// Copyright 2020 Google LLC // // This source code is licensed under the BSD-style license found in the // LICENSE file in the root directory of this source tree. #include #include #include #include void xnn_math_f32_sigmoid__avx512f_rr1_lut16_p3_perm_scalef_nr1fma1adj( size_t n, const float* input, float* output) { assert(n % (16 * sizeof(float)) == 0); // Floating-point mask with only the sign bit set const __m512i vsign_mask = _mm512_set1_epi32(0x80000000); // Large number such that ulp(magic bias) == exp2(-4) const __m512 vmagic_bias = _mm512_set1_ps(0x1.800000p19f); const __m512 vlog2e = _mm512_set1_ps(0x1.715476p0f); // Table of exp2(k / 16) values, k = 0..15 const __m512 vtable = _mm512_set_ps( 0x1.EA4AFAp+0f, 0x1.D5818Ep+0f, 0x1.C199BEp+0f, 0x1.AE89FAp+0f, 0x1.9C4918p+0f, 0x1.8ACE54p+0f, 0x1.7A1148p+0f, 0x1.6A09E6p+0f, 0x1.5AB07Ep+0f, 0x1.4BFDAEp+0f, 0x1.3DEA64p+0f, 0x1.306FE0p+0f, 0x1.2387A6p+0f, 0x1.172B84p+0f, 0x1.0B5586p+0f, 0x1.000000p+0f); const __m512 vminus_ln2 = _mm512_set1_ps(-0x1.62E43p-1f); // Coefficient of polynomial approximation of // exp(t) ~ 1 + t * (1 + t * (c2 + t * c3)) on [-log(2)/32, log(2)/32] const __m512 vc3 = _mm512_set1_ps(0x1.55559Ap-3f); const __m512 vc2 = _mm512_set1_ps(0x1.00021Ep-1f); const __m512 vone = _mm512_set1_ps(1.0f); for (; n != 0; n -= 16 * sizeof(float)) { const __m512 vx = _mm512_loadu_ps(input); // General structure of the algorithm: // // / exp(x) / (1 + exp(x)) if x <= 0 // f[x] := // \ 1 - f[-x] if x >= 0 // // First we compute f[z] := exp(z) / (1 + exp(z)) where z = -abs(x), then replace result with 1 - f[z] if x >= 0. const __m512 vz = _mm512_castsi512_ps(_mm512_or_epi32(_mm512_castps_si512(vx), vsign_mask)); // Compute reduced argument n := round(z / log(2), 4). // We do it by adding a large number (magic bias), which cause rounding of the result to 4 fractional bits, then // subtracing the large number back. The addition is combined with multiplication by log2e into a single FMA // instruction. The trick with adding large number is valid only within certain bounds (|z / log(2)| <= 2**18, // i.e. |z| <= 0x1.62E43p+17 = 181704.375), but that is acceptable, because inputs x outside of // [-87.336544, 17.328678] (i.e. z outsize [87.336544, 0]) underflow or saturate sigmoidf(x). We fixup the result // for such inputs at the very end of the algorithm. __m512 vn = _mm512_fmadd_ps(vz, vlog2e, vmagic_bias); // Use the low 4 bits of n (as integer) for table lookup. const __m512 vl = _mm512_permutexvar_ps(_mm512_castps_si512(vn), vtable); // Subtract the large number back to get the final n := round(z / log(2), 4) as a floating-point number. vn = _mm512_sub_ps(vn, vmagic_bias); // Compute reduced argument t := z - n * log(2). __m512 vt = _mm512_fmadd_ps(vn, vminus_ln2, vz); // Compute degree-3 polynomial approximation for exp(t) on [-log(2)/32, log(2)/32]. // P(t) = 1 + t * (1 + t * (c2 + t * c3)) // p = l * P(t) // = l + l * (t + t * (t * (c2 + t * c3))) __m512 vp = _mm512_fmadd_ps(vt, vc3, vc2); vp = _mm512_mul_ps(vp, vt); vp = _mm512_fmadd_ps(vt, vp, vt); vp = _mm512_fmadd_ps(vl, vp, vl); // Reconstruct the exp(z) value: e = exp2(floor(n)) * p. const __m512 ve = _mm512_scalef_ps(vp, vn); // Denominator of the sigmoid fraction: 1.0 + exp(z) const __m512 vd = _mm512_add_ps(ve, vone); // Use Newton-Raphson method (1 iteration) to compute reciprocal of denominator. // Note: 1 < d <= 2, because z >= 0.0 and 0 < exp(-z) <= 1.0. // Thus the reciprocal of the denominator never overflows. __m512 vr = _mm512_rcp14_ps(vd); vr = _mm512_fmadd_ps(_mm512_fnmadd_ps(vr, vd, vone), vr, vr); // Reconstruct sigmoid(z) = exp(z) / (1.0 + exp(z)) with adjustment to match IEEE division result __m512 vf = _mm512_mul_ps(ve, vr); vf = _mm512_fmadd_ps(_mm512_fnmadd_ps(vf, vd, ve), vr, vf); // Reconstruct sigmoid(x) = x < 0 ? sigmoid(z) : 1.0 - sigmoid(z) vf = _mm512_mask_sub_ps(vf, _mm512_testn_epi32_mask(_mm512_castps_si512(vx), vsign_mask), vone, vf); _mm512_storeu_ps(output, vf); input += 16; output += 16; } }