Flop Count Calculations + x86 Measurements + c13 Fix
This commit is contained in:
@@ -197,6 +197,7 @@ static inline __m256 exp256_ps(__m256 x) {
|
||||
const __m256 ln2 = _mm256_set1_ps(0.6931471805599453f);
|
||||
__m256 r = _mm256_fnmadd_ps(nf, ln2, x); // # 8 MULTs, 8 ADDs
|
||||
|
||||
const __m256 c13 = _mm256_set1_ps(c13_scalar);
|
||||
const __m256 c12 = _mm256_set1_ps(c12_scalar);
|
||||
const __m256 c11 = _mm256_set1_ps(c11_scalar);
|
||||
const __m256 c10 = _mm256_set1_ps(c10_scalar);
|
||||
@@ -211,7 +212,7 @@ static inline __m256 exp256_ps(__m256 x) {
|
||||
const __m256 c1 = _mm256_set1_ps(c1_scalar);
|
||||
const __m256 c0 = _mm256_set1_ps(c0_scalar);
|
||||
|
||||
__m256 p = c12;
|
||||
__m256 p = c13;
|
||||
|
||||
// # 13 * (8 MULTs, 8 ADDs)
|
||||
// # = 104 MULTs, 104 ADDs
|
||||
|
||||
Reference in New Issue
Block a user