Vectorized Class — pytorch Architecture

Source Code

aten/src/ATen/cpu/vec/vec128/vec128_float_neon.h lines 77–459

template <>
class Vectorized<float> {
 private:
  float32x4_t values;

 public:
  using value_type = float;
  using size_type = int;
  static constexpr size_type size() {
    return 4;
  }
  Vectorized() {
    values = vmovq_n_f32(0);
  }
  Vectorized(float32x4_t v) : values(v) {}
  Vectorized(float val) : values{vdupq_n_f32(val)} {}
  Vectorized(float val0, float val1, float val2, float val3)
      : values{val0, val1, val2, val3} {}
  Vectorized(float (&arr)[4]) : Vectorized(arr[0], arr[1], arr[2], arr[3]) {}
  operator float32x4_t() const {
    return values;
  }
  template <int64_t mask>
  static Vectorized<float> blend(
      const Vectorized<float>& a,
      const Vectorized<float>& b) {
    Vectorized<float> vec;
    vec.values = BlendRegs < 0,
    (mask & 0x01) != 0 > ::impl(a.values, b.values, vec.values);
    vec.values = BlendRegs < 1,
    (mask & 0x02) != 0 > ::impl(a.values, b.values, vec.values);
    vec.values = BlendRegs < 2,
    (mask & 0x04) != 0 > ::impl(a.values, b.values, vec.values);
    vec.values = BlendRegs < 3,
    (mask & 0x08) != 0 > ::impl(a.values, b.values, vec.values);
    return vec;
  }
  static Vectorized<float> blendv(
      const Vectorized<float>& a,
      const Vectorized<float>& b,
      const Vectorized<float>& mask) {
    // TODO
    // NB: This requires that each value, i.e., each uint value,
    // of the mask either all be zeros or all be 1s.
    // We perhaps need some kind of an assert?
    // But that will affect performance.
    Vectorized<float> vec(mask.values);
    vec.values =
        vbslq_f32(vreinterpretq_u32_f32(vec.values), b.values, a.values);
    return vec;
  }
  template <typename step_t>
  static Vectorized<float> arange(
      float base = 0.f,
      step_t step = static_cast<step_t>(1)) {
    const Vectorized<float> base_vec(base);
    const Vectorized<float> step_vec(step);
    const Vectorized<float> step_sizes(0, 1, 2, 3);
    return fmadd(step_sizes, step_vec, base_vec);
  }
  static Vectorized<float> set(
      const Vectorized<float>& a,
      const Vectorized<float>& b,
      int64_t count = size()) {
    switch (count) {
      case 0:
        return a;
      case 1: {
        Vectorized<float> vec;
        static uint32x4_t mask_low = {0xFFFFFFFF, 0x0, 0x0, 0x0};
        vec.values = vreinterpretq_f32_u32(mask_low);
        vec.values =
            vbslq_f32(vreinterpretq_u32_f32(vec.values), b.values, a.values);
        return vec;
      }
      case 2: {
        Vectorized<float> vec;
        static uint32x4_t mask_low = {0xFFFFFFFF, 0xFFFFFFFF, 0x0, 0x0};
        vec.values = vreinterpretq_f32_u32(mask_low);
        vec.values =
            vbslq_f32(vreinterpretq_u32_f32(vec.values), b.values, a.values);
        return vec;
      }
      case 3: {
        Vectorized<float> vec;
        static uint32x4_t mask_low = {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0x0};
        vec.values = vreinterpretq_f32_u32(mask_low);
        vec.values =
            vbslq_f32(vreinterpretq_u32_f32(vec.values), b.values, a.values);
        return vec;
      }
    }
    return b;
  }
  static Vectorized<float> loadu(const void* ptr, int64_t count = size()) {
    if (count == size()) {
      return vld1q_f32(reinterpret_cast<const float*>(ptr));
    } else {
      __at_align__ float tmp_values[size()];
      for (const auto i : c10::irange(size())) {
        tmp_values[i] = 0.0;
      }
      std::memcpy(
          tmp_values,
          reinterpret_cast<const float*>(ptr),
          count * sizeof(float));
      return vld1q_f32(reinterpret_cast<const float*>(tmp_values));
    }
  }
  void store(void* ptr, int64_t count = size()) const {
    if (count == size()) {
      vst1q_f32(reinterpret_cast<float*>(ptr), values);
    } else {
      float tmp_values[size()];
      vst1q_f32(reinterpret_cast<float*>(tmp_values), values);
      std::memcpy(ptr, tmp_values, count * sizeof(float));
    }
  }
  // Very slow implementation of indexing.
  // Only required because vec256_qint refers to this.
  // Once we specialize that implementation for ARM
  // this should be removed. TODO (kimishpatel)
  float operator[](int idx) const {
    __at_align__ float tmp[size()];
    store(tmp);
    return tmp[idx];
  }
  float operator[](int idx) {
    __at_align__ float tmp[size()];
    store(tmp);
    return tmp[idx];
  }
  int zero_mask() const {
    uint32x4_t is_zero_vec = vceqzq_f32(values);
    const int32x4_t shift = vcombine_s32(
        vcreate_s32(0x0 | (int64_t(0x1) << 32)),
        vcreate_s32(0x2 | (int64_t(0x3) << 32)));
    uint32x4_t bits_vec =
        vshlq_u32(vandq_u32(is_zero_vec, vdupq_n_u32(1)), shift);
    return vaddvq_u32(bits_vec);
  }
  Vectorized<float> isnan() const {
    return vreinterpretq_f32_u32(vmvnq_u32(vceqq_f32(values, values)));
  }
  bool has_inf_nan() const {
    __at_align__ float tmp[size()];
    store(tmp);
    for (const auto i : c10::irange(size())) {
      if (_isnan(tmp[i]) || _isinf(tmp[i])) {
        return true;
      }
    }
    return false;
  }
  Vectorized<float> map(float (*const f)(float)) const {
    __at_align__ float tmp[size()];
    store(tmp);
    for (const auto i : c10::irange(size())) {
      tmp[i] = f(tmp[i]);
    }
    return loadu(tmp);
  }
  Vectorized<float> map2(
      const Vectorized<float>& second,
      float (*const f)(float, float)) const {
    __at_align__ float tmp[size()];
    __at_align__ float tmp_second[size()];
    store(tmp);
    second.store(tmp_second);
    for (const auto i : c10::irange(size())) {
      tmp[i] = f(tmp[i], tmp_second[i]);
    }
    return loadu(tmp);
  }
  Vectorized<float> abs() const {
    return Vectorized<float>(vabsq_f32(values));
  }
  Vectorized<float> angle() const {
    auto zero = Vectorized<float>(0);
    auto pi = Vectorized<float>(c10::pi<float>);
    auto tmp = blendv(zero, pi, *this < zero);
    return blendv(tmp, *this, isnan());
  }
  Vectorized<float> real() const {
    return *this;
  }
  Vectorized<float> imag() const {
    return Vectorized<float>(0.f);
  }
  Vectorized<float> conj() const {
    return *this;
  }
#define DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC_WITH_SLEEF_NAME(      \
    name, sleef_name)                                                        \
  Vectorized<float> name() const {                                           \
    return USE_SLEEF(Vectorized<float>(sleef_name(values)), map(std::name)); \
  }

#define DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(name)      \
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC_WITH_SLEEF_NAME( \
      name, Sleef_##name##f4_u10)

  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(acos)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(acosh)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(asin)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(asinh)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(atan)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(atanh)

#define DEFINE_SLEEF_COMPATIBLE_BINARY_ELEMENTWISE_FUNC_WITH_SLEEF_NAME( \
    name, sleef_name)                                                    \
  Vectorized<float> name(const Vectorized<float>& arg) const {           \
    return USE_SLEEF(                                                    \
        Vectorized<float>(sleef_name(values, arg.values)),               \
        map2(arg, std::name));                                           \
  }

#define DEFINE_SLEEF_COMPATIBLE_BINARY_ELEMENTWISE_FUNC(name)      \
  DEFINE_SLEEF_COMPATIBLE_BINARY_ELEMENTWISE_FUNC_WITH_SLEEF_NAME( \
      name, Sleef_##name##f4_u10)

  DEFINE_SLEEF_COMPATIBLE_BINARY_ELEMENTWISE_FUNC(atan2)
  DEFINE_SLEEF_COMPATIBLE_BINARY_ELEMENTWISE_FUNC_WITH_SLEEF_NAME(
      copysign,
      Sleef_copysignf4)
  Vectorized<float> erf() const;
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC_WITH_SLEEF_NAME(
      erfc,
      Sleef_erfcf4_u15)
  Vectorized<float> erfinv() const {
    return map(calc_erfinv);
  }
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(exp)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(exp2)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(expm1)
  // Implementation copied from Arm Optimized Routine
  // https://github.com/ARM-software/optimized-routines/blob/master/math/aarch64/advsimd/expf.c
  inline Vectorized<float> vexpq_f32_u20() const {
    // bail out to sleef if it's a special case:
    // i.e. there's an input s.t. |input| > 87.3....
    const float32x4_t special_bound = vdupq_n_f32(0x1.5d5e2ap+6f);
    uint32x4_t cmp = vcagtq_f32(values, special_bound);
    if (vpaddd_u64(vreinterpretq_u64_u32(cmp)) != 0) {
      return exp();
    }

    const float32x4_t inv_ln2 = vdupq_n_f32(0x1.715476p+0f);
    const float ln2_hi = 0x1.62e4p-1f;
    const float ln2_lo = 0x1.7f7d1cp-20f;
    const float c0 = 0x1.0e4020p-7f;
    const float c2 = 0x1.555e66p-3f;
    const float32x4_t ln2_c02 = {ln2_hi, ln2_lo, c0, c2};

    const uint32x4_t exponent_bias = vdupq_n_u32(0x3f800000);
    const float32x4_t c1 = vdupq_n_f32(0x1.573e2ep-5f);
    const float32x4_t c3 = vdupq_n_f32(0x1.fffdb6p-2f);
    const float32x4_t c4 = vdupq_n_f32(0x1.ffffecp-1f);

    /* exp(x) = 2^n (1 + poly(r)), with 1 + poly(r) in [1/sqrt(2),sqrt(2)]
      x = ln2*n + r, with r in [-ln2/2, ln2/2].  */

    float32x4_t n = vrndaq_f32(vmulq_f32(values, inv_ln2));
    float32x4_t r = vfmsq_laneq_f32(values, n, ln2_c02, 0);
    r = vfmsq_laneq_f32(r, n, ln2_c02, 1);
    uint32x4_t e = vshlq_n_u32(vreinterpretq_u32_s32(vcvtq_s32_f32(n)), 23);
    float32x4_t scale = vreinterpretq_f32_u32(vaddq_u32(e, exponent_bias));

    float32x4_t r2 = vmulq_f32(r, r);
    float32x4_t p = vfmaq_laneq_f32(c1, r, ln2_c02, 2);
    float32x4_t q = vfmaq_laneq_f32(c3, r, ln2_c02, 3);
    q = vfmaq_f32(q, p, r2);
    p = vmulq_f32(c4, r);
    float32x4_t poly = vfmaq_f32(p, q, r2);

    return vfmaq_f32(scale, poly, scale);
  }
  Vectorized<float> exp_u20() const {
    return vexpq_f32_u20();
  }
  Vectorized<float> fexp_u20() const {
    return exp_u20();
  }
  DEFINE_SLEEF_COMPATIBLE_BINARY_ELEMENTWISE_FUNC_WITH_SLEEF_NAME(
      fmod,
      Sleef_fmodf4)
  DEFINE_SLEEF_COMPATIBLE_BINARY_ELEMENTWISE_FUNC_WITH_SLEEF_NAME(
      hypot,
      Sleef_hypotf4_u05)
  Vectorized<float> i0() const {
    return map(calc_i0);
  }
  Vectorized<float> i0e() const {
    return map(calc_i0e);
  }
  Vectorized<float> digamma() const {
    return map(calc_digamma);
  }
  Vectorized<float> igamma(const Vectorized<float>& x) const {
    return map2(x, calc_igamma);
  }
  Vectorized<float> igammac(const Vectorized<float>& x) const {
    return map2(x, calc_igammac);
  }
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(log)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(log10)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(log1p)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(log2)
  DEFINE_SLEEF_COMPATIBLE_BINARY_ELEMENTWISE_FUNC_WITH_SLEEF_NAME(
      nextafter,
      Sleef_nextafterf4)
  Vectorized<float> frac() const;
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(sin)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(sinh)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(cos)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(cosh)
  Vectorized<float> ceil() const {
    return map(at::native::ceil_impl);
  }
  Vectorized<float> floor() const {
    return map(at::native::floor_impl);
  }
  Vectorized<float> neg() const {
    return Vectorized<float>(vnegq_f32(values));
  }
  Vectorized<float> round() const {
    // We do not use std::round because we would like to round midway numbers to
    // the nearest even integer.
    return map(at::native::round_impl);
  }
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(tan)
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(tanh)
  Vectorized<float> trunc() const {
    return Vectorized<float>(vrndq_f32(values));
  }
  DEFINE_SLEEF_COMPATIBLE_UNARY_ELEMENTWISE_FUNC(lgamma)
  Vectorized<float> sqrt() const {
    return Vectorized<float>(vsqrtq_f32(values));
  }
  Vectorized<float> reciprocal() const {
    return Vectorized<float>(vdivq_f32(vdupq_n_f32(1.0f), values));
  }
  Vectorized<float> rsqrt() const {
    return this->sqrt().reciprocal();
  }
  DEFINE_SLEEF_COMPATIBLE_BINARY_ELEMENTWISE_FUNC(pow)
  Vectorized<float> operator==(const Vectorized<float>& other) const {
    return Vectorized<float>(
        vreinterpretq_f32_u32(vceqq_f32(values, other.values)));
  }

  Vectorized<float> operator!=(const Vectorized<float>& other) const {
    float32x4_t r0 =
        vreinterpretq_f32_u32(vmvnq_u32(vceqq_f32(values, other.values)));
    return Vectorized<float>(r0);
  }

  Vectorized<float> operator<(const Vectorized<float>& other) const {
    return Vectorized<float>(
        vreinterpretq_f32_u32(vcltq_f32(values, other.values)));
  }

  Vectorized<float> operator<=(const Vectorized<float>& other) const {
    return Vectorized<float>(
        vreinterpretq_f32_u32(vcleq_f32(values, other.values)));
  }

  Vectorized<float> operator>(const Vectorized<float>& other) const {
    return Vectorized<float>(
        vreinterpretq_f32_u32(vcgtq_f32(values, other.values)));
  }

  Vectorized<float> operator>=(const Vectorized<float>& other) const {
    return Vectorized<float>(
        vreinterpretq_f32_u32(vcgeq_f32(values, other.values)));
  }

  Vectorized<float> eq(const Vectorized<float>& other) const;
  Vectorized<float> ne(const Vectorized<float>& other) const;
  Vectorized<float> gt(const Vectorized<float>& other) const;
  Vectorized<float> ge(const Vectorized<float>& other) const;
  Vectorized<float> lt(const Vectorized<float>& other) const;
  Vectorized<float> le(const Vectorized<float>& other) const;
};