tokitai-operator 0.1.0

Verified DL kernel compiler: formally-checked GEMM, p-adic, sheaf, contract-carrying ops. Paper-artifact grade.
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

use tokitai_operator::backend::conformance::CpuOracleConformanceSuite;
use tokitai_operator::backend::hardware::{HardwareTarget, MemorySpace};
use tokitai_operator::backend::memory::{
    BufferLayout, DeviceBuffer, TransferDirection, TransferFallbackReason, TransferPlan,
    TransferStatus,
};
use tokitai_operator::domain::{Domain, DomainId, PadicDomain};
use tokitai_operator::object::sheaf::Cover;
use tokitai_operator::object::{DeviceId, ObjectMeta, Representation, Shape};

#[test]
fn p182_cpu_dense_tensor_maps_to_host_device_buffer_without_changing_domain() {
    let meta = ObjectMeta::tensor(
        DomainId::new("integer"),
        Shape::from(vec![2, 3]),
        Representation::dense_cpu(),
    );
    let buffer = DeviceBuffer::host_tensor("dense-i64", HardwareTarget::cpu_scalar(), meta, 8)
        .with_oracle_fixture("dense_i64_add_reduce");

    assert_eq!(buffer.memory_space, MemorySpace::Host);
    assert_eq!(buffer.layout, BufferLayout::Contiguous { element_count: 6 });
    assert_eq!(buffer.byte_len, Some(48));
    assert_eq!(buffer.meta.domain.0, "integer");
    assert_eq!(
        buffer.oracle_fixture.as_deref(),
        Some("dense_i64_add_reduce")
    );
    assert!(buffer.is_host_accessible());
}

#[test]
fn p182_padic_and_sheaf_witnesses_remain_representable_as_buffer_metadata() {
    let q5 = PadicDomain::new(5, 3).unwrap();
    let padic_meta = ObjectMeta::tensor(
        q5.id(),
        Shape::from(vec![4]),
        Representation::DenseTensor {
            layout: tokitai_operator::object::Layout::PadicLimbArray,
            device: DeviceId::cpu(),
        },
    );
    let padic_buffer =
        DeviceBuffer::host_tensor("padic", HardwareTarget::cpu_scalar(), padic_meta, 16)
            .with_oracle_fixture("padic_valuation_skip_sum_products");
    assert_eq!(
        padic_buffer.layout,
        BufferLayout::PadicLimbArray {
            element_count: 4,
            precision_digits: 3
        }
    );
    assert_eq!(
        padic_buffer.oracle_fixture.as_deref(),
        Some("padic_valuation_skip_sum_products")
    );

    let cover = Cover::new("U", ["A", "B"]);
    let sheaf_meta = ObjectMeta::tensor(
        DomainId::new(format!("cover:{}", cover.target.0)),
        Shape::from(vec![cover.opens.len()]),
        Representation::CoverIndexedSection {
            device: DeviceId::cpu(),
        },
    );
    let sheaf_buffer =
        DeviceBuffer::host_tensor("sheaf", HardwareTarget::cpu_scalar(), sheaf_meta, 8)
            .with_oracle_fixture("sheaf_finite_site_glue");
    assert_eq!(
        sheaf_buffer.layout,
        BufferLayout::CoverIndexed {
            open_count: cover.opens.len()
        }
    );
    assert_eq!(
        sheaf_buffer.oracle_fixture.as_deref(),
        Some("sheaf_finite_site_glue")
    );
}

#[test]
fn p182_transfer_plan_records_noop_host_transfer_for_cpu_oracle_fixture() {
    let report = CpuOracleConformanceSuite::new().run_all().unwrap();
    let fixture = report
        .fixtures
        .iter()
        .find(|fixture| fixture.name == "dense_i64_add_reduce")
        .unwrap();
    let meta = ObjectMeta::tensor(
        DomainId::new("integer"),
        Shape::from(vec![4]),
        Representation::dense_cpu(),
    );
    let source = DeviceBuffer::host_tensor("src", HardwareTarget::cpu_scalar(), meta.clone(), 8)
        .with_oracle_fixture(&fixture.name);
    let destination = DeviceBuffer::host_tensor("dst", HardwareTarget::cpu_scalar(), meta, 8)
        .with_oracle_fixture(&fixture.name);
    let transfer = TransferPlan::plan(source, destination);

    assert_eq!(transfer.direction, TransferDirection::HostToHost);
    assert_eq!(transfer.status, TransferStatus::NoOp);
    assert!(transfer.preserves_semantics);
    assert!(transfer.reason.is_none());
    transfer.require_supported().unwrap();
}

#[test]
fn p182_transfer_plan_rejects_unimplemented_host_to_device_path_explicitly() {
    let meta = ObjectMeta::tensor(
        DomainId::new("integer"),
        Shape::from(vec![2]),
        Representation::dense_cpu(),
    );
    let source = DeviceBuffer::host_tensor("src", HardwareTarget::cpu_scalar(), meta.clone(), 8);
    let mut destination = DeviceBuffer::host_tensor("dst", HardwareTarget::cpu_scalar(), meta, 8);
    destination.memory_space = MemorySpace::Device;

    let transfer = TransferPlan::plan(source, destination);
    assert_eq!(transfer.direction, TransferDirection::HostToDevice);
    assert!(matches!(
        transfer.status,
        TransferStatus::Unsupported(TransferFallbackReason::MissingDeviceMemory {
            source: MemorySpace::Host,
            destination: MemorySpace::Device
        })
    ));
    assert!(!transfer.preserves_semantics);
    assert!(
        transfer
            .require_supported()
            .unwrap_err()
            .to_string()
            .contains("missing device-memory implementation")
    );
}