metaltile-std 0.1.0

MetalTile kernel standard library — benchmark metadata and type definitions
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

//! Copyright 2026 0xClandestine, Ekryski, TheTom, Ambisphaeric
//! SPDX-License-Identifier: Apache-2.0
//! GPU correctness for `mlx::binary` — elementwise binary ops
//! (add, mul, sub, div, max, min, pow, atan2, remainder, logaddexp).
//!
//! Each test: synthetic ramp inputs → naive f32 CPU reference →
//! dispatch via Context::dispatch_with_grid → assert max_abs_diff.
//!
//! macOS-gated. Shared gpu_lock.

#![cfg(target_os = "macos")]

mod common;

use std::collections::BTreeMap;

use common::{Dt, gpu_lock, max_abs_diff, pack_bytes, unpack_bytes};
use metaltile_core::{dtype::DType, ir::Kernel};
use metaltile_runtime::Context;
use metaltile_std::mlx::binary::{
    mt_atan2,
    mt_div,
    mt_logaddexp,
    mt_max_elem,
    mt_min_elem,
    mt_mul,
    mt_pow,
    mt_sub,
};

// Dispatch a two-input, one-output binary kernel (Grid3D — one thread per element).
fn run_binary(kernel_ir: fn(DType) -> Kernel, a: &[f32], b: &[f32], dt: Dt, n: usize) -> Vec<f32> {
    let mut buffers: BTreeMap<String, Vec<u8>> = BTreeMap::new();
    buffers.insert("a".into(), pack_bytes(a, dt));
    buffers.insert("b".into(), pack_bytes(b, dt));
    buffers.insert("out".into(), pack_bytes(&vec![0.0f32; n], dt));

    let ctx = Context::new().expect("Context::new on macOS");
    let kernel = kernel_ir(dt.to_dtype());
    // binary ops are Grid3D (one thread per output element).
    let tpg = 256usize;
    let groups = n.div_ceil(tpg);
    let result = ctx
        .dispatch_with_grid(&kernel, &buffers, &BTreeMap::new(), [groups, 1, 1], [tpg, 1, 1])
        .expect("binary dispatch");
    let mut out = unpack_bytes(result.outputs.get("out").expect("out"), dt);
    out.truncate(n);
    out
}

// `vector_add` (metaltile's flagship example kernel) names its output
// param `c`, not `out`, so it does not flow through the shared
// `run_binary` helper — a correct dedicated test is a tracked follow-up.

// --- mt_mul ---

#[test]
fn binary_mul_matches_cpu_f32() {
    let _g = gpu_lock();
    let n = 1024usize;
    let a: Vec<f32> = (0..n).map(|i| (i % 17) as f32 * 0.05 - 0.4).collect();
    let b: Vec<f32> = (0..n).map(|i| (i % 13) as f32 * 0.04 - 0.25).collect();
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(x, y)| x * y).collect();
    let actual = run_binary(mt_mul::kernel_ir_for, &a, &b, Dt::F32, n);
    assert!(max_abs_diff(&actual, &expected) < 1e-5, "mul f32 mismatch");
}

// --- mt_sub ---

#[test]
fn binary_sub_matches_cpu_f32() {
    let _g = gpu_lock();
    let n = 512usize;
    let a: Vec<f32> = (0..n).map(|i| (i % 19) as f32 * 0.07 - 0.6).collect();
    let b: Vec<f32> = (0..n).map(|i| (i % 11) as f32 * 0.05 - 0.3).collect();
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(x, y)| x - y).collect();
    let actual = run_binary(mt_sub::kernel_ir_for, &a, &b, Dt::F32, n);
    assert!(max_abs_diff(&actual, &expected) < 1e-5, "sub f32 mismatch");
}

// --- mt_div ---

#[test]
fn binary_div_matches_cpu_f32() {
    let _g = gpu_lock();
    let n = 512usize;
    // Avoid near-zero denominator: shift b away from 0.
    let a: Vec<f32> = (0..n).map(|i| (i % 17) as f32 * 0.06 - 0.4).collect();
    let b: Vec<f32> = (0..n).map(|i| (i % 13) as f32 * 0.08 + 0.2).collect();
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(x, y)| x / y).collect();
    let actual = run_binary(mt_div::kernel_ir_for, &a, &b, Dt::F32, n);
    assert!(max_abs_diff(&actual, &expected) < 1e-4, "div f32 mismatch");
}

// --- mt_max_elem ---

#[test]
fn binary_max_elem_matches_cpu_f32() {
    let _g = gpu_lock();
    let n = 512usize;
    let a: Vec<f32> = (0..n).map(|i| (i % 17) as f32 * 0.05 - 0.4).collect();
    let b: Vec<f32> = (0..n).map(|i| (i % 13) as f32 * 0.06 - 0.35).collect();
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(x, y)| x.max(*y)).collect();
    let actual = run_binary(mt_max_elem::kernel_ir_for, &a, &b, Dt::F32, n);
    assert!(max_abs_diff(&actual, &expected) < 1e-5, "max_elem f32 mismatch");
}

// --- mt_min_elem ---

#[test]
fn binary_min_elem_matches_cpu_f32() {
    let _g = gpu_lock();
    let n = 512usize;
    let a: Vec<f32> = (0..n).map(|i| (i % 17) as f32 * 0.05 - 0.4).collect();
    let b: Vec<f32> = (0..n).map(|i| (i % 13) as f32 * 0.06 - 0.35).collect();
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(x, y)| x.min(*y)).collect();
    let actual = run_binary(mt_min_elem::kernel_ir_for, &a, &b, Dt::F32, n);
    assert!(max_abs_diff(&actual, &expected) < 1e-5, "min_elem f32 mismatch");
}

// --- mt_pow ---

#[test]
fn binary_pow_matches_cpu_f32() {
    let _g = gpu_lock();
    let n = 256usize;
    // Keep base positive to avoid complex-valued results.
    let a: Vec<f32> = (0..n).map(|i| (i % 9) as f32 * 0.1 + 0.2).collect();
    let b: Vec<f32> = (0..n).map(|i| (i % 5) as f32 * 0.4 + 0.2).collect();
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(x, y)| x.powf(*y)).collect();
    let actual = run_binary(mt_pow::kernel_ir_for, &a, &b, Dt::F32, n);
    assert!(max_abs_diff(&actual, &expected) < 1e-3, "pow f32 mismatch");
}

// --- mt_atan2 ---

#[test]
fn binary_atan2_matches_cpu_f32() {
    let _g = gpu_lock();
    let n = 512usize;
    let y_vals: Vec<f32> = (0..n).map(|i| (i % 17) as f32 * 0.1 - 0.8).collect();
    let x_vals: Vec<f32> = (0..n).map(|i| (i % 11) as f32 * 0.1 - 0.5).collect();
    let expected: Vec<f32> = y_vals.iter().zip(x_vals.iter()).map(|(y, x)| y.atan2(*x)).collect();
    // Kernel arg order: y, x, out — matches the fn signature mt_atan2(y, x, out).
    let mut buffers: BTreeMap<String, Vec<u8>> = BTreeMap::new();
    buffers.insert("y".into(), pack_bytes(&y_vals, Dt::F32));
    buffers.insert("x".into(), pack_bytes(&x_vals, Dt::F32));
    buffers.insert("out".into(), pack_bytes(&vec![0.0f32; n], Dt::F32));
    let ctx = Context::new().expect("Context::new");
    let kernel = mt_atan2::kernel_ir_for(DType::F32);
    let tpg = 256;
    let groups = n.div_ceil(tpg);
    let result = ctx
        .dispatch_with_grid(&kernel, &buffers, &BTreeMap::new(), [groups, 1, 1], [tpg, 1, 1])
        .expect("atan2 dispatch");
    let actual: Vec<f32> = unpack_bytes(result.outputs.get("out").expect("out"), Dt::F32)
        .into_iter()
        .take(n)
        .collect();
    assert!(max_abs_diff(&actual, &expected) < 1e-4, "atan2 f32 mismatch");
}

// mt_remainder: floored-vs-truncated division semantics need
// reconciling against the MLX reference — a correct test is a tracked
// follow-up.

// --- mt_logaddexp ---

#[test]
fn binary_logaddexp_matches_cpu_f32() {
    let _g = gpu_lock();
    let n = 512usize;
    let a: Vec<f32> = (0..n).map(|i| (i % 11) as f32 * 0.3 - 1.5).collect();
    let b: Vec<f32> = (0..n).map(|i| (i % 7) as f32 * 0.4 - 1.0).collect();
    let expected: Vec<f32> =
        a.iter().zip(b.iter()).map(|(x, y)| (x.exp() + y.exp()).ln()).collect();
    let actual = run_binary(mt_logaddexp::kernel_ir_for, &a, &b, Dt::F32, n);
    // logaddexp f32 compounds two exp + ln; tol widened to 1e-3.
    assert!(max_abs_diff(&actual, &expected) < 1e-3, "logaddexp f32 mismatch");
}