furiosa-opt-std 0.3.0

Standard library for Furiosa NPU TCP Virtual ISA programming.
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

//! Hardware numeric type operations for Renegade NPU.

use std::ops::Mul;

pub(crate) const F8E4_ZERO: u8 = 0x00;
pub(crate) const F8E4_NEG_ZERO: u8 = 0x80;
pub(crate) const F8E4_ONE: u8 = 0x38;

pub(crate) fn f8_e4_to_f32(bits: u8) -> f32 {
    f32::from_bits(F8E4_TO_F32[bits as usize])
}

pub(crate) fn f8_e4_from_f32(v: f32) -> u8 {
    if v.is_nan() {
        return 0x7f; // qNaN
    }
    let n = v.to_bits();
    let exp = (n >> 23) & 0xff;
    let man = n & 0x7fffff;
    let abs_value = if exp > 127 - 7 {
        let valid = (n >> 20) & 0x7ff;
        let val = if (n & 0xfffff) + (valid & 1) > 0x80000 {
            valid + 1
        } else {
            valid
        };
        if val >= (((127 + 8) << 3) | 0x6) {
            0x7e_u32 // MAX
        } else {
            val - ((127 - 7) << 3)
        }
    } else if exp == 127 - 7 {
        if man >= 0x700000 {
            8
        } else if man > 0x500000 {
            7
        } else if man >= 0x300000 {
            6
        } else if man > 0x100000 {
            5
        } else {
            4
        }
    } else if exp == 127 - 8 {
        if man >= 0x600000 {
            4
        } else if man > 0x200000 {
            3
        } else {
            2
        }
    } else if exp == 127 - 9 {
        if man >= 0x400000 { 2 } else { 1 }
    } else if exp == 127 - 10 && man != 0 {
        1
    } else {
        0
    };
    (((n >> 31) << 7) | abs_value) as u8
}

pub(crate) fn f8_e4_is_zero(bits: u8) -> bool {
    bits == F8E4_ZERO || bits == F8E4_NEG_ZERO
}

#[rustfmt::skip]
const F8E4_TO_F32: &[u32; 256] = &[
    0x00000000, 0x3b000000, 0x3b800000, 0x3bc00000, 0x3c000000, 0x3c200000, 0x3c400000, 0x3c600000,
    0x3c800000, 0x3c900000, 0x3ca00000, 0x3cb00000, 0x3cc00000, 0x3cd00000, 0x3ce00000, 0x3cf00000,
    0x3d000000, 0x3d100000, 0x3d200000, 0x3d300000, 0x3d400000, 0x3d500000, 0x3d600000, 0x3d700000,
    0x3d800000, 0x3d900000, 0x3da00000, 0x3db00000, 0x3dc00000, 0x3dd00000, 0x3de00000, 0x3df00000,
    0x3e000000, 0x3e100000, 0x3e200000, 0x3e300000, 0x3e400000, 0x3e500000, 0x3e600000, 0x3e700000,
    0x3e800000, 0x3e900000, 0x3ea00000, 0x3eb00000, 0x3ec00000, 0x3ed00000, 0x3ee00000, 0x3ef00000,
    0x3f000000, 0x3f100000, 0x3f200000, 0x3f300000, 0x3f400000, 0x3f500000, 0x3f600000, 0x3f700000,
    0x3f800000, 0x3f900000, 0x3fa00000, 0x3fb00000, 0x3fc00000, 0x3fd00000, 0x3fe00000, 0x3ff00000,
    0x40000000, 0x40100000, 0x40200000, 0x40300000, 0x40400000, 0x40500000, 0x40600000, 0x40700000,
    0x40800000, 0x40900000, 0x40a00000, 0x40b00000, 0x40c00000, 0x40d00000, 0x40e00000, 0x40f00000,
    0x41000000, 0x41100000, 0x41200000, 0x41300000, 0x41400000, 0x41500000, 0x41600000, 0x41700000,
    0x41800000, 0x41900000, 0x41a00000, 0x41b00000, 0x41c00000, 0x41d00000, 0x41e00000, 0x41f00000,
    0x42000000, 0x42100000, 0x42200000, 0x42300000, 0x42400000, 0x42500000, 0x42600000, 0x42700000,
    0x42800000, 0x42900000, 0x42a00000, 0x42b00000, 0x42c00000, 0x42d00000, 0x42e00000, 0x42f00000,
    0x43000000, 0x43100000, 0x43200000, 0x43300000, 0x43400000, 0x43500000, 0x43600000, 0x43700000,
    0x43800000, 0x43900000, 0x43a00000, 0x43b00000, 0x43c00000, 0x43d00000, 0x43e00000, 0x7fc00000,
    0x80000000, 0xbb000000, 0xbb800000, 0xbbc00000, 0xbc000000, 0xbc200000, 0xbc400000, 0xbc600000,
    0xbc800000, 0xbc900000, 0xbca00000, 0xbcb00000, 0xbcc00000, 0xbcd00000, 0xbce00000, 0xbcf00000,
    0xbd000000, 0xbd100000, 0xbd200000, 0xbd300000, 0xbd400000, 0xbd500000, 0xbd600000, 0xbd700000,
    0xbd800000, 0xbd900000, 0xbda00000, 0xbdb00000, 0xbdc00000, 0xbdd00000, 0xbde00000, 0xbdf00000,
    0xbe000000, 0xbe100000, 0xbe200000, 0xbe300000, 0xbe400000, 0xbe500000, 0xbe600000, 0xbe700000,
    0xbe800000, 0xbe900000, 0xbea00000, 0xbeb00000, 0xbec00000, 0xbed00000, 0xbee00000, 0xbef00000,
    0xbf000000, 0xbf100000, 0xbf200000, 0xbf300000, 0xbf400000, 0xbf500000, 0xbf600000, 0xbf700000,
    0xbf800000, 0xbf900000, 0xbfa00000, 0xbfb00000, 0xbfc00000, 0xbfd00000, 0xbfe00000, 0xbff00000,
    0xc0000000, 0xc0100000, 0xc0200000, 0xc0300000, 0xc0400000, 0xc0500000, 0xc0600000, 0xc0700000,
    0xc0800000, 0xc0900000, 0xc0a00000, 0xc0b00000, 0xc0c00000, 0xc0d00000, 0xc0e00000, 0xc0f00000,
    0xc1000000, 0xc1100000, 0xc1200000, 0xc1300000, 0xc1400000, 0xc1500000, 0xc1600000, 0xc1700000,
    0xc1800000, 0xc1900000, 0xc1a00000, 0xc1b00000, 0xc1c00000, 0xc1d00000, 0xc1e00000, 0xc1f00000,
    0xc2000000, 0xc2100000, 0xc2200000, 0xc2300000, 0xc2400000, 0xc2500000, 0xc2600000, 0xc2700000,
    0xc2800000, 0xc2900000, 0xc2a00000, 0xc2b00000, 0xc2c00000, 0xc2d00000, 0xc2e00000, 0xc2f00000,
    0xc3000000, 0xc3100000, 0xc3200000, 0xc3300000, 0xc3400000, 0xc3500000, 0xc3600000, 0xc3700000,
    0xc3800000, 0xc3900000, 0xc3a00000, 0xc3b00000, 0xc3c00000, 0xc3d00000, 0xc3e00000, 0x7fc00000,
];

// ============================================================================
// Fixed-point ↔ float conversion
// ============================================================================

const MANTISSA_MASK: u32 = (1 << 23) - 1;
const EXPONENT_MASK: u32 = (1 << 8) - 1;
const EXPONENT_SHIFT: i32 = 23;
const SIGN_SHIFT: i32 = 31;
const EXPONENT_OFFSET: u32 = (1 << 7) - 1;

pub(crate) fn float_to_fixedpoint(input: f32, integer_width: u32) -> i32 {
    assert!(integer_width < 32);
    if input.is_nan() {
        return if input.is_sign_positive() { i32::MAX } else { i32::MIN };
    }

    let fraction_width = 31 - integer_width;
    let f64_result: f64 = f64::from(input)
        .mul(2_f64.powi(fraction_width as i32))
        .round()
        .clamp(i32::MIN as f64, i32::MAX as f64);
    f64_result as i32
}

pub(crate) fn fixedpoint_to_float(input: i32, integer_width: u32) -> f32 {
    assert!(integer_width < 32);
    let abs_value: u32 = input.unsigned_abs();

    let mask = (1 << 31) - 1;
    let leading_zeros = (abs_value & mask).leading_zeros() - 1;

    let aligned_abs_value = abs_value << leading_zeros;
    let rounded = round_to_nearest_even(aligned_abs_value, 31, 24);
    let mantissa = rounded & MANTISSA_MASK;

    let sign = if input < 0 { 1 } else { 0 };

    let exponent = if input == i32::MIN {
        integer_width + EXPONENT_OFFSET
    } else if leading_zeros == 31 {
        0
    } else {
        let mut result = integer_width + (EXPONENT_OFFSET - 1) - leading_zeros;
        if (rounded >> 24) != 0 {
            result += 1;
        }
        result
    };
    let float_as_bits = (sign << SIGN_SHIFT) | ((exponent & EXPONENT_MASK) << EXPONENT_SHIFT) | mantissa;
    f32::from_bits(float_as_bits)
}

fn round_to_nearest_even(input: u32, iw: u32, ow: u32) -> u32 {
    let in_width_mask = (1u32 << iw) - 1;
    let input_valid = input & in_width_mask;
    let shift = (iw - ow) - 2;

    let shift_mask = (1u32 << shift) - 1;
    let discarded = input_valid & shift_mask;
    let sticky = discarded != 0;

    let interest = input_valid >> shift;
    let round_factor = ((interest & 0x7) << 1) | u32::from(sticky);

    let round_up = matches!(
        round_factor,
        0b0110 | 0b1110 | 0b0101 | 0b1101 | 0b0111 | 0b1111 | 0b1100
    );

    (interest >> 2) + u32::from(round_up)
}