paulistrings 0.1.2

Classical simulation of quantum circuits in the Pauli basis, parallel across threads, NUMA domains, MPI ranks and GPUs
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

//! Shared fixtures for the crate's own tests, benches and examples, re-exported flat from the submodules.
//!
//! Compiled only under `cfg(test)` or the `test-utils` feature (the crate's self-dev-dependency); `#[doc(hidden)]`, not public API.

mod circuits;
mod compare;
#[cfg(feature = "cuda")]
mod gpu;
mod oracle;
mod random;
mod rows;
mod transport;

pub use circuits::*;
pub use compare::*;
#[cfg(feature = "cuda")]
pub use gpu::{
    assert_invariants_device, from_host_with_nvrtc_options, refine_to, set_fault_plan, DeviceSplit,
    FaultPlan, WireFault,
};
pub use oracle::*;
pub use random::*;
pub use rows::{circuit_generators, count_remote_deltas, GeneratorWeight};
pub use transport::*;

pub use crate::engine::bucketed::apply_layer_bucketed;
pub use crate::engine::partitioned::in_process::InProcessTransport;

/// Layers between two bucket-count agreements of the partitioned driver.
pub const BITS_AGREE_EVERY: usize = crate::engine::driver::BITS_AGREE_EVERY;

pub use crate::pauli_sum::hash::B_MAX_BITS;
pub use crate::pauli_sum::storage::{
    desired_bits, DEFAULT_HASH_SEED, DEFAULT_MIN_BUCKETS, DEFAULT_TARGET_BUCKET_LEN,
};

/// Rough estimate of the cost of one `Instant::now()` read, in nanoseconds, for this hardware class; used by the `phase_breakdown` probe's overhead line (`timer_reads() * TIMER_READ_OVERHEAD_NS`).
#[cfg(feature = "phase-timing")]
pub const TIMER_READ_OVERHEAD_NS: u64 = 25;

const ZERO: num_complex::Complex64 = num_complex::Complex64::new(0.0, 0.0);

/// Returns early from the calling test when no CUDA device is available, so a device test is
/// never `#[ignore]`d — it just does nothing on a box without a GPU.
#[cfg(feature = "cuda")]
#[macro_export]
macro_rules! require_cuda {
    () => {
        if !$crate::gpu::cuda_available() {
            return;
        }
    };
}

/// `transport` with its largest message capped at `bytes`, so a test reaches the chunked exchange.
#[cfg(feature = "mpi")]
pub fn mpi_with_chunk_bytes(
    transport: crate::mpi::MpiTransport,
    bytes: usize,
) -> crate::mpi::MpiTransport {
    transport.with_chunk_bytes(bytes)
}

/// The MPI library version string this build probed through `mpicc`.
#[cfg(feature = "mpi")]
pub fn mpi_build_library_version() -> &'static str {
    crate::mpi::MpiTransport::build_library_version()
}

#[cfg(test)]
mod tests;