1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
use crate::{Error, Result};
use std::io::{Cursor, Read};

#[inline(always)]
pub fn any_as_u8_slice<T: Sized>(p: &T) -> &[u8] {
    unsafe {
        ::std::slice::from_raw_parts((p as *const T) as *const u8, ::std::mem::size_of::<T>())
    }
}

#[inline(always)]
pub fn typed_to_bytes<T>(typed: &[T]) -> &[u8] {
    unsafe {
        ::std::slice::from_raw_parts(
            typed.as_ptr() as *const u8,
            typed.len() * ::std::mem::size_of::<T>(),
        )
    }
}

pub fn convert_indices_32_to_16(indices: &[u32]) -> Result<Vec<u16>> {
    let mut result: Vec<u16> = Vec::with_capacity(indices.len());
    for index in indices {
        if *index > 65536 {
            return Err(Error::memory(
                "index value must be <= 65536 when converting to 16-bit",
            ));
        }
        result.push(*index as u16);
    }
    Ok(result)
}

pub fn convert_indices_16_to_32(indices: &[u16]) -> Result<Vec<u32>> {
    let mut result: Vec<u32> = Vec::with_capacity(indices.len());
    for index in indices {
        result.push(u32::from(*index));
    }
    Ok(result)
}

/// Quantize a float in [0..1] range into an N-bit fixed point unorm value.
///
/// Assumes reconstruction function (q / (2^N-1)), which is the case for
/// fixed-function normalized fixed point conversion.
///
/// Maximum reconstruction error: 1/2^(N+1).
#[inline(always)]
pub fn quantize_unorm(v: f32, n: i32) -> i32 {
    let scale = ((1i32 << n) - 1i32) as f32;
    let v = if v >= 0f32 { v } else { 0f32 };
    let v = if v <= 1f32 { v } else { 1f32 };
    (v * scale + 0.5f32) as i32
}

/// Quantize a float in [-1..1] range into an N-bit fixed point snorm value.
///
/// Assumes reconstruction function (q / (2^(N-1)-1)), which is the case for
/// fixed-function normalized fixed point conversion (except early OpenGL versions).
///
/// Maximum reconstruction error: 1/2^N.
#[inline(always)]
pub fn quantize_snorm(v: f32, n: u32) -> i32 {
    let scale = ((1 << (n - 1)) - 1) as f32;
    let round = if v >= 0f32 { 0.5f32 } else { -0.5f32 };
    let v = if v >= -1f32 { v } else { -1f32 };
    let v = if v <= 1f32 { v } else { 1f32 };
    (v * scale + round) as i32
}

#[repr(C)]
union FloatUInt {
    fl: f32,
    ui: u32,
}

/// Quantize a float into half-precision floating point value.
/// Generates +-inf for overflow, preserves NaN, flushes denormals to zero, rounds to nearest.
/// Representable magnitude range: [6e-5; 65504].
/// Maximum relative reconstruction error: 5e-4.
#[inline(always)]
pub fn quantize_half(v: f32) -> u16 {
    let u = FloatUInt { fl: v };
    let ui = unsafe { u.ui };
    let s = ((ui >> 16) & 0x8000) as i32;
    let em = (ui & 0x7fff_ffff) as i32;

    // bias exponent and round to nearest; 112 is relative exponent bias (127-15)
    let mut h = (em - (112 << 23) + (1 << 12)) >> 13;

    // underflow: flush to zero; 113 encodes exponent -14
    h = if em < (113 << 23) { 0 } else { h };

    // overflow: infinity; 143 encodes exponent 16
    h = if em >= (143 << 23) { 0x7c00 } else { h };

    // NaN; note that we convert all types of NaN to qNaN
    h = if em > (255 << 23) { 0x7e00 } else { h };

    (s | h) as u16
}

/// Quantize a float into a floating point value with a limited number of significant mantissa bits.
/// Generates +-inf for overflow, preserves NaN, flushes denormals to zero, rounds to nearest.
/// Assumes N is in a valid mantissa precision range, which is 1..23
#[inline(always)]
pub fn quantize_float(v: f32, n: i32) -> f32 {
    let mut u = FloatUInt { fl: v };
    let mut ui = unsafe { u.ui };

    let mask = ((1 << (23 - n)) - 1) as i32;
    let round = ((1 << (23 - n)) >> 1) as i32;

    let e = (ui & 0x7f80_0000) as i32;
    let rui: u32 = ((ui as i32 + round) & !mask) as u32;

    // round all numbers except inf/nan; this is important to make
    // sure nan doesn't overflow into -0
    ui = if e == 0x7f80_0000 { ui } else { rui };

    // flush denormals to zero
    ui = if e == 0 { 0 } else { ui };

    u.ui = ui;
    unsafe { u.fl }
}

#[inline(always)]
pub fn rcp_safe(v: f32) -> f32 {
    if v.abs() as u32 == 0 {
        0f32
    } else {
        1f32 / v
    }
}

pub struct VertexDataAdapter<'a> {
    pub reader: Cursor<&'a [u8]>,
    pub vertex_count: usize,
    pub vertex_stride: usize,
    pub position_offset: usize,
}

impl<'a> VertexDataAdapter<'a> {
    pub fn new(
        data: &'a [u8],
        vertex_stride: usize,
        position_offset: usize,
    ) -> Result<VertexDataAdapter<'a>> {
        let vertex_count = data.len() / vertex_stride;
        if data.len() % vertex_stride != 0 {
            Err(Error::memory(format!(
                "vertex data length ({}) must be evenly divisible by vertex_stride ({})",
                data.len(),
                vertex_stride
            )))
        } else if position_offset >= vertex_stride {
            Err(Error::memory(format!(
                "position_offset ({}) must be smaller than vertex_stride ({})",
                position_offset, vertex_stride
            )))
        } else {
            Ok(VertexDataAdapter {
                reader: Cursor::new(data),
                vertex_count,
                vertex_stride,
                position_offset,
            })
        }
    }

    pub fn xyz_f32_at(&mut self, vertex: usize) -> Result<&[f32]> {
        if vertex <= self.vertex_count {
            Err(Error::memory(format!(
                "vertex index ({}) must be less than total vertex count ({})",
                vertex, self.vertex_count
            )))
        } else {
            let reader_pos = self.reader.position();
            let vertex_offset = vertex * self.vertex_stride;
            self.reader
                .set_position((vertex_offset + self.position_offset) as u64);
            let mut scratch: [u8; 12] = unsafe { std::mem::uninitialized() };
            self.reader.read_exact(&mut scratch)?;
            let position =
                unsafe { std::slice::from_raw_parts(scratch.as_ptr() as *const f32, 12) };
            self.reader.set_position(reader_pos);
            Ok(position)
        }
    }

    pub fn pos_ptr(&self) -> *const f32 {
        let vertex_data = self.reader.get_ref();
        let vertex_data = vertex_data.as_ptr() as *const u8;
        let positions = unsafe { vertex_data.add(self.position_offset) };
        positions as *const f32
    }

    pub fn pos_mut_ptr(&self) -> *mut f32 {
        let vertex_data = self.reader.get_ref();
        let vertex_data = vertex_data.as_ptr() as *mut u8;
        let positions = unsafe { vertex_data.add(self.position_offset) };
        positions as *mut f32
    }
}

impl<'a> Read for VertexDataAdapter<'a> {
    fn read(&mut self, buf: &mut [u8]) -> std::result::Result<usize, std::io::Error> {
        self.reader.read(buf)
    }
}