vector_ta/indicators/

dx.rs

use crate::utilities::data_loader::{source_type, Candles};
use crate::utilities::enums::Kernel;
use crate::utilities::helpers::{
    alloc_with_nan_prefix, detect_best_batch_kernel, detect_best_kernel, init_matrix_prefixes,
    make_uninit_matrix,
};
use aligned_vec::{AVec, CACHELINE_ALIGN};
#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
use core::arch::x86_64::*;
#[cfg(not(target_arch = "wasm32"))]
use rayon::prelude::*;
use std::convert::AsRef;
use std::error::Error;
use std::mem::MaybeUninit;
use thiserror::Error;

#[cfg(feature = "python")]
use crate::utilities::kernel_validation::validate_kernel;
#[cfg(feature = "python")]
use numpy::{IntoPyArray, PyArray1, PyArrayMethods, PyReadonlyArray1};
#[cfg(feature = "python")]
use pyo3::exceptions::PyValueError;
#[cfg(feature = "python")]
use pyo3::prelude::*;
#[cfg(feature = "python")]
use pyo3::types::{PyDict, PyList};

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
use serde::{Deserialize, Serialize};
#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
use wasm_bindgen::prelude::*;

#[derive(Debug, Clone)]
pub enum DxData<'a> {
    Candles {
        candles: &'a Candles,
    },
    HlcSlices {
        high: &'a [f64],
        low: &'a [f64],
        close: &'a [f64],
    },
}

#[derive(Debug, Clone)]
pub struct DxOutput {
    pub values: Vec<f64>,
}

#[derive(Debug, Clone)]
#[cfg_attr(
    all(target_arch = "wasm32", feature = "wasm"),
    derive(Serialize, Deserialize)
)]
pub struct DxParams {
    pub period: Option<usize>,
}

impl Default for DxParams {
    fn default() -> Self {
        Self { period: Some(14) }
    }
}

impl DxParams {
    pub fn generate_batch_params(period_range: (usize, usize, usize)) -> Vec<Self> {
        let (start, end, step) = period_range;
        let step = if step == 0 { 1 } else { step };
        let mut params = Vec::new();

        let mut period = start;
        while period <= end {
            params.push(Self {
                period: Some(period),
            });
            period += step;
        }

        params
    }
}

#[derive(Debug, Clone)]
pub struct DxInput<'a> {
    pub data: DxData<'a>,
    pub params: DxParams,
}

impl<'a> DxInput<'a> {
    #[inline]
    pub fn from_candles(candles: &'a Candles, params: DxParams) -> Self {
        Self {
            data: DxData::Candles { candles },
            params,
        }
    }
    #[inline]
    pub fn from_hlc_slices(
        high: &'a [f64],
        low: &'a [f64],
        close: &'a [f64],
        params: DxParams,
    ) -> Self {
        Self {
            data: DxData::HlcSlices { high, low, close },
            params,
        }
    }
    #[inline]
    pub fn with_default_candles(candles: &'a Candles) -> Self {
        Self {
            data: DxData::Candles { candles },
            params: DxParams::default(),
        }
    }
    #[inline]
    pub fn get_period(&self) -> usize {
        self.params.period.unwrap_or(14)
    }
}

#[derive(Copy, Clone, Debug)]
pub struct DxBuilder {
    period: Option<usize>,
    kernel: Kernel,
}

impl Default for DxBuilder {
    fn default() -> Self {
        Self {
            period: None,
            kernel: Kernel::Auto,
        }
    }
}

impl DxBuilder {
    #[inline(always)]
    pub fn new() -> Self {
        Self::default()
    }
    #[inline(always)]
    pub fn period(mut self, n: usize) -> Self {
        self.period = Some(n);
        self
    }
    #[inline(always)]
    pub fn kernel(mut self, k: Kernel) -> Self {
        self.kernel = k;
        self
    }
    #[inline(always)]
    pub fn apply(self, c: &Candles) -> Result<DxOutput, DxError> {
        let p = DxParams {
            period: self.period,
        };
        let i = DxInput::from_candles(c, p);
        dx_with_kernel(&i, self.kernel)
    }
    #[inline(always)]
    pub fn apply_hlc(self, high: &[f64], low: &[f64], close: &[f64]) -> Result<DxOutput, DxError> {
        let p = DxParams {
            period: self.period,
        };
        let i = DxInput::from_hlc_slices(high, low, close, p);
        dx_with_kernel(&i, self.kernel)
    }
    #[inline(always)]
    pub fn into_stream(self) -> Result<DxStream, DxError> {
        let p = DxParams {
            period: self.period,
        };
        DxStream::try_new(p)
    }
}

#[derive(Debug, Error)]
pub enum DxError {
    #[error("dx: Empty data provided for DX.")]
    EmptyInputData,
    #[error("dx: Could not select candle field: {0}")]
    SelectCandleFieldError(String),
    #[error("dx: Invalid period: period = {period}, data length = {data_len}")]
    InvalidPeriod { period: usize, data_len: usize },
    #[error("dx: Not enough valid data: needed = {needed}, valid = {valid}")]
    NotEnoughValidData { needed: usize, valid: usize },
    #[error("dx: All high, low, and close values are NaN.")]
    AllValuesNaN,
    #[error("dx: output length mismatch: expected = {expected}, got = {got}")]
    OutputLengthMismatch { expected: usize, got: usize },
    #[error("dx: invalid kernel for batch: {0:?}")]
    InvalidKernelForBatch(Kernel),
    #[error("dx: invalid range: start={start}, end={end}, step={step}")]
    InvalidRange {
        start: usize,
        end: usize,
        step: usize,
    },
    #[error("dx: invalid input: {0}")]
    InvalidInput(String),
}

#[inline(always)]
fn dx_prepare<'a>(
    input: &'a DxInput,
    kernel: Kernel,
) -> Result<(&'a [f64], &'a [f64], &'a [f64], usize, usize, Kernel), DxError> {
    let (high, low, close) = match &input.data {
        DxData::Candles { candles } => {
            let h = candles
                .select_candle_field("high")
                .map_err(|e| DxError::SelectCandleFieldError(e.to_string()))?;
            let l = candles
                .select_candle_field("low")
                .map_err(|e| DxError::SelectCandleFieldError(e.to_string()))?;
            let c = candles
                .select_candle_field("close")
                .map_err(|e| DxError::SelectCandleFieldError(e.to_string()))?;
            (h, l, c)
        }
        DxData::HlcSlices { high, low, close } => (*high, *low, *close),
    };
    let len = high.len().min(low.len()).min(close.len());
    if len == 0 {
        return Err(DxError::EmptyInputData);
    }
    let period = input.get_period();
    if period == 0 || period > len {
        return Err(DxError::InvalidPeriod {
            period,
            data_len: len,
        });
    }

    let first = (0..len)
        .find(|&i| !high[i].is_nan() && !low[i].is_nan() && !close[i].is_nan())
        .ok_or(DxError::AllValuesNaN)?;
    if len - first < period {
        return Err(DxError::NotEnoughValidData {
            needed: period,
            valid: len - first,
        });
    }
    let chosen = match kernel {
        Kernel::Auto => Kernel::Scalar,
        k => k,
    };
    Ok((high, low, close, len, first, chosen))
}

#[inline]
pub fn dx(input: &DxInput) -> Result<DxOutput, DxError> {
    dx_with_kernel(input, Kernel::Auto)
}

pub fn dx_with_kernel(input: &DxInput, kernel: Kernel) -> Result<DxOutput, DxError> {
    let (h, l, c, len, first, chosen) = dx_prepare(input, kernel)?;
    let warm = first + input.get_period() - 1;
    let mut out = alloc_with_nan_prefix(len, warm);
    unsafe {
        match chosen {
            Kernel::Scalar | Kernel::ScalarBatch => {
                dx_scalar(h, l, c, input.get_period(), first, &mut out)
            }
            #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
            Kernel::Avx2 | Kernel::Avx2Batch => {
                dx_avx2(h, l, c, input.get_period(), first, &mut out)
            }
            #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
            Kernel::Avx512 | Kernel::Avx512Batch => {
                dx_avx512(h, l, c, input.get_period(), first, &mut out)
            }
            _ => unreachable!(),
        }
    }
    Ok(DxOutput { values: out })
}

#[cfg(not(all(target_arch = "wasm32", feature = "wasm")))]
#[inline]
pub fn dx_into(input: &DxInput, out: &mut [f64]) -> Result<(), DxError> {
    dx_into_slice(out, input, Kernel::Auto)
}

#[inline]
pub fn dx_scalar(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    period: usize,
    first_valid_idx: usize,
    out: &mut [f64],
) {
    let len = high.len().min(low.len()).min(close.len());
    if len == 0 {
        return;
    }

    let p_f64 = period as f64;
    let hundred = 100.0f64;

    let mut prev_high = high[first_valid_idx];
    let mut prev_low = low[first_valid_idx];
    let mut prev_close = close[first_valid_idx];

    let mut plus_dm_sum = 0.0f64;
    let mut minus_dm_sum = 0.0f64;
    let mut tr_sum = 0.0f64;
    let mut initial_count: usize = 0;

    unsafe {
        let mut i = first_valid_idx + 1;
        while i < len {
            let h = *high.get_unchecked(i);
            let l = *low.get_unchecked(i);
            let cl = *close.get_unchecked(i);

            if h.is_nan() | l.is_nan() | cl.is_nan() {
                *out.get_unchecked_mut(i) = if i > 0 {
                    *out.get_unchecked(i - 1)
                } else {
                    f64::NAN
                };
                prev_high = h;
                prev_low = l;
                prev_close = cl;
                i += 1;
                continue;
            }

            let up_move = h - prev_high;
            let down_move = prev_low - l;
            let mut plus_dm = 0.0f64;
            let mut minus_dm = 0.0f64;
            if up_move > 0.0 && up_move > down_move {
                plus_dm = up_move;
            } else if down_move > 0.0 && down_move > up_move {
                minus_dm = down_move;
            }

            let tr1 = h - l;
            let tr2 = (h - prev_close).abs();
            let tr3 = (l - prev_close).abs();
            let tr = tr1.max(tr2).max(tr3);

            if initial_count < (period - 1) {
                plus_dm_sum += plus_dm;
                minus_dm_sum += minus_dm;
                tr_sum += tr;
                initial_count += 1;

                if initial_count == (period - 1) {
                    let plus_di = (plus_dm_sum / tr_sum) * hundred;
                    let minus_di = (minus_dm_sum / tr_sum) * hundred;
                    let sum_di = plus_di + minus_di;
                    *out.get_unchecked_mut(i) = if sum_di != 0.0 {
                        hundred * ((plus_di - minus_di).abs() / sum_di)
                    } else {
                        0.0
                    };
                }
            } else {
                plus_dm_sum = plus_dm_sum - (plus_dm_sum / p_f64) + plus_dm;
                minus_dm_sum = minus_dm_sum - (minus_dm_sum / p_f64) + minus_dm;
                tr_sum = tr_sum - (tr_sum / p_f64) + tr;

                let plus_di = if tr_sum != 0.0 {
                    (plus_dm_sum / tr_sum) * hundred
                } else {
                    0.0
                };
                let minus_di = if tr_sum != 0.0 {
                    (minus_dm_sum / tr_sum) * hundred
                } else {
                    0.0
                };
                let sum_di = plus_di + minus_di;
                *out.get_unchecked_mut(i) = if sum_di != 0.0 {
                    hundred * ((plus_di - minus_di).abs() / sum_di)
                } else {
                    *out.get_unchecked(i - 1)
                };
            }

            prev_high = h;
            prev_low = l;
            prev_close = cl;

            i += 1;
        }
    }
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline]
pub fn dx_avx512(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    period: usize,
    first_valid: usize,
    out: &mut [f64],
) {
    if period <= 32 {
        unsafe { dx_avx512_short(high, low, close, period, first_valid, out) }
    } else {
        unsafe { dx_avx512_long(high, low, close, period, first_valid, out) }
    }
}

#[inline]
pub fn dx_avx2(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    period: usize,
    first_valid: usize,
    out: &mut [f64],
) {
    dx_scalar(high, low, close, period, first_valid, out)
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline]
pub fn dx_avx512_short(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    period: usize,
    first_valid: usize,
    out: &mut [f64],
) {
    dx_scalar(high, low, close, period, first_valid, out)
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline]
pub fn dx_avx512_long(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    period: usize,
    first_valid: usize,
    out: &mut [f64],
) {
    dx_scalar(high, low, close, period, first_valid, out)
}

#[derive(Debug, Clone)]
pub struct DxStream {
    period: usize,

    p_f64: f64,
    hundred: f64,

    plus_dm_sum: f64,
    minus_dm_sum: f64,
    tr_sum: f64,

    prev_high: f64,
    prev_low: f64,
    prev_close: f64,

    initial_count: usize,
    filled: bool,

    last_dx: f64,
}

impl DxStream {
    pub fn try_new(params: DxParams) -> Result<Self, DxError> {
        let period = params.period.unwrap_or(14);
        if period == 0 {
            return Err(DxError::InvalidPeriod {
                period,
                data_len: 0,
            });
        }
        Ok(Self {
            period,
            p_f64: period as f64,
            hundred: 100.0,

            plus_dm_sum: 0.0,
            minus_dm_sum: 0.0,
            tr_sum: 0.0,

            prev_high: f64::NAN,
            prev_low: f64::NAN,
            prev_close: f64::NAN,

            initial_count: 0,
            filled: false,
            last_dx: f64::NAN,
        })
    }

    #[inline(always)]
    pub fn update(&mut self, high: f64, low: f64, close: f64) -> Option<f64> {
        if self.prev_high.is_nan() || self.prev_low.is_nan() || self.prev_close.is_nan() {
            self.prev_high = high;
            self.prev_low = low;
            self.prev_close = close;
            return None;
        }

        if high.is_nan() || low.is_nan() || close.is_nan() {
            let carried = if self.filled { self.last_dx } else { f64::NAN };
            self.prev_high = high;
            self.prev_low = low;
            self.prev_close = close;
            return Some(carried);
        }

        let up_move = high - self.prev_high;
        let down_move = self.prev_low - low;
        let plus_dm = if up_move > 0.0 && up_move > down_move {
            up_move
        } else {
            0.0
        };
        let minus_dm = if down_move > 0.0 && down_move > up_move {
            down_move
        } else {
            0.0
        };

        let tr1 = high - low;
        let tr2 = (high - self.prev_close).abs();
        let tr3 = (low - self.prev_close).abs();
        let tr = tr1.max(tr2).max(tr3);

        let mut out: Option<f64> = None;

        if self.initial_count < (self.period - 1) {
            self.plus_dm_sum += plus_dm;
            self.minus_dm_sum += minus_dm;
            self.tr_sum += tr;
            self.initial_count += 1;

            if self.initial_count == (self.period - 1) {
                let plus_di = (self.plus_dm_sum / self.tr_sum) * self.hundred;
                let minus_di = (self.minus_dm_sum / self.tr_sum) * self.hundred;
                let sum_di = plus_di + minus_di;

                let dx = if sum_di != 0.0 {
                    self.hundred * ((plus_di - minus_di).abs() / sum_di)
                } else {
                    0.0
                };
                self.filled = true;
                self.last_dx = dx;
                out = Some(dx);
            }
        } else {
            self.plus_dm_sum = self.plus_dm_sum - (self.plus_dm_sum / self.p_f64) + plus_dm;
            self.minus_dm_sum = self.minus_dm_sum - (self.minus_dm_sum / self.p_f64) + minus_dm;
            self.tr_sum = self.tr_sum - (self.tr_sum / self.p_f64) + tr;

            let plus_di = if self.tr_sum != 0.0 {
                (self.plus_dm_sum / self.tr_sum) * self.hundred
            } else {
                0.0
            };
            let minus_di = if self.tr_sum != 0.0 {
                (self.minus_dm_sum / self.tr_sum) * self.hundred
            } else {
                0.0
            };
            let sum_di = plus_di + minus_di;

            let dx = if sum_di != 0.0 {
                self.hundred * ((plus_di - minus_di).abs() / sum_di)
            } else if self.filled {
                self.last_dx
            } else {
                f64::NAN
            };
            self.last_dx = dx;
            out = Some(dx);
        }

        self.prev_high = high;
        self.prev_low = low;
        self.prev_close = close;

        out
    }
}

#[derive(Clone, Debug)]
pub struct DxBatchRange {
    pub period: (usize, usize, usize),
}

impl Default for DxBatchRange {
    fn default() -> Self {
        Self {
            period: (14, 263, 1),
        }
    }
}

impl DxBatchRange {
    pub fn from_tuple(period: (usize, usize, usize)) -> Self {
        Self { period }
    }
}

#[derive(Clone, Debug, Default)]
pub struct DxBatchBuilder {
    range: DxBatchRange,
    kernel: Kernel,
}

impl DxBatchBuilder {
    pub fn new() -> Self {
        Self::default()
    }
    pub fn kernel(mut self, k: Kernel) -> Self {
        self.kernel = k;
        self
    }

    #[inline]
    pub fn period_range(mut self, start: usize, end: usize, step: usize) -> Self {
        self.range.period = (start, end, step);
        self
    }
    #[inline]
    pub fn period_static(mut self, p: usize) -> Self {
        self.range.period = (p, p, 0);
        self
    }

    pub fn apply_hlc(
        self,
        high: &[f64],
        low: &[f64],
        close: &[f64],
    ) -> Result<DxBatchOutput, DxError> {
        dx_batch_with_kernel(high, low, close, &self.range, self.kernel)
    }

    pub fn apply_candles(self, c: &Candles) -> Result<DxBatchOutput, DxError> {
        let high = source_type(c, "high");
        let low = source_type(c, "low");
        let close = source_type(c, "close");
        self.apply_hlc(high, low, close)
    }
}

pub struct DxBatchOutput {
    pub values: Vec<f64>,
    pub combos: Vec<DxParams>,
    pub rows: usize,
    pub cols: usize,
}

impl DxBatchOutput {
    pub fn row_for_params(&self, p: &DxParams) -> Option<usize> {
        self.combos
            .iter()
            .position(|c| c.period.unwrap_or(14) == p.period.unwrap_or(14))
    }
    pub fn values_for(&self, p: &DxParams) -> Option<&[f64]> {
        self.row_for_params(p).map(|row| {
            let start = row * self.cols;
            &self.values[start..start + self.cols]
        })
    }
}

#[inline(always)]
fn expand_grid_checked(r: &DxBatchRange) -> Result<Vec<DxParams>, DxError> {
    let (start, end, step) = r.period;

    if step == 0 || start == end {
        return Ok(vec![DxParams {
            period: Some(start),
        }]);
    }

    let mut out: Vec<usize> = Vec::new();
    if start < end {
        let mut v = start;
        while v <= end {
            out.push(v);
            match v.checked_add(step) {
                Some(next) if next != v => v = next,
                _ => break,
            }
        }
    } else {
        let mut v = start;

        loop {
            out.push(v);
            if v <= end {
                break;
            }
            let dec = v.saturating_sub(step);
            if dec == v {
                break;
            }
            v = dec;
        }

        out.sort_unstable();
    }
    if out.is_empty() {
        return Err(DxError::InvalidRange { start, end, step });
    }
    Ok(out
        .into_iter()
        .map(|p| DxParams { period: Some(p) })
        .collect())
}

pub fn dx_batch_with_kernel(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    sweep: &DxBatchRange,
    k: Kernel,
) -> Result<DxBatchOutput, DxError> {
    let kernel = match k {
        Kernel::Auto => Kernel::ScalarBatch,
        other if other.is_batch() => other,
        other => return Err(DxError::InvalidKernelForBatch(other)),
    };

    let simd = match kernel {
        Kernel::Avx512Batch => Kernel::Avx512,
        Kernel::Avx2Batch => Kernel::Avx2,
        Kernel::ScalarBatch => Kernel::Scalar,
        _ => unreachable!(),
    };
    dx_batch_par_slice(high, low, close, sweep, simd)
}

#[inline(always)]
pub fn dx_batch_slice(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    sweep: &DxBatchRange,
    kern: Kernel,
) -> Result<DxBatchOutput, DxError> {
    dx_batch_inner(high, low, close, sweep, kern, false)
}

#[inline(always)]
pub fn dx_batch_par_slice(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    sweep: &DxBatchRange,
    kern: Kernel,
) -> Result<DxBatchOutput, DxError> {
    dx_batch_inner(high, low, close, sweep, kern, true)
}

#[inline(always)]
fn dx_batch_inner_into(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    sweep: &DxBatchRange,
    kern: Kernel,
    parallel: bool,
    out: &mut [f64],
) -> Result<Vec<DxParams>, DxError> {
    let combos_vec = expand_grid_checked(sweep)?;
    let combos = combos_vec;

    let len = high.len().min(low.len()).min(close.len());
    let first = (0..len)
        .find(|&i| !high[i].is_nan() && !low[i].is_nan() && !close[i].is_nan())
        .ok_or(DxError::AllValuesNaN)?;
    let max_p = combos.iter().map(|c| c.period.unwrap()).max().unwrap();
    if len - first < max_p {
        return Err(DxError::NotEnoughValidData {
            needed: max_p,
            valid: len - first,
        });
    }

    let rows = combos.len();
    let total = rows
        .checked_mul(len)
        .ok_or_else(|| DxError::InvalidInput("rows*cols overflow".into()))?;
    if out.len() != total {
        return Err(DxError::OutputLengthMismatch {
            expected: total,
            got: out.len(),
        });
    }

    let actual = match kern {
        Kernel::Auto => detect_best_batch_kernel(),
        k => k,
    };
    let simd = match actual {
        Kernel::ScalarBatch | Kernel::Scalar => Kernel::Scalar,
        #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
        Kernel::Avx2Batch | Kernel::Avx2 => Kernel::Avx2,
        #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
        Kernel::Avx512Batch | Kernel::Avx512 => Kernel::Avx512,
        _ => unreachable!(),
    };

    let (plus_dm, minus_dm, tr, carry) = dx_precompute_terms(high, low, close, first, len);

    let do_row = |row: usize, dst_row: &mut [f64]| unsafe {
        let p = combos[row].period.unwrap();
        dx_row_scalar_precomputed(&plus_dm, &minus_dm, &tr, &carry, first, p, dst_row);
    };

    if parallel {
        #[cfg(not(target_arch = "wasm32"))]
        out.par_chunks_mut(len)
            .enumerate()
            .for_each(|(r, s)| do_row(r, s));
        #[cfg(target_arch = "wasm32")]
        for (r, s) in out.chunks_mut(len).enumerate() {
            do_row(r, s);
        }
    } else {
        for (r, s) in out.chunks_mut(len).enumerate() {
            do_row(r, s);
        }
    }
    Ok(combos)
}

#[inline(always)]
fn dx_precompute_terms(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    first: usize,
    len: usize,
) -> (AVec<f64>, AVec<f64>, AVec<f64>, Vec<u8>) {
    let mut plus_dm: AVec<f64> = AVec::with_capacity(CACHELINE_ALIGN, len);
    let mut minus_dm: AVec<f64> = AVec::with_capacity(CACHELINE_ALIGN, len);
    let mut tr: AVec<f64> = AVec::with_capacity(CACHELINE_ALIGN, len);

    let mut carry: Vec<u8> = vec![0; len];

    for _ in 0..len {
        plus_dm.push(0.0);
    }
    for _ in 0..len {
        minus_dm.push(0.0);
    }
    for _ in 0..len {
        tr.push(0.0);
    }

    if len == 0 || first + 1 >= len {
        return (plus_dm, minus_dm, tr, carry);
    }

    for i in (first + 1)..len {
        let h = high[i];
        let l = low[i];
        let c = close[i];
        if h.is_nan() || l.is_nan() || c.is_nan() {
            carry[i] = 1;
            continue;
        }

        let up_move = h - high[i - 1];
        let down_move = low[i - 1] - l;
        let pdm = if up_move > 0.0 && up_move > down_move {
            up_move
        } else {
            0.0
        };
        let mdm = if down_move > 0.0 && down_move > up_move {
            down_move
        } else {
            0.0
        };

        let tr1 = h - l;
        let tr2 = (h - close[i - 1]).abs();
        let tr3 = (l - close[i - 1]).abs();
        let t = tr1.max(tr2).max(tr3);

        plus_dm[i] = pdm;
        minus_dm[i] = mdm;
        tr[i] = t;
    }

    (plus_dm, minus_dm, tr, carry)
}

#[inline(always)]
unsafe fn dx_row_scalar_precomputed(
    plus_dm: &[f64],
    minus_dm: &[f64],
    tr: &[f64],
    carry: &[u8],
    first: usize,
    period: usize,
    out: &mut [f64],
) {
    let len = out.len();
    if len == 0 || first + 1 >= len {
        return;
    }

    let p_f64 = period as f64;
    let hundred = 100.0f64;

    let mut plus_dm_sum = 0.0f64;
    let mut minus_dm_sum = 0.0f64;
    let mut tr_sum = 0.0f64;
    let mut initial_count: usize = 0;

    let mut i = first + 1;
    while i < len {
        if *carry.get_unchecked(i) != 0 {
            *out.get_unchecked_mut(i) = if i > 0 {
                *out.get_unchecked(i - 1)
            } else {
                f64::NAN
            };
            i += 1;
            continue;
        }

        let pdm = *plus_dm.get_unchecked(i);
        let mdm = *minus_dm.get_unchecked(i);
        let t = *tr.get_unchecked(i);

        if initial_count < (period - 1) {
            plus_dm_sum += pdm;
            minus_dm_sum += mdm;
            tr_sum += t;
            initial_count += 1;
            if initial_count == (period - 1) {
                let plus_di = (plus_dm_sum / tr_sum) * hundred;
                let minus_di = (minus_dm_sum / tr_sum) * hundred;
                let sum_di = plus_di + minus_di;
                *out.get_unchecked_mut(i) = if sum_di != 0.0 {
                    hundred * ((plus_di - minus_di).abs() / sum_di)
                } else {
                    0.0
                };
            }
        } else {
            plus_dm_sum = plus_dm_sum - (plus_dm_sum / p_f64) + pdm;
            minus_dm_sum = minus_dm_sum - (minus_dm_sum / p_f64) + mdm;
            tr_sum = tr_sum - (tr_sum / p_f64) + t;
            let plus_di = if tr_sum != 0.0 {
                (plus_dm_sum / tr_sum) * hundred
            } else {
                0.0
            };
            let minus_di = if tr_sum != 0.0 {
                (minus_dm_sum / tr_sum) * hundred
            } else {
                0.0
            };
            let sum_di = plus_di + minus_di;
            *out.get_unchecked_mut(i) = if sum_di != 0.0 {
                hundred * ((plus_di - minus_di).abs() / sum_di)
            } else {
                *out.get_unchecked(i - 1)
            };
        }

        i += 1;
    }
}

fn dx_batch_inner(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    sweep: &DxBatchRange,
    kern: Kernel,
    parallel: bool,
) -> Result<DxBatchOutput, DxError> {
    let combos = expand_grid_checked(sweep)?;
    let rows = combos.len();
    let cols = high.len().min(low.len()).min(close.len());
    if cols == 0 {
        return Err(DxError::EmptyInputData);
    }
    let _ = rows
        .checked_mul(cols)
        .ok_or_else(|| DxError::InvalidInput("rows*cols overflow".into()))?;

    let first = (0..cols)
        .find(|&i| !high[i].is_nan() && !low[i].is_nan() && !close[i].is_nan())
        .ok_or(DxError::AllValuesNaN)?;
    let warm: Vec<usize> = combos
        .iter()
        .map(|c| first + c.period.unwrap() - 1)
        .collect();

    let mut buf_mu = make_uninit_matrix(rows, cols);
    init_matrix_prefixes(&mut buf_mu, cols, &warm);

    let mut guard = core::mem::ManuallyDrop::new(buf_mu);
    let out_slice: &mut [f64] =
        unsafe { core::slice::from_raw_parts_mut(guard.as_mut_ptr() as *mut f64, guard.len()) };

    let _ = dx_batch_inner_into(high, low, close, sweep, kern, parallel, out_slice)?;

    let values = unsafe {
        Vec::from_raw_parts(
            guard.as_mut_ptr() as *mut f64,
            guard.len(),
            guard.capacity(),
        )
    };
    Ok(DxBatchOutput {
        values,
        combos,
        rows,
        cols,
    })
}

#[inline(always)]
unsafe fn dx_row_scalar(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    first: usize,
    period: usize,
    out: &mut [f64],
) {
    dx_scalar(high, low, close, period, first, out)
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline(always)]
unsafe fn dx_row_avx2(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    first: usize,
    period: usize,
    out: &mut [f64],
) {
    dx_scalar(high, low, close, period, first, out)
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline(always)]
pub unsafe fn dx_row_avx512(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    first: usize,
    period: usize,
    out: &mut [f64],
) {
    if period <= 32 {
        dx_row_avx512_short(high, low, close, first, period, out);
    } else {
        dx_row_avx512_long(high, low, close, first, period, out);
    }
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline(always)]
unsafe fn dx_row_avx512_short(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    first: usize,
    period: usize,
    out: &mut [f64],
) {
    dx_scalar(high, low, close, period, first, out)
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline(always)]
unsafe fn dx_row_avx512_long(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    first: usize,
    period: usize,
    out: &mut [f64],
) {
    dx_scalar(high, low, close, period, first, out)
}

#[inline(always)]
pub fn expand_grid_dx(r: &DxBatchRange) -> Vec<DxParams> {
    expand_grid_checked(r).unwrap_or_else(|_| vec![])
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::skip_if_unsupported;
    use crate::utilities::data_loader::read_candles_from_csv;

    fn check_dx_partial_params(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;

        let default_params = DxParams { period: None };
        let input = DxInput::from_candles(&candles, default_params);
        let output = dx_with_kernel(&input, kernel)?;
        assert_eq!(output.values.len(), candles.close.len());
        Ok(())
    }

    #[cfg(not(all(target_arch = "wasm32", feature = "wasm")))]
    #[test]
    fn test_dx_into_matches_api() -> Result<(), Box<dyn Error>> {
        let n = 512usize;
        let mut close = vec![0.0f64; n];
        for i in 0..n {
            let t = i as f64;
            close[i] = 100.0 + 0.1 * t + (t * 0.2).sin() * 2.0;
        }
        let mut high = vec![0.0f64; n];
        let mut low = vec![0.0f64; n];
        for i in 0..n {
            let t = i as f64;

            high[i] = close[i] + 0.6 + 0.05 * (t * 0.3).sin();
            low[i] = close[i] - 0.6 - 0.05 * (t * 0.3).cos();
            if low[i] > high[i] {
                core::mem::swap(&mut low[i], &mut high[i]);
            }
        }

        let params = DxParams { period: Some(14) };
        let input = DxInput::from_hlc_slices(&high, &low, &close, params);

        let base = dx(&input)?.values;

        let mut into_out = vec![0.0f64; n];
        dx_into(&input, &mut into_out)?;

        #[inline]
        fn eq_or_both_nan(a: f64, b: f64) -> bool {
            (a.is_nan() && b.is_nan()) || (a == b) || ((a - b).abs() <= 1e-12)
        }

        assert_eq!(base.len(), into_out.len());
        for i in 0..n {
            assert!(
                eq_or_both_nan(base[i], into_out[i]),
                "dx_into mismatch at {}: base={}, into={}",
                i,
                base[i],
                into_out[i]
            );
        }
        Ok(())
    }

    fn check_dx_accuracy(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;

        let input = DxInput::from_candles(&candles, DxParams::default());
        let result = dx_with_kernel(&input, kernel)?;
        let expected_last_five = [
            43.72121533411883,
            41.47251493226443,
            43.43041386436222,
            43.22673458811955,
            51.65514026197179,
        ];
        let start = result.values.len().saturating_sub(5);
        for (i, &val) in result.values[start..].iter().enumerate() {
            let diff = (val - expected_last_five[i]).abs();
            assert!(
                diff < 1e-4,
                "[{}] DX {:?} mismatch at idx {}: got {}, expected {}",
                test_name,
                kernel,
                i,
                val,
                expected_last_five[i]
            );
        }
        Ok(())
    }

    fn check_dx_default_candles(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;

        let input = DxInput::with_default_candles(&candles);
        match input.data {
            DxData::Candles { .. } => {}
            _ => panic!("Expected DxData::Candles"),
        }
        let output = dx_with_kernel(&input, kernel)?;
        assert_eq!(output.values.len(), candles.close.len());
        Ok(())
    }

    fn check_dx_zero_period(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let high = [2.0, 2.5, 3.0];
        let low = [1.0, 1.2, 2.1];
        let close = [1.5, 2.3, 2.2];
        let params = DxParams { period: Some(0) };
        let input = DxInput::from_hlc_slices(&high, &low, &close, params);
        let res = dx_with_kernel(&input, kernel);
        assert!(
            res.is_err(),
            "[{}] DX should fail with zero period",
            test_name
        );
        Ok(())
    }

    fn check_dx_period_exceeds_length(
        test_name: &str,
        kernel: Kernel,
    ) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let high = [3.0, 4.0];
        let low = [2.0, 3.0];
        let close = [2.5, 3.5];
        let params = DxParams { period: Some(14) };
        let input = DxInput::from_hlc_slices(&high, &low, &close, params);
        let res = dx_with_kernel(&input, kernel);
        assert!(
            res.is_err(),
            "[{}] DX should fail with period exceeding length",
            test_name
        );
        Ok(())
    }

    fn check_dx_very_small_dataset(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let high = [3.0];
        let low = [2.0];
        let close = [2.5];
        let params = DxParams { period: Some(14) };
        let input = DxInput::from_hlc_slices(&high, &low, &close, params);
        let res = dx_with_kernel(&input, kernel);
        assert!(
            res.is_err(),
            "[{}] DX should fail with insufficient data",
            test_name
        );
        Ok(())
    }

    fn check_dx_reinput(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;

        let first_params = DxParams { period: Some(14) };
        let first_input = DxInput::from_candles(&candles, first_params);
        let first_result = dx_with_kernel(&first_input, kernel)?;

        let second_params = DxParams { period: Some(14) };
        let second_input = DxInput::from_hlc_slices(
            &first_result.values,
            &first_result.values,
            &first_result.values,
            second_params,
        );
        let second_result = dx_with_kernel(&second_input, kernel)?;
        assert_eq!(second_result.values.len(), first_result.values.len());
        for i in 28..second_result.values.len() {
            assert!(
                !second_result.values[i].is_nan(),
                "[{}] Expected no NaN after index 28, found NaN at idx {}",
                test_name,
                i
            );
        }
        Ok(())
    }

    fn check_dx_nan_handling(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;
        let input = DxInput::from_candles(&candles, DxParams { period: Some(14) });
        let res = dx_with_kernel(&input, kernel)?;
        assert_eq!(res.values.len(), candles.close.len());
        if res.values.len() > 50 {
            for (i, &val) in res.values[50..].iter().enumerate() {
                assert!(
                    !val.is_nan(),
                    "[{}] Found unexpected NaN at out-index {}",
                    test_name,
                    50 + i
                );
            }
        }
        Ok(())
    }

    fn check_dx_streaming(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;
        let high = source_type(&candles, "high");
        let low = source_type(&candles, "low");
        let close = source_type(&candles, "close");
        let period = 14;

        let input = DxInput::from_candles(
            &candles,
            DxParams {
                period: Some(period),
            },
        );
        let batch_output = dx_with_kernel(&input, kernel)?.values;

        let mut stream = DxStream::try_new(DxParams {
            period: Some(period),
        })?;
        let mut stream_values = Vec::with_capacity(candles.close.len());
        for ((&h, &l), &c) in high.iter().zip(low).zip(close) {
            match stream.update(h, l, c) {
                Some(dx_val) => stream_values.push(dx_val),
                None => stream_values.push(f64::NAN),
            }
        }

        assert_eq!(batch_output.len(), stream_values.len());
        for (i, (&b, &s)) in batch_output.iter().zip(stream_values.iter()).enumerate() {
            if b.is_nan() && s.is_nan() {
                continue;
            }
            let diff = (b - s).abs();
            assert!(
                diff < 1e-9,
                "[{}] DX streaming f64 mismatch at idx {}: batch={}, stream={}, diff={}",
                test_name,
                i,
                b,
                s,
                diff
            );
        }
        Ok(())
    }

    macro_rules! generate_all_dx_tests {
        ($($test_fn:ident),*) => {
            paste::paste! {
                $(
                    #[test]
                    fn [<$test_fn _scalar_f64>]() {
                        let _ = $test_fn(stringify!([<$test_fn _scalar_f64>]), Kernel::Scalar);
                    }
                )*
                #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
                $(
                    #[test]
                    fn [<$test_fn _avx2_f64>]() {
                        let _ = $test_fn(stringify!([<$test_fn _avx2_f64>]), Kernel::Avx2);
                    }
                    #[test]
                    fn [<$test_fn _avx512_f64>]() {
                        let _ = $test_fn(stringify!([<$test_fn _avx512_f64>]), Kernel::Avx512);
                    }
                )*
            }
        }
    }
    #[cfg(debug_assertions)]
    fn check_dx_no_poison(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);

        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;

        let test_params = vec![
            DxParams::default(),
            DxParams { period: Some(2) },
            DxParams { period: Some(5) },
            DxParams { period: Some(7) },
            DxParams { period: Some(10) },
            DxParams { period: Some(14) },
            DxParams { period: Some(20) },
            DxParams { period: Some(30) },
            DxParams { period: Some(50) },
            DxParams { period: Some(100) },
            DxParams { period: Some(200) },
        ];

        for (param_idx, params) in test_params.iter().enumerate() {
            let input = DxInput::from_candles(&candles, params.clone());
            let output = dx_with_kernel(&input, kernel)?;

            for (i, &val) in output.values.iter().enumerate() {
                if val.is_nan() {
                    continue;
                }

                let bits = val.to_bits();

                if bits == 0x11111111_11111111 {
                    panic!(
                        "[{}] Found alloc_with_nan_prefix poison value {} (0x{:016X}) at index {} \
						 with params: period={} (param set {})",
                        test_name,
                        val,
                        bits,
                        i,
                        params.period.unwrap_or(14),
                        param_idx
                    );
                }

                if bits == 0x22222222_22222222 {
                    panic!(
                        "[{}] Found init_matrix_prefixes poison value {} (0x{:016X}) at index {} \
						 with params: period={} (param set {})",
                        test_name,
                        val,
                        bits,
                        i,
                        params.period.unwrap_or(14),
                        param_idx
                    );
                }

                if bits == 0x33333333_33333333 {
                    panic!(
                        "[{}] Found make_uninit_matrix poison value {} (0x{:016X}) at index {} \
						 with params: period={} (param set {})",
                        test_name,
                        val,
                        bits,
                        i,
                        params.period.unwrap_or(14),
                        param_idx
                    );
                }
            }
        }

        Ok(())
    }

    #[cfg(not(debug_assertions))]
    fn check_dx_no_poison(_test_name: &str, _kernel: Kernel) -> Result<(), Box<dyn Error>> {
        Ok(())
    }

    #[cfg(test)]
    #[allow(clippy::float_cmp)]
    fn check_dx_property(
        test_name: &str,
        kernel: Kernel,
    ) -> Result<(), Box<dyn std::error::Error>> {
        use proptest::prelude::*;
        skip_if_unsupported!(kernel, test_name);

        let strat = (2usize..=50)
            .prop_flat_map(|period| {
                (
                    100.0f64..5000.0f64,
                    (period + 20)..400,
                    0.001f64..0.05f64,
                    -0.01f64..0.01f64,
                    Just(period),
                )
            })
            .prop_map(|(base_price, data_len, volatility, trend, period)| {
                let mut high = Vec::with_capacity(data_len);
                let mut low = Vec::with_capacity(data_len);
                let mut close = Vec::with_capacity(data_len);

                let mut price = base_price;
                for i in 0..data_len {
                    let trend_component = trend * i as f64;
                    let random_component = ((i * 7 + 13) % 17) as f64 / 17.0 - 0.5;
                    price =
                        base_price + trend_component + random_component * volatility * base_price;

                    let daily_volatility = volatility * price;
                    let h = price + daily_volatility * (0.5 + ((i * 3) % 7) as f64 / 14.0);
                    let l = price - daily_volatility * (0.5 + ((i * 5) % 7) as f64 / 14.0);
                    let c = l + (h - l) * (0.3 + ((i * 11) % 7) as f64 / 10.0);

                    high.push(h);
                    low.push(l);
                    close.push(c);
                }

                (high, low, close, period)
            });

        proptest::test_runner::TestRunner::default()
			.run(&strat, |(high, low, close, period)| {
				let params = DxParams { period: Some(period) };
				let input = DxInput::from_hlc_slices(&high, &low, &close, params.clone());

				let DxOutput { values: out } = dx_with_kernel(&input, kernel).unwrap();
				let DxOutput { values: ref_out } = dx_with_kernel(&input, Kernel::Scalar).unwrap();


				for (i, &val) in out.iter().enumerate() {
					if !val.is_nan() {
						prop_assert!(
							val >= -1e-9 && val <= 100.0 + 1e-9,
							"[{}] DX value {} at index {} is outside [0, 100] range",
							test_name, val, i
						);
					}
				}




				let warmup = period - 1;
				for i in 0..warmup {
					prop_assert!(
						out[i].is_nan(),
						"[{}] Expected NaN during warmup at index {}, got {}",
						test_name, i, out[i]
					);
				}


				if out.len() > warmup + 10 {
					for i in (warmup + 10)..out.len() {
						prop_assert!(
							!out[i].is_nan(),
							"[{}] Unexpected NaN after warmup at index {}",
							test_name, i
						);
					}
				}


				for (i, (&val, &ref_val)) in out.iter().zip(ref_out.iter()).enumerate() {
					if val.is_nan() && ref_val.is_nan() {
						continue;
					}

					let diff = (val - ref_val).abs();
					prop_assert!(
						diff < 1e-9,
						"[{}] Kernel mismatch at index {}: {} vs {} (diff: {})",
						test_name, i, val, ref_val, diff
					);
				}


				let all_same_high = high.windows(2).all(|w| (w[0] - w[1]).abs() < 1e-10);
				let all_same_low = low.windows(2).all(|w| (w[0] - w[1]).abs() < 1e-10);
				let all_same_close = close.windows(2).all(|w| (w[0] - w[1]).abs() < 1e-10);

				if all_same_high && all_same_low && all_same_close {

					if out.len() > warmup + 10 {
						let stable_vals = &out[warmup + 10..];
						for (i, &val) in stable_vals.iter().enumerate() {
							if !val.is_nan() {
								prop_assert!(
									val < 1.0,
									"[{}] With constant prices, expected DX < 1.0, got {} at index {}",
									test_name, val, warmup + 10 + i
								);
							}
						}
					}
				}


				if period <= 20 && out.len() > 100 {

					let mid = out.len() / 2;
					let first_half_avg_price = close[..mid].iter().sum::<f64>() / mid as f64;
					let second_half_avg_price = close[mid..].iter().sum::<f64>() / (out.len() - mid) as f64;
					let price_change = ((second_half_avg_price - first_half_avg_price) / first_half_avg_price).abs();


					if price_change > 0.05 {

						let first_half_dx = &out[warmup..mid];
						let second_half_dx = &out[mid..];

						let first_avg = first_half_dx.iter()
							.filter(|v| !v.is_nan())
							.sum::<f64>() / first_half_dx.len() as f64;
						let second_avg = second_half_dx.iter()
							.filter(|v| !v.is_nan())
							.sum::<f64>() / second_half_dx.len() as f64;


						prop_assert!(
							second_avg > 20.0 || first_avg > 20.0,
							"[{}] Expected higher average DX in trending market. First half avg: {}, Second half avg: {}",
							test_name, first_avg, second_avg
						);
					}
				}



				if period <= 14 && out.len() > 50 {

					let trend_base = close[0];
					let perfect_trend = (0..50)
						.map(|i| {
							let price = trend_base + (i as f64 * trend_base * 0.01);
							let h = price * 1.005;
							let l = price * 0.995;
							let c = price;
							(h, l, c)
						})
						.collect::<Vec<_>>();

					let perfect_high: Vec<f64> = perfect_trend.iter().map(|&(h, _, _)| h).collect();
					let perfect_low: Vec<f64> = perfect_trend.iter().map(|&(_, l, _)| l).collect();
					let perfect_close: Vec<f64> = perfect_trend.iter().map(|&(_, _, c)| c).collect();

					let perfect_input = DxInput::from_hlc_slices(&perfect_high, &perfect_low, &perfect_close, params.clone());
					let DxOutput { values: perfect_out } = dx_with_kernel(&perfect_input, kernel).unwrap();


					if perfect_out.len() > warmup + 10 {
						let stable_dx = &perfect_out[warmup + 10..];
						let avg_dx = stable_dx.iter()
							.filter(|v| !v.is_nan())
							.sum::<f64>() / stable_dx.len() as f64;

						prop_assert!(
							avg_dx > 50.0,
							"[{}] Expected high DX (>50) in perfect trend, got avg {}",
							test_name, avg_dx
						);
					}
				}



				if period <= 14 && out.len() > 50 {

					let range_base = close[0];
					let ranging_data = (0..50)
						.map(|i| {

							let price = if i % 4 < 2 {
								range_base * 1.01
							} else {
								range_base * 0.99
							};
							let h = price * 1.002;
							let l = price * 0.998;
							let c = price;
							(h, l, c)
						})
						.collect::<Vec<_>>();

					let ranging_high: Vec<f64> = ranging_data.iter().map(|&(h, _, _)| h).collect();
					let ranging_low: Vec<f64> = ranging_data.iter().map(|&(_, l, _)| l).collect();
					let ranging_close: Vec<f64> = ranging_data.iter().map(|&(_, _, c)| c).collect();

					let ranging_input = DxInput::from_hlc_slices(&ranging_high, &ranging_low, &ranging_close, params.clone());
					let DxOutput { values: ranging_out } = dx_with_kernel(&ranging_input, kernel).unwrap();


					if ranging_out.len() > warmup + 10 {
						let stable_dx = &ranging_out[warmup + 10..];
						let avg_dx = stable_dx.iter()
							.filter(|v| !v.is_nan())
							.sum::<f64>() / stable_dx.len() as f64;



						prop_assert!(
							avg_dx < 65.0,
							"[{}] Expected moderate DX (<65) in ranging market, got avg {}",
							test_name, avg_dx
						);
					}
				}

				Ok(())
			})
			.unwrap();

        Ok(())
    }

    generate_all_dx_tests!(
        check_dx_partial_params,
        check_dx_accuracy,
        check_dx_default_candles,
        check_dx_zero_period,
        check_dx_period_exceeds_length,
        check_dx_very_small_dataset,
        check_dx_reinput,
        check_dx_nan_handling,
        check_dx_streaming,
        check_dx_no_poison
    );

    #[cfg(test)]
    generate_all_dx_tests!(check_dx_property);

    fn check_batch_default_row(test: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test);

        let file = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let c = read_candles_from_csv(file)?;

        let output = DxBatchBuilder::new().kernel(kernel).apply_candles(&c)?;

        let def = DxParams::default();
        let row = output.values_for(&def).expect("default row missing");

        assert_eq!(row.len(), c.close.len());

        let expected = [
            43.72121533411883,
            41.47251493226443,
            43.43041386436222,
            43.22673458811955,
            51.65514026197179,
        ];
        let start = row.len() - 5;
        for (i, &v) in row[start..].iter().enumerate() {
            assert!(
                (v - expected[i]).abs() < 1e-4,
                "[{test}] default-row mismatch at idx {i}: {v} vs {expected:?}"
            );
        }
        Ok(())
    }

    macro_rules! gen_batch_tests {
        ($fn_name:ident) => {
            paste::paste! {
                #[test] fn [<$fn_name _scalar>]()      {
                    let _ = $fn_name(stringify!([<$fn_name _scalar>]), Kernel::ScalarBatch);
                }
                #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
                #[test] fn [<$fn_name _avx2>]()        {
                    let _ = $fn_name(stringify!([<$fn_name _avx2>]), Kernel::Avx2Batch);
                }
                #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
                #[test] fn [<$fn_name _avx512>]()      {
                    let _ = $fn_name(stringify!([<$fn_name _avx512>]), Kernel::Avx512Batch);
                }
                #[test] fn [<$fn_name _auto_detect>]() {
                    let _ = $fn_name(stringify!([<$fn_name _auto_detect>]), Kernel::Auto);
                }
            }
        };
    }
    fn check_batch_sweep(test: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test);

        let file = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let c = read_candles_from_csv(file)?;

        let output = DxBatchBuilder::new()
            .kernel(kernel)
            .period_range(10, 30, 5)
            .apply_candles(&c)?;

        let expected_combos = 5;
        assert_eq!(output.combos.len(), expected_combos);
        assert_eq!(output.rows, expected_combos);
        assert_eq!(output.cols, c.close.len());

        Ok(())
    }

    #[cfg(debug_assertions)]
    fn check_batch_no_poison(test: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test);

        let file = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let c = read_candles_from_csv(file)?;

        let test_configs = vec![
            (2, 10, 2),
            (5, 25, 5),
            (30, 60, 15),
            (2, 5, 1),
            (10, 20, 2),
            (14, 14, 0),
            (5, 50, 15),
            (100, 200, 50),
        ];

        for (cfg_idx, &(p_start, p_end, p_step)) in test_configs.iter().enumerate() {
            let output = DxBatchBuilder::new()
                .kernel(kernel)
                .period_range(p_start, p_end, p_step)
                .apply_candles(&c)?;

            for (idx, &val) in output.values.iter().enumerate() {
                if val.is_nan() {
                    continue;
                }

                let bits = val.to_bits();
                let row = idx / output.cols;
                let col = idx % output.cols;
                let combo = &output.combos[row];

                if bits == 0x11111111_11111111 {
                    panic!(
                        "[{}] Config {}: Found alloc_with_nan_prefix poison value {} (0x{:016X}) \
						 at row {} col {} (flat index {}) with params: period={}",
                        test,
                        cfg_idx,
                        val,
                        bits,
                        row,
                        col,
                        idx,
                        combo.period.unwrap_or(14)
                    );
                }

                if bits == 0x22222222_22222222 {
                    panic!(
                        "[{}] Config {}: Found init_matrix_prefixes poison value {} (0x{:016X}) \
						 at row {} col {} (flat index {}) with params: period={}",
                        test,
                        cfg_idx,
                        val,
                        bits,
                        row,
                        col,
                        idx,
                        combo.period.unwrap_or(14)
                    );
                }

                if bits == 0x33333333_33333333 {
                    panic!(
                        "[{}] Config {}: Found make_uninit_matrix poison value {} (0x{:016X}) \
						 at row {} col {} (flat index {}) with params: period={}",
                        test,
                        cfg_idx,
                        val,
                        bits,
                        row,
                        col,
                        idx,
                        combo.period.unwrap_or(14)
                    );
                }
            }
        }

        Ok(())
    }

    #[cfg(not(debug_assertions))]
    fn check_batch_no_poison(_test: &str, _kernel: Kernel) -> Result<(), Box<dyn Error>> {
        Ok(())
    }

    gen_batch_tests!(check_batch_default_row);
    gen_batch_tests!(check_batch_sweep);
    gen_batch_tests!(check_batch_no_poison);
}

#[cfg(feature = "python")]
#[pyfunction(name = "dx")]
#[pyo3(signature = (high, low, close, period, kernel=None))]
pub fn dx_py<'py>(
    py: Python<'py>,
    high: PyReadonlyArray1<f64>,
    low: PyReadonlyArray1<f64>,
    close: PyReadonlyArray1<f64>,
    period: usize,
    kernel: Option<&str>,
) -> PyResult<Bound<'py, PyArray1<f64>>> {
    let h = high.as_slice()?;
    let l = low.as_slice()?;
    let c = close.as_slice()?;
    let kern = validate_kernel(kernel, false)?;
    let params = DxParams {
        period: Some(period),
    };
    let inp = DxInput::from_hlc_slices(h, l, c, params);
    let vec_out: Vec<f64> = py
        .allow_threads(|| dx_with_kernel(&inp, kern).map(|o| o.values))
        .map_err(|e| PyValueError::new_err(e.to_string()))?;
    Ok(vec_out.into_pyarray(py))
}

#[cfg(feature = "python")]
#[pyfunction(name = "dx_batch")]
#[pyo3(signature = (high, low, close, period_range, kernel=None))]
pub fn dx_batch_py<'py>(
    py: Python<'py>,
    high: PyReadonlyArray1<f64>,
    low: PyReadonlyArray1<f64>,
    close: PyReadonlyArray1<f64>,
    period_range: (usize, usize, usize),
    kernel: Option<&str>,
) -> PyResult<Bound<'py, PyDict>> {
    use numpy::{PyArray1, PyArrayMethods};
    let h = high.as_slice()?;
    let l = low.as_slice()?;
    let c = close.as_slice()?;
    let sweep = DxBatchRange::from_tuple(period_range);
    let combos = expand_grid_checked(&sweep).map_err(|e| PyValueError::new_err(e.to_string()))?;
    let rows = combos.len();

    let cols = h.len().min(l.len()).min(c.len());
    let kern = validate_kernel(kernel, true)?;
    let DxBatchOutput { values, .. } = py
        .allow_threads(|| dx_batch_with_kernel(h, l, c, &sweep, kern))
        .map_err(|e| PyValueError::new_err(e.to_string()))?;

    let out_arr = PyArray1::from_vec(py, values);

    let dict = PyDict::new(py);
    dict.set_item("values", out_arr.reshape((rows, cols))?)?;
    dict.set_item(
        "periods",
        combos
            .iter()
            .map(|p| p.period.unwrap() as u64)
            .collect::<Vec<_>>()
            .into_pyarray(py),
    )?;
    Ok(dict.into())
}

#[cfg(all(feature = "python", feature = "cuda"))]
use crate::cuda::dx_wrapper::CudaDx;
#[cfg(all(feature = "python", feature = "cuda"))]
use crate::cuda::moving_averages::alma_wrapper::DeviceArrayF32 as DeviceArrayF32Cuda;
#[cfg(all(feature = "python", feature = "cuda"))]
use cust::context::Context as CudaContext;
#[cfg(all(feature = "python", feature = "cuda"))]
use std::sync::Arc;

#[cfg(all(feature = "python", feature = "cuda"))]
#[pyfunction(name = "dx_cuda_batch_dev")]
#[pyo3(signature = (high_f32, low_f32, close_f32, period_range, device_id=0))]
pub fn dx_cuda_batch_dev_py<'py>(
    py: Python<'py>,
    high_f32: numpy::PyReadonlyArray1<'py, f32>,
    low_f32: numpy::PyReadonlyArray1<'py, f32>,
    close_f32: numpy::PyReadonlyArray1<'py, f32>,
    period_range: (usize, usize, usize),
    device_id: usize,
) -> PyResult<(DxDeviceArrayF32Py, Bound<'py, PyDict>)> {
    use crate::cuda::cuda_available;
    if !cuda_available() {
        return Err(PyValueError::new_err("CUDA not available"));
    }
    let h = high_f32.as_slice()?;
    let l = low_f32.as_slice()?;
    let c = close_f32.as_slice()?;
    let sweep = DxBatchRange::from_tuple(period_range);
    let (inner, combos, ctx, dev_id) = py.allow_threads(|| {
        let cuda = CudaDx::new(device_id).map_err(|e| PyValueError::new_err(e.to_string()))?;
        let ctx = cuda.context_arc();
        let dev_id = cuda.device_id();
        cuda.dx_batch_dev(h, l, c, &sweep)
            .map(|(arr, combos)| (arr, combos, ctx, dev_id))
            .map_err(|e| PyValueError::new_err(e.to_string()))
    })?;
    let dict = PyDict::new(py);
    dict.set_item(
        "periods",
        combos
            .iter()
            .map(|p| p.period.unwrap() as u64)
            .collect::<Vec<_>>()
            .into_pyarray(py),
    )?;
    Ok((
        DxDeviceArrayF32Py {
            inner,
            _ctx: ctx,
            device_id: dev_id,
        },
        dict,
    ))
}

#[cfg(all(feature = "python", feature = "cuda"))]
#[pyfunction(name = "dx_cuda_many_series_one_param_dev")]
#[pyo3(signature = (high_tm_f32, low_tm_f32, close_tm_f32, cols, rows, period, device_id=0))]
pub fn dx_cuda_many_series_one_param_dev_py(
    py: Python<'_>,
    high_tm_f32: numpy::PyReadonlyArray1<'_, f32>,
    low_tm_f32: numpy::PyReadonlyArray1<'_, f32>,
    close_tm_f32: numpy::PyReadonlyArray1<'_, f32>,
    cols: usize,
    rows: usize,
    period: usize,
    device_id: usize,
) -> PyResult<DxDeviceArrayF32Py> {
    use crate::cuda::cuda_available;
    if !cuda_available() {
        return Err(PyValueError::new_err("CUDA not available"));
    }
    let h = high_tm_f32.as_slice()?;
    let l = low_tm_f32.as_slice()?;
    let c = close_tm_f32.as_slice()?;
    let (inner, ctx, dev_id) = py.allow_threads(|| {
        let cuda = CudaDx::new(device_id).map_err(|e| PyValueError::new_err(e.to_string()))?;
        let ctx = cuda.context_arc();
        let dev_id = cuda.device_id();
        cuda.dx_many_series_one_param_time_major_dev(h, l, c, cols, rows, period)
            .map(|arr| (arr, ctx, dev_id))
            .map_err(|e| PyValueError::new_err(e.to_string()))
    })?;
    Ok(DxDeviceArrayF32Py {
        inner,
        _ctx: ctx,
        device_id: dev_id,
    })
}

#[cfg(all(feature = "python", feature = "cuda"))]
#[pyclass(module = "ta_indicators.cuda", unsendable)]
pub struct DxDeviceArrayF32Py {
    pub(crate) inner: DeviceArrayF32Cuda,
    pub(crate) _ctx: Arc<CudaContext>,
    pub(crate) device_id: u32,
}

#[cfg(all(feature = "python", feature = "cuda"))]
#[pymethods]
impl DxDeviceArrayF32Py {
    #[getter]
    fn __cuda_array_interface__<'py>(&self, py: Python<'py>) -> PyResult<Bound<'py, PyDict>> {
        let inner = &self.inner;
        let d = PyDict::new(py);

        d.set_item("shape", (inner.rows, inner.cols))?;

        d.set_item("typestr", "<f4")?;

        d.set_item(
            "strides",
            (
                inner.cols * std::mem::size_of::<f32>(),
                std::mem::size_of::<f32>(),
            ),
        )?;
        let size = inner.rows.saturating_mul(inner.cols);
        let ptr = if size == 0 {
            0usize
        } else {
            inner.device_ptr() as usize
        };
        d.set_item("data", (ptr, false))?;

        d.set_item("version", 3)?;
        Ok(d)
    }

    fn __dlpack_device__(&self) -> PyResult<(i32, i32)> {
        unsafe {
            use cust::sys::cuPointerGetAttribute;
            let attr = cust::sys::CUpointer_attribute::CU_POINTER_ATTRIBUTE_DEVICE_ORDINAL;
            let mut dev_ordinal: i32 = -1;
            let res = cuPointerGetAttribute(
                &mut dev_ordinal as *mut _ as *mut std::ffi::c_void,
                attr,
                self.inner.device_ptr(),
            );
            if res == cust::sys::CUresult::CUDA_SUCCESS && dev_ordinal >= 0 {
                return Ok((2, dev_ordinal));
            }
            Ok((2, self.device_id as i32))
        }
    }

    #[pyo3(signature = (stream=None, max_version=None, dl_device=None, copy=None))]
    fn __dlpack__<'py>(
        &mut self,
        py: Python<'py>,
        stream: Option<PyObject>,
        max_version: Option<PyObject>,
        dl_device: Option<PyObject>,
        copy: Option<PyObject>,
    ) -> PyResult<PyObject> {
        let (kdl, alloc_dev) = self.__dlpack_device__()?;
        if let Some(dev_obj) = dl_device.as_ref() {
            if let Ok((dev_ty, dev_id)) = dev_obj.extract::<(i32, i32)>(py) {
                if dev_ty != kdl || dev_id != alloc_dev {
                    let wants_copy = copy
                        .as_ref()
                        .and_then(|c| c.extract::<bool>(py).ok())
                        .unwrap_or(false);
                    if wants_copy {
                        return Err(PyValueError::new_err(
                            "device copy not implemented for __dlpack__",
                        ));
                    } else {
                        return Err(PyValueError::new_err("dl_device mismatch for __dlpack__"));
                    }
                }
            }
        }
        let _ = stream;

        let dummy = cust::memory::DeviceBuffer::<f32>::from_slice(&[])
            .map_err(|e| PyValueError::new_err(e.to_string()))?;
        let inner = std::mem::replace(
            &mut self.inner,
            DeviceArrayF32Cuda {
                buf: dummy,
                rows: 0,
                cols: 0,
            },
        );

        let rows = inner.rows;
        let cols = inner.cols;
        let buf = inner.buf;

        let max_version_bound = max_version.map(|obj| obj.into_bound(py));

        crate::utilities::dlpack_cuda::export_f32_cuda_dlpack_2d(
            py,
            buf,
            rows,
            cols,
            alloc_dev,
            max_version_bound,
        )
    }
}

#[cfg(feature = "python")]
#[pyclass(name = "DxStream")]
pub struct DxStreamPy {
    inner: DxStream,
}

#[cfg(feature = "python")]
#[pymethods]
impl DxStreamPy {
    #[new]
    pub fn new(period: usize) -> PyResult<Self> {
        let params = DxParams {
            period: Some(period),
        };
        let inner = DxStream::try_new(params)
            .map_err(|e| pyo3::exceptions::PyValueError::new_err(e.to_string()))?;
        Ok(Self { inner })
    }

    pub fn update(&mut self, high: f64, low: f64, close: f64) -> Option<f64> {
        self.inner.update(high, low, close)
    }
}

#[inline]
pub fn dx_into_slice(dst: &mut [f64], input: &DxInput, kern: Kernel) -> Result<(), DxError> {
    let (h, l, c, len, first, chosen) = dx_prepare(input, kern)?;
    if dst.len() != len {
        return Err(DxError::OutputLengthMismatch {
            expected: len,
            got: dst.len(),
        });
    }
    unsafe {
        match chosen {
            Kernel::Scalar | Kernel::ScalarBatch => {
                dx_scalar(h, l, c, input.get_period(), first, dst)
            }
            #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
            Kernel::Avx2 | Kernel::Avx2Batch => dx_avx2(h, l, c, input.get_period(), first, dst),
            #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
            Kernel::Avx512 | Kernel::Avx512Batch => {
                dx_avx512(h, l, c, input.get_period(), first, dst)
            }
            _ => unreachable!(),
        }
    }
    let warm = first + input.get_period() - 1;
    for v in &mut dst[..warm] {
        *v = f64::NAN;
    }
    Ok(())
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn dx_js(high: &[f64], low: &[f64], close: &[f64], period: usize) -> Result<Vec<f64>, JsValue> {
    let input = DxInput::from_hlc_slices(
        high,
        low,
        close,
        DxParams {
            period: Some(period),
        },
    );
    let mut out = vec![0.0; high.len().min(low.len()).min(close.len())];
    dx_into_slice(&mut out, &input, detect_best_kernel())
        .map_err(|e| JsValue::from_str(&e.to_string()))?;
    Ok(out)
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn dx_into(
    h_ptr: *const f64,
    l_ptr: *const f64,
    c_ptr: *const f64,
    out_ptr: *mut f64,
    len: usize,
    period: usize,
) -> Result<(), JsValue> {
    if [
        h_ptr as usize,
        l_ptr as usize,
        c_ptr as usize,
        out_ptr as usize,
    ]
    .iter()
    .any(|&p| p == 0)
    {
        return Err(JsValue::from_str("null pointer"));
    }
    unsafe {
        let h = core::slice::from_raw_parts(h_ptr, len);
        let l = core::slice::from_raw_parts(l_ptr, len);
        let c = core::slice::from_raw_parts(c_ptr, len);
        let inp = DxInput::from_hlc_slices(
            h,
            l,
            c,
            DxParams {
                period: Some(period),
            },
        );

        if out_ptr == h_ptr as *mut f64
            || out_ptr == l_ptr as *mut f64
            || out_ptr == c_ptr as *mut f64
        {
            let mut tmp = vec![0.0; len];
            dx_into_slice(&mut tmp, &inp, detect_best_kernel())
                .map_err(|e| JsValue::from_str(&e.to_string()))?;
            let dst = core::slice::from_raw_parts_mut(out_ptr, len);
            dst.copy_from_slice(&tmp);
        } else {
            let out = core::slice::from_raw_parts_mut(out_ptr, len);
            dx_into_slice(out, &inp, detect_best_kernel())
                .map_err(|e| JsValue::from_str(&e.to_string()))?;
        }
        Ok(())
    }
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn dx_alloc(len: usize) -> *mut f64 {
    let mut vec = Vec::<f64>::with_capacity(len);
    let ptr = vec.as_mut_ptr();
    std::mem::forget(vec);
    ptr
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn dx_free(ptr: *mut f64, len: usize) {
    if !ptr.is_null() {
        unsafe {
            let _ = Vec::from_raw_parts(ptr, len, len);
        }
    }
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[derive(Serialize, Deserialize)]
pub struct DxBatchConfig {
    pub period_range: (usize, usize, usize),
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[derive(Serialize, Deserialize)]
pub struct DxBatchJsOutput {
    pub values: Vec<f64>,
    pub combos: Vec<DxParams>,
    pub rows: usize,
    pub cols: usize,
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen(js_name = "dx_batch")]
pub fn dx_batch_unified_js(
    high: &[f64],
    low: &[f64],
    close: &[f64],
    config: JsValue,
) -> Result<JsValue, JsValue> {
    let cfg: DxBatchConfig = serde_wasm_bindgen::from_value(config)
        .map_err(|e| JsValue::from_str(&format!("Invalid config: {}", e)))?;
    let sweep = DxBatchRange::from_tuple(cfg.period_range);

    let rows = expand_grid_checked(&sweep)
        .map_err(|e| JsValue::from_str(&e.to_string()))?
        .len();
    let cols = high.len().min(low.len()).min(close.len());
    let mut buf_mu = make_uninit_matrix(rows, cols);

    let first = (0..cols)
        .find(|&i| !high[i].is_nan() && !low[i].is_nan() && !close[i].is_nan())
        .ok_or_else(|| JsValue::from_str("AllValuesNaN"))?;
    let warm: Vec<usize> = expand_grid_checked(&sweep)
        .map_err(|e| JsValue::from_str(&e.to_string()))?
        .iter()
        .map(|p| first + p.period.unwrap() - 1)
        .collect();
    init_matrix_prefixes(&mut buf_mu, cols, &warm);

    let mut guard = core::mem::ManuallyDrop::new(buf_mu);
    let out_slice: &mut [f64] =
        unsafe { core::slice::from_raw_parts_mut(guard.as_mut_ptr() as *mut f64, guard.len()) };
    let combos = dx_batch_inner_into(
        high,
        low,
        close,
        &sweep,
        detect_best_batch_kernel(),
        false,
        out_slice,
    )
    .map_err(|e| JsValue::from_str(&e.to_string()))?;
    let values = unsafe {
        Vec::from_raw_parts(
            guard.as_mut_ptr() as *mut f64,
            guard.len(),
            guard.capacity(),
        )
    };
    let js = DxBatchJsOutput {
        values,
        combos,
        rows,
        cols,
    };
    serde_wasm_bindgen::to_value(&js)
        .map_err(|e| JsValue::from_str(&format!("Serialization error: {}", e)))
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn dx_batch_into(
    high_ptr: *const f64,
    low_ptr: *const f64,
    close_ptr: *const f64,
    out_ptr: *mut f64,
    len: usize,
    period_start: usize,
    period_end: usize,
    period_step: usize,
) -> Result<(), JsValue> {
    if high_ptr.is_null() || low_ptr.is_null() || close_ptr.is_null() || out_ptr.is_null() {
        return Err(JsValue::from_str("Null pointer provided"));
    }

    unsafe {
        let high = std::slice::from_raw_parts(high_ptr, len);
        let low = std::slice::from_raw_parts(low_ptr, len);
        let close = std::slice::from_raw_parts(close_ptr, len);
        let batch_range = DxBatchRange::from_tuple((period_start, period_end, period_step));
        let combos = DxParams::generate_batch_params((period_start, period_end, period_step));
        let n_combos = combos.len();

        if high_ptr == out_ptr || low_ptr == out_ptr || close_ptr == out_ptr {
            let result = dx_batch_with_kernel(high, low, close, &batch_range, Kernel::Auto)
                .map_err(|e| JsValue::from_str(&e.to_string()))?;
            let out = std::slice::from_raw_parts_mut(out_ptr, len * n_combos);
            out.copy_from_slice(&result.values);
        } else {
            let params = combos;
            let out = std::slice::from_raw_parts_mut(out_ptr, len * n_combos);

            let first = high
                .iter()
                .zip(low)
                .zip(close)
                .position(|((&h, &l), &c)| !h.is_nan() && !l.is_nan() && !c.is_nan())
                .ok_or_else(|| JsValue::from_str("All values are NaN"))?;

            let mut buf_uninit = make_uninit_matrix(params.len(), len);
            let warmup_periods: Vec<usize> = params
                .iter()
                .map(|p| first + p.period.unwrap() - 1)
                .collect();
            init_matrix_prefixes(&mut buf_uninit, len, &warmup_periods);

            let buf_ptr = buf_uninit.as_mut_ptr() as *mut f64;
            std::mem::forget(buf_uninit);
            let slice_out = std::slice::from_raw_parts_mut(buf_ptr, params.len() * len);

            for (i, param) in params.iter().enumerate() {
                let row_offset = i * len;
                let row = &mut slice_out[row_offset..row_offset + len];

                let warmup = first + param.period.unwrap() - 1;
                dx_scalar(high, low, close, param.period.unwrap(), first, row);
            }

            out.copy_from_slice(slice_out);
        }
        Ok(())
    }
}