Trying to add an __m128 using an and mask in SSE programming

Question

I am trying to use the result of a compare operation to add to an SSE variable. I have just realised that when using the _mm_cmplt_ps operation if the result is true it returns a NAN because 0xffffffff can't be represented which is of no use to me:

__m128 va;
__m128 vb;
__m128 result =_mm_set1_ps(0.0f);
vb = _mm_cmplt_ps(va,vb);
result = _mm_add_ps(result,vb);  // problem is that I would like to convert vb to 1.0's and 0.0's

Do you want to, for each component, end up with 1.0 if va was less than vb? — fun4jimmy, Apr 04 '13 at 10:56

Paul R · Accepted Answer · 2017-01-26T16:29:46.717

8

You can just apply the comparison mask to a vector of 1.0 values, e.g.

__m128  va, vb;
__m128  vcmp = _mm_cmplt_ps(va, vb);
__m128  vresult = _mm_and_ps(_mm_set1_ps(1.0f), vcmp);

edited Jan 26 '17 at 16:29

answered Apr 04 '13 at 10:59

Paul R

208,748
37
389
560

I would want the result to be a mixture of 1 .0 and 0.0 based on the value of the comparison? – user1850254 Apr 04 '13 at 11:04
`temp` is a mixture of 1.0s and 0.0s after it has been masked, so if that's all you need then omit the final addition. I've updated the code above now to reflect this. – Paul R Apr 04 '13 at 11:06

Trying to add an __m128 using an and mask in SSE programming

1 Answers1

Linked

Related