);
short4 vb = (short4) (5, 32767, -13, 17);
short4 vc = va + vb;
short4 vd = va - vb;
short4 ve = __add_sat(va, vb);
short4 vf = __sub_sat(va, vb);
print_short4("va=", va);
print_short4("vb=", vb);
print_short4("vc=(va+vb)=", vc);
print_short4("vd=(va-vb)=", vd);
print_short4("ve=__add_sat(va,vb)=", ve);
print_short4("vf=__sub_sat(va,vb)=", vf);
return 0;
} Compile the example as follows:
%> cl6x -mv6400 --vectypes -o1 -k vbif_ex.c -z -o vbif_ex.out -llnk.cmdNote that the lnk.cmd file contains a reference to rts6400.lib. The rts6400.libs library contains c6x_veclib.obj, which defines the built-in functions, __add_sat() and __sub_sat().
Running this example produces the following output:
va= <1, 2, 3, -32766>
vb= <5, 32767, -13, 17>
vc=(va+vb)= <6, -32767, -10, -32749>
vd=(va-vb)= <-4, -32765, 16, 32753>
ve=__add_sat(va,vb)= <6, 32767, -10, -32749>
vf=__sub_sat(va,vb)= <-4, -32765, 16, -32768>