This intrinsic inserts a SEL instruction into the instruction stream generated by the compiler. It enables you to select bytes from the input parameters, whereby the bytes that are selected depend upon the results of previous SIMD instruction intrinsics. The results of previous SIMD instruction intrinsics are represented by the Greater than or Equal flags in the Application Program Status Register (APSR).

The __sel intrinsic works equally well on both halfword and byte operand intrinsic results. This is because halfword operand operations set two (duplicate) GE bits per value. For example, the __sasx intrinsic.

unsigned int __sel(unsigned int val1, unsigned int val2)

where:

The __sel intrinsic selects bytes from the input parameters and returns them in the return value, res, according to the following criteria:

if APSR.GE[0] == 1 then res[7:0] = val1[7:0] else res[7:0] = val2[7:0]
if APSR.GE[1] == 1 then res[15:8] = val1[15:8] else res[15:8] = val2[15:8]
if APSR.GE[2] == 1 then res[23:16] = val1[23:16] else res[23:16] = val2[23:16]
if APSR.GE[3] == 1 then res[31;24] = val1[31:24] else res = val2[31:24]

Example:

unsigned int ge_filter(unsigned int val1, unsigned int val2)
{
  unsigned int res;

    res = __sel(val1,val2); 
    return res;
}

unsigned int foo(unsigned int a, unsigned int b)
{
  int res;
  int filtered_res;

    res = __sasx(a,b);  /* This intrinsic sets the GE flags */
    filtered_res = ge_filter(res); /* Filter the results of the __sasx */
                                   /* intrinsic. Some results are filtered */
                                   /* out based on the GE flags. */
    return filtered_res;
}