thermite 0.2.1

High-performance, generic, ISA-portable SIMD library with a policy-configurable transcendental math library
Documentation 
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
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248

use generic_array::{GenericArray, typenum};

use super::{
    spirv_count_ones_u32, spirv_count_ones_u64, spirv_leading_zeros_u32, spirv_leading_zeros_u64, spirv_swap_bytes_u32,
    spirv_swap_bytes_u64, spirv_trailing_zeros_u32, spirv_trailing_zeros_u64,
};
use crate::backend::spirv::arch::{self as arch, glsl};
use crate::{
    isa::InstructionSet,
    register::{
        BitshiftRegister, BitwiseRegister, CoreRegister, Element, IndexableRegister, IntegerRegister,
        InterleaveRegister, MaskElement, NumericRegister, PartialOrdRegister, PermuteRegister, Register,
        ShuffleRegister, Storage, UnsignedIntegerRegister, ZeroUpper,
    },
};

#[rustfmt::skip]
macro_rules! decl_spirv_unsigned_scalar { ($u:ty => $width:literal) => { paste::paste! {

#[thermite_macros::inline_always]
impl CoreRegister for [<u $width>] {
    type NativeIsa = crate::backend::scalar::Scalar;
    type Lanes   = typenum::U1;
    type Storage = $u;
    type Mask    = bool;

    const IS_EMULATED:         bool = false;
    const EMPTY:               Storage<Self> = 0;
    const HAS_EQUAL_SIZE_MASK: bool = false;

    fn blendv(mask: bool, lhs: Self, rhs: Self) -> Self {
        unsafe { arch::op_opselect::<Self, bool>(mask, rhs, lhs) }
    }
    fn z (mask: bool, value: Self) -> Self { unsafe { arch::op_opselect::<Self, bool>(mask, value, 0) } }
    fn nz(mask: bool, value: Self) -> Self { unsafe { arch::op_opselect::<Self, bool>(mask, 0, value) } }
    fn zeroupper_z<Z: ZeroUpper>(value: Self) -> Self {
        if const { Z::N >= 1 } { value } else { Self::EMPTY }
    }

    fn from_mask(mask: bool) -> Self { Self::from_bool(mask) }
}

#[rustfmt::skip]
#[thermite_macros::inline_always]
impl BitwiseRegister for [<u $width>] {
    fn bitxor(lhs: Self, rhs: Self) -> Self { lhs ^ rhs }
    fn bitand(lhs: Self, rhs: Self) -> Self { lhs & rhs }
    fn bitor (lhs: Self, rhs: Self) -> Self { lhs | rhs }
    fn not(value: Self) -> Self { !value }
}

#[rustfmt::skip]
#[thermite_macros::inline_always]
impl InterleaveRegister for [<u $width>] {
    fn interleave (a: Self, b: Self) -> (Self, Self) { (a, b) }
    fn deinterleave(a: Self, b: Self) -> (Self, Self) { (a, b) }
}

#[thermite_macros::inline_always]
impl Register for [<u $width>] {
    type Element  = $u;
    type Signed   = [<i $width>];
    type Unsigned = [<u $width>];

    fn into_mask(value: Self) -> bool { value.to_bool() }
    fn msb_to_mask(value: Self) -> bool {
        // Reinterpret as signed then arithmetic-shift to propagate the MSB
        let signed = unsafe { arch::op_opbitcast::<[<i $width>], Self>(value) };
        <[<i $width>]>::into_mask(signed >> (<$u>::BITS - 1))
    }

    fn new(value: GenericArray<Self::Element, Self::Lanes>) -> Self { value[0] }
    fn single(value: Self::Element) -> Self { value }
    fn splat (value: Self::Element) -> Self { value }
    fn broadcast<const I: usize>(value: Self) -> Self { value }
    fn reverse(value: Self) -> Self { value }

    fn swap_bytes(value: Self) -> Self {
        [<spirv_swap_bytes_u $width>](value)
    }

    const HAS_PERMUTEV: bool = false;
    fn permutev(value: Self, _idxs: GenericArray<u32, Self::Lanes>) -> Self { value }
    fn swizzle(a: Self, b: Self, idxs: GenericArray<u32, Self::Lanes>) -> Self {
        if idxs[0] & 0b1 == 0 { a } else { b }
    }
}

#[thermite_macros::inline_always]
impl<I: UnsignedIntegerRegister<Lanes = Self::Lanes>> IndexableRegister<I> for [<u $width>] {}

#[thermite_macros::inline_always]
impl BitshiftRegister for [<u $width>] {
    const HAS_TRUE_SHIFTV:      bool = true;
    const HAS_WIDE_BYTE_SHIFTS: bool = false;

    fn bshli<const IMM8: i32>(value: Self) -> Self { value.unbounded_shl((8 * IMM8) as u32) }
    fn bshri<const IMM8: i32>(value: Self) -> Self { value.unbounded_shr((8 * IMM8) as u32) }
    fn shl (value: Self, shift: u32) -> Self { value.unbounded_shl(shift) }
    fn shr (value: Self, shift: u32) -> Self { value.unbounded_shr(shift) }
    fn shlv(value: Self, shifts: $u) -> Self { value.unbounded_shl(shifts as u32) }
    fn shrv(value: Self, shifts: $u) -> Self { value.unbounded_shr(shifts as u32) }
    fn shli<const IMM8: i32>(value: Self) -> Self { value.unbounded_shl(IMM8 as u32) }
    fn shri<const IMM8: i32>(value: Self) -> Self { value.unbounded_shr(IMM8 as u32) }

    fn rol (value: Self, shift: u32) -> Self { value.rotate_left (shift) }
    fn ror (value: Self, shift: u32) -> Self { value.rotate_right(shift) }
    fn rolv(value: Self, shifts: $u) -> Self { Self::rol(value, shifts as _) }
    fn rorv(value: Self, shifts: $u) -> Self { Self::ror(value, shifts as _) }
        // OpBitReverse: native SPIR-V instruction - override the default swap+shift chain.
    fn reverse_bits(value: Self) -> Self { unsafe { arch::op_opbitreverse::<Self>(value) } }
}

#[thermite_macros::inline_always]
impl ShuffleRegister for [<u $width>] {
    fn shuffle<const IMM8: i32>(lhs: Self, rhs: Self) -> Self {
        if IMM8 & 0b01 == 0 { lhs } else { rhs }
    }
}

#[thermite_macros::inline_always]
impl PermuteRegister for [<u $width>] {
    fn permute<const IMM8: i32>(value: Self) -> Self { value }
}

#[rustfmt::skip]
#[thermite_macros::inline_always]
impl PartialOrdRegister for [<u $width>] {
    fn gt(lhs: Self, rhs: Self) -> bool { lhs > rhs }
    fn eq(lhs: Self, rhs: Self) -> bool { lhs == rhs }
    fn ge(lhs: Self, rhs: Self) -> bool { lhs >= rhs }
    fn lt(lhs: Self, rhs: Self) -> bool { lhs < rhs }
    fn le(lhs: Self, rhs: Self) -> bool { lhs <= rhs }
    fn ne(lhs: Self, rhs: Self) -> bool { lhs != rhs }
}

#[thermite_macros::inline_always]
impl NumericRegister for [<u $width>] {
    const ZERO: Self = 0;
    const ONE:  Self = 1;
    const TWO:  Self = 2;
    const MIN:  Self = <$u>::MIN;
    const MAX:  Self = <$u>::MAX;

    fn min_element (value: Self) -> Self::Element { value }
    fn max_element (value: Self) -> Self::Element { value }
    fn sum_elements(value: Self) -> Self::Element { value }
    fn prod_elements(value: Self) -> Self::Element { value }
    fn pairwise_sum(lo: Self, hi: Self) -> Self { lo.wrapping_add(hi) }
    fn offset() -> Self { 1 }
    fn indexed() -> Self { 0 }
    fn add(lhs: Self, rhs: Self) -> Self { lhs.wrapping_add(rhs) }
    fn sub(lhs: Self, rhs: Self) -> Self { lhs.wrapping_sub(rhs) }
    fn mul(lhs: Self, rhs: Self) -> Self { lhs.wrapping_mul(rhs) }
    fn div(lhs: Self, rhs: Self) -> Self { lhs.wrapping_div(rhs) }
    fn rem(lhs: Self, rhs: Self) -> Self { lhs.wrapping_rem(rhs) }
    fn sort(value: Self) -> Self { value }
    fn min(lhs: Self, rhs: Self) -> Self { unsafe { arch::glsl_op2::<Self, Self, Self, {glsl::U_MIN}, false>(lhs, rhs) } }
    fn max(lhs: Self, rhs: Self) -> Self { unsafe { arch::glsl_op2::<Self, Self, Self, {glsl::U_MAX}, false>(lhs, rhs) } }
}

#[thermite_macros::inline_always]
impl IntegerRegister for [<u $width>] {
    fn mulhi(lhs: Self, rhs: Self) -> Self {
        // OpUMulExtended returns (lo, hi) pair; we want only hi
        unsafe { arch::spirv_umul_extended(lhs, rhs).1 }
    }

    fn mullo(lhs: Self, rhs: Self) -> Self { lhs.wrapping_mul(rhs) }

    // OpIAddCarry provides the carry bit at no extra cost; avoids widening to u64/u128.
    fn saturating_add(lhs: Self, rhs: Self) -> Self {
        let (sum, carry) = unsafe { arch::spirv_iadd_carry(lhs, rhs) };
        let overflow = carry != 0;
        unsafe { arch::op_opselect::<Self, bool>(overflow, sum, <$u>::MAX) }
    }
    // OpISubBorrow provides the borrow bit at no extra cost.
    fn saturating_sub(lhs: Self, rhs: Self) -> Self {
        let (diff, borrow) = unsafe { arch::spirv_isub_borrow(lhs, rhs) };
        let underflow = borrow != 0;
        unsafe { arch::op_opselect::<Self, bool>(underflow, diff, 0) }
    }

    fn wrapping_sum (value: Self) -> Self::Element { value }
    fn wrapping_product(value: Self) -> Self::Element { value }

    fn div_branched(value: Self, divider: crate::divider::Divider<Self::Element>) -> Self {
        divider.divide(value)
    }
    fn div_branchfree(value: Self, divider: crate::divider::BranchfreeDivider<Self::Element>) -> Self {
        divider.divide(value)
    }
    fn divv_branchfree(value: Self, dividers: crate::divider::vector::VectorDivider<Self>) -> Self {
        crate::divider::BranchfreeDivider::<$u>::new(dividers.multipliers.0, dividers.shifts.0 as i8 as u8).divide(value)
    }

    const HAS_HARDWARE_POPCNT: bool = true;

    // Use width-appropriate helpers; for 64-bit, OpBitCount is 32-bit-only so
    // we split into two halves (see spirv_count_ones_u64 in mod.rs).
    fn count_ones(value: Self) -> Self {
        unsafe { [<spirv_count_ones_u $width>](value) as $u }
    }
    fn count_zeros(value: Self) -> Self {
        unsafe { [<spirv_count_ones_u $width>](!value) as $u }
    }
    fn leading_zeros(value: Self) -> Self {
        unsafe { [<spirv_leading_zeros_u $width>](value) as $u }
    }
    fn trailing_zeros(value: Self) -> Self {
        unsafe { [<spirv_trailing_zeros_u $width>](value) as $u }
    }
    // Use BitwiseRegister::not so the inverted value stays the right type
    fn leading_ones(value: Self) -> Self { <Self as IntegerRegister>::leading_zeros(<Self as BitwiseRegister>::not(value)) }
    fn trailing_ones(value: Self) -> Self { <Self as IntegerRegister>::trailing_zeros(<Self as BitwiseRegister>::not(value)) }
}

#[thermite_macros::inline_always]
impl UnsignedIntegerRegister for [<u $width>] {
    fn next_power_of_two_m1(value: Self) -> Self {
        if value == 0 {
            return 0;
        }
        // Fill all bits below and including the MSB: MAX >> leading_zeros(value - 1)
        let lz = Self::leading_zeros(value.wrapping_sub(1));
        <$u>::MAX.unbounded_shr(lz as u32)
    }

    fn is_power_of_two(value: Self) -> bool {
        value != 0 && (value & value.wrapping_sub(1)) == 0
    }

    fn parity(value: Self) -> Self {
        <Self as IntegerRegister>::count_ones(value) & 1
    }

    // GLSLstd450 FindUMsb(x) = floor(log2(x)) for x > 0, typically 0xFFFFFFFF for x = 0.
    // ilog2p1(x) = FindUMsb(x) + 1; for x = 0, 0xFFFFFFFF + 1 wraps to 0 correctly.
    fn ilog2p1(value: Self) -> Self {
        let msb = unsafe { arch::glsl_op1::<Self, Self, { glsl::FIND_U_MSB }, false>(value) };
        msb.wrapping_add(1)
    }
}

}}} // end macro

decl_spirv_unsigned_scalar!(u32 => 32);
decl_spirv_unsigned_scalar!(u64 => 64);