Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Oxidize the numeric code in the Isometry gate class #12197

Merged
merged 25 commits into from
Apr 30, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
25 commits
Select commit Hold shift + click to select a range
6dca86e
Oxidize the numeric code in the Isometry gate class
mtreinish Apr 16, 2024
5c21069
Remove unused import
mtreinish Apr 17, 2024
cf3c5ff
Use rust impl for small utility functions too
mtreinish Apr 17, 2024
8107fb2
Oxidize the linalg in UCGate too
mtreinish Apr 17, 2024
63e19e3
Merge remote-tracking branch 'origin/main' into isometry-rust
mtreinish Apr 17, 2024
b0b7a33
Migrate another numeric helper method of UCGate
mtreinish Apr 17, 2024
eaf43f0
Remove now unused code paths
mtreinish Apr 18, 2024
23f5e2e
Remove bitstring usage with bitwise ops
mtreinish Apr 19, 2024
cb8fc67
Mostly replace Vec<u8> usage with bitwise operations
mtreinish Apr 19, 2024
7b61dcc
Merge branch 'main' into isometry-rust
mtreinish Apr 19, 2024
6a6f3bc
Apply suggestions from code review
mtreinish Apr 26, 2024
6e86efe
Remove python side call sites
mtreinish Apr 26, 2024
3676469
Fix integer typing in uc_gate.rs
mtreinish Apr 26, 2024
29ffc99
Simplify basis state bitshift loop logic
mtreinish Apr 26, 2024
366a934
Build set of control labels outside construct_basis_states
mtreinish Apr 26, 2024
4843ef5
Use 2 element array for reverse_qubit_state
mtreinish Apr 26, 2024
d26bd4a
Micro optimize reverse_qubit_state
mtreinish Apr 26, 2024
091d229
Use 1d numpy arrays for diagonal inputs
mtreinish Apr 26, 2024
d540511
Merge remote-tracking branch 'origin/main' into isometry-rust
mtreinish Apr 26, 2024
604e33a
Fix lint
mtreinish Apr 26, 2024
3bd4358
Update crates/accelerate/src/isometry.rs
mtreinish Apr 29, 2024
c04aa17
Merge branch 'main' into isometry-rust
mtreinish Apr 29, 2024
b8b82a6
Add back sqrt() accidentally removed by inline suggestion
mtreinish Apr 29, 2024
fcf587f
Use a constant for rz pi/2 elements
mtreinish Apr 30, 2024
2c07b79
Merge remote-tracking branch 'origin/main' into isometry-rust
mtreinish Apr 30, 2024
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 10 additions & 0 deletions Cargo.lock

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

1 change: 1 addition & 0 deletions crates/accelerate/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ num-complex = "0.4"
num-bigint = "0.4"
rustworkx-core = "0.14"
faer = "0.18.2"
itertools = "0.12.1"
qiskit-circuit.workspace = true

[dependencies.smallvec]
Expand Down
367 changes: 367 additions & 0 deletions crates/accelerate/src/isometry.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,367 @@
// This code is part of Qiskit.
//
// (C) Copyright IBM 2024
//
// This code is licensed under the Apache License, Version 2.0. You may
// obtain a copy of this license in the LICENSE.txt file in the root directory
// of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
//
// Any modifications or derivative works of this code must retain this
// copyright notice, and modified files need to carry a notice indicating
// that they have been altered from the originals.

use std::ops::BitAnd;

use approx::abs_diff_eq;
use num_complex::{Complex64, ComplexFloat};
use pyo3::prelude::*;
use pyo3::wrap_pyfunction;
use pyo3::Python;

use hashbrown::HashSet;
use itertools::Itertools;
use ndarray::prelude::*;
use numpy::{IntoPyArray, PyReadonlyArray1, PyReadonlyArray2};

use crate::two_qubit_decompose::ONE_QUBIT_IDENTITY;

/// Find special unitary matrix that maps [c0,c1] to [r,0] or [0,r] if basis_state=0 or
/// basis_state=1 respectively
#[pyfunction]
pub fn reverse_qubit_state(
py: Python,
state: [Complex64; 2],
basis_state: usize,
epsilon: f64,
) -> PyObject {
reverse_qubit_state_inner(&state, basis_state, epsilon)
.into_pyarray_bound(py)
.into()
}

#[inline(always)]
fn l2_norm(vec: &[Complex64]) -> f64 {
Copy link
Contributor

@jlapeyre jlapeyre Apr 29, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd prefer that this were somehow called the 2-norm or p-norm for p=2 ("somehow" with a valid identifier). l2 is a space, usually a sequence space. But upon googling for uses of p norm, l2 norm, etc. it seems that, as far as the internet is concerned, especially machine-learning internet, lp, Lp, can mean whatever you want them to mean. [EDIT. looks like this is translated from np.linalg.norm, which, along with Julia's LinearAlgebra, calls this the 2-norm)]

We might want a function for p norms with p as a parameter, with 2 as the default, and make sure constants are propagated, or whatever it takes to get the correct norm at compile time. But, unless we have an existing use for p != 2 that can be postponed indefinitely.

In any case, at some point, we will wish that functions like this had been collected in a repo-wide module. (probably not in the ten thousandth "utils.xxx"). And in that case it's worth having somewhat more general trait bounds. Not repeating this function could be important for correctness and performance. For example, the Julia 2-norm function appears to sometimes scale the elements to avoid overflow. However, the function mynorm(v) = sqrt(sum(x -> x^2, v)); is more than 3 times faster than top-level entry point norm. We'd want an api and implementation that reflects our common uses, which might be almost all 2- and 4-element vectors. Indeed, an optimization, which we don't need to make at the moment, would take advantage the length of the vector, which is known here at compile time to be two.

The main point for now is to consider at least collecting these somewhere where they are visible so that the questions of performance and correctness can be more easily addressed at some point

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't want to get into a bikeshedding conversation about naming, but l2_norm or lp_norm is a pretty standard terminology for this. If you look at the comment in the Julia code below what you linked they use it there too:

https://github.com/JuliaLang/julia/blob/68da780b93518204d874410307791702d5200e29/stdlib/LinearAlgebra/src/generic.jl#L496

# Compute L_p norm ‖x‖ₚ = sum(abs(x).^p)^(1/p)

That being said the response to the main comment here is that we can look at creating a dedicated mathematical functions module, or even a separate crate if it's generic enough for these kind of things in a follow up PR. Ideally we'd contribute this to something like ndarray or faer imo though assuming a generic implementation. I only did it like this because it was so small and simple and the usage was very minimal. So far the only function I think that falls into this category is the 2x2 determinant function which is used in this PR, the two qubit decomposer, and the one qubit decomposer (where it was originally added). But again it's just a single line so I didn't feel like it was worth creating a dedicate module for it.

vec.iter()
.fold(0., |acc, elem| acc + elem.norm_sqr())
.sqrt()
}

fn reverse_qubit_state_inner(
state: &[Complex64; 2],
basis_state: usize,
epsilon: f64,
) -> Array2<Complex64> {
let r = l2_norm(state);
let r_inv = 1. / r;
if r < epsilon {
Array2::eye(2)
} else if basis_state == 0 {
array![
[state[0].conj() * r_inv, state[1].conj() * r_inv],
[-state[1] * r_inv, state[0] * r_inv],
]
} else {
array![
[-state[1] * r_inv, state[0] * r_inv],
[state[0].conj() * r_inv, state[1].conj() * r_inv],
]
}
}

/// This method finds the single-qubit gates for a UCGate to disentangle a qubit:
/// we consider the n-qubit state v[:,0] starting with k zeros (in the computational basis).
/// The qubit with label n-s-1 is disentangled into the basis state k_s(k,s).

#[pyfunction]
pub fn find_squs_for_disentangling(
mtreinish marked this conversation as resolved.
Show resolved Hide resolved
py: Python,
v: PyReadonlyArray2<Complex64>,
k: usize,
s: usize,
epsilon: f64,
n: usize,
) -> Vec<PyObject> {
let v = v.as_array();
let k_prime = 0;
let i_start = if b(k, s + 1) == 0 {
a(k, s + 1)
} else {
a(k, s + 1) + 1
};
let mut output: Vec<Array2<Complex64>> = (0..i_start).map(|_| Array2::eye(2)).collect();
let mut squs: Vec<Array2<Complex64>> = (i_start..2_usize.pow((n - s - 1) as u32))
.map(|i| {
reverse_qubit_state_inner(
&[
v[[2 * i * 2_usize.pow(s as u32) + b(k, s), k_prime]],
v[[(2 * i + 1) * 2_usize.pow(s as u32) + b(k, s), k_prime]],
],
k_s(k, s),
epsilon,
)
})
.collect();
output.append(&mut squs);
output
.into_iter()
.map(|x| x.into_pyarray_bound(py).into())
.collect()
}

#[pyfunction]
pub fn apply_ucg(
py: Python,
m: PyReadonlyArray2<Complex64>,
k: usize,
single_qubit_gates: Vec<PyReadonlyArray2<Complex64>>,
) -> PyObject {
let mut m = m.as_array().to_owned();
let shape = m.shape();
let num_qubits = shape[0].ilog2();
let num_col = shape[1];
let spacing: usize = 2_usize.pow(num_qubits - k as u32 - 1);
for j in 0..2_usize.pow(num_qubits - 1) {
let i = (j / spacing) * spacing + j;
let gate_index = i / (2_usize.pow(num_qubits - k as u32));
for col in 0..num_col {
let gate = single_qubit_gates[gate_index].as_array();
let a = m[[i, col]];
let b = m[[i + spacing, col]];
m[[i, col]] = gate[[0, 0]] * a + gate[[0, 1]] * b;
m[[i + spacing, col]] = gate[[1, 0]] * a + gate[[1, 1]] * b;
}
}
m.into_pyarray_bound(py).into()
}

#[inline(always)]
fn bin_to_int(bin: &[u8]) -> usize {
bin.iter()
.fold(0_usize, |acc, digit| (acc << 1) + *digit as usize)
}

#[pyfunction]
pub fn apply_diagonal_gate(
py: Python,
m: PyReadonlyArray2<Complex64>,
action_qubit_labels: Vec<usize>,
diag: PyReadonlyArray1<Complex64>,
) -> PyResult<PyObject> {
let diag = diag.as_slice()?;
let mut m = m.as_array().to_owned();
let shape = m.shape();
let num_qubits = shape[0].ilog2();
let num_col = shape[1];
for state in std::iter::repeat([0_u8, 1_u8])
.take(num_qubits as usize)
.multi_cartesian_product()
{
let diag_index = action_qubit_labels
.iter()
.fold(0_usize, |acc, i| (acc << 1) + state[*i] as usize);
let i = bin_to_int(&state);
Comment on lines +155 to +162
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Completely pre-existing, but holy moly. This is a wild way of spelling for i in 0..2.pow(num_qubits) {} (and swapping to corresponding bit-shift logic on diag_index).

No need to change it here unless you felt like it - the whole implementation needs a bit more TLC than that.

for j in 0..num_col {
m[[i, j]] = diag[diag_index] * m[[i, j]]
}
}
Ok(m.into_pyarray_bound(py).into())
}

#[pyfunction]
pub fn apply_diagonal_gate_to_diag(
mut m_diagonal: Vec<Complex64>,
action_qubit_labels: Vec<usize>,
diag: PyReadonlyArray1<Complex64>,
num_qubits: usize,
) -> PyResult<Vec<Complex64>> {
let diag = diag.as_slice()?;
if m_diagonal.is_empty() {
return Ok(m_diagonal);
}
for state in std::iter::repeat([0_u8, 1_u8])
.take(num_qubits)
.multi_cartesian_product()
.take(m_diagonal.len())
{
let diag_index = action_qubit_labels
.iter()
.fold(0_usize, |acc, i| (acc << 1) + state[*i] as usize);
let i = bin_to_int(&state);
m_diagonal[i] *= diag[diag_index]
}
Ok(m_diagonal)
}

/// Helper method for _apply_multi_controlled_gate. This constructs the basis states the MG gate
/// is acting on for a specific state state_free of the qubits we neither control nor act on
fn construct_basis_states(
state_free: &[u8],
control_set: &HashSet<usize>,
target_label: usize,
) -> [usize; 2] {
let size = state_free.len() + control_set.len() + 1;
let mut e1: usize = 0;
let mut e2: usize = 0;
let mut j = 0;
for i in 0..size {
e1 <<= 1;
e2 <<= 1;
if control_set.contains(&i) {
e1 += 1;
e2 += 1;
} else if i == target_label {
e2 += 1;
} else {
assert!(j <= 1);
e1 += state_free[j] as usize;
e2 += state_free[j] as usize;
j += 1
}
}
[e1, e2]
}

#[pyfunction]
pub fn apply_multi_controlled_gate(
py: Python,
m: PyReadonlyArray2<Complex64>,
control_labels: Vec<usize>,
target_label: usize,
gate: PyReadonlyArray2<Complex64>,
) -> PyObject {
let mut m = m.as_array().to_owned();
let gate = gate.as_array();
let shape = m.shape();
let num_qubits = shape[0].ilog2();
let num_col = shape[1];
let free_qubits = num_qubits as usize - control_labels.len() - 1;
let control_set: HashSet<usize> = control_labels.into_iter().collect();
if free_qubits == 0 {
let [e1, e2] = construct_basis_states(&[], &control_set, target_label);
for i in 0..num_col {
let temp: Vec<_> = gate
.dot(&aview2(&[[m[[e1, i]]], [m[[e2, i]]]]))
.into_iter()
.take(2)
.collect();
m[[e1, i]] = temp[0];
m[[e2, i]] = temp[1];
}
return m.into_pyarray_bound(py).into();
}
for state_free in std::iter::repeat([0_u8, 1_u8])
.take(free_qubits)
.multi_cartesian_product()
{
let [e1, e2] = construct_basis_states(&state_free, &control_set, target_label);
for i in 0..num_col {
let temp: Vec<_> = gate
.dot(&aview2(&[[m[[e1, i]]], [m[[e2, i]]]]))
.into_iter()
.take(2)
.collect();
m[[e1, i]] = temp[0];
m[[e2, i]] = temp[1];
}
}
m.into_pyarray_bound(py).into()
}

#[pyfunction]
pub fn ucg_is_identity_up_to_global_phase(
single_qubit_gates: Vec<PyReadonlyArray2<Complex64>>,
epsilon: f64,
) -> bool {
let global_phase: Complex64 = if single_qubit_gates[0].as_array()[[0, 0]].abs() >= epsilon {
single_qubit_gates[0].as_array()[[0, 0]].finv()
} else {
return false;
};
for raw_gate in single_qubit_gates {
let gate = raw_gate.as_array();
if !abs_diff_eq!(
gate.mapv(|x| x * global_phase),
aview2(&ONE_QUBIT_IDENTITY),
epsilon = 1e-8 // Default tolerance from numpy for allclose()
) {
return false;
}
}
true
}
Comment on lines +270 to +291
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is pre-existing, but global_phase is not the global phase until its abs happens to be 1. This test is actually testing whether the gate is close to a scaled identity. Same with diag_is_identity_up_to_global_phase below.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe global_phase should be renamed global_scale if it can't be constrained always to have magnitude 1. And likewise, this function should be renamed. But maybe these can wait for a rewrite.


#[pyfunction]
fn diag_is_identity_up_to_global_phase(diag: Vec<Complex64>, epsilon: f64) -> bool {
let global_phase: Complex64 = if diag[0].abs() >= epsilon {
diag[0].finv()
} else {
return false;
};
for d in diag {
if (global_phase * d - 1.0).abs() >= epsilon {
return false;
}
}
true
}

#[pyfunction]
pub fn merge_ucgate_and_diag(
py: Python,
single_qubit_gates: Vec<PyReadonlyArray2<Complex64>>,
diag: Vec<Complex64>,
) -> Vec<PyObject> {
single_qubit_gates
.iter()
.enumerate()
.map(|(i, raw_gate)| {
let gate = raw_gate.as_array();
let res = aview2(&[
[diag[2 * i], Complex64::new(0., 0.)],
[Complex64::new(0., 0.), diag[2 * i + 1]],
])
.dot(&gate);
res.into_pyarray_bound(py).into()
})
.collect()
}

#[inline(always)]
#[pyfunction]
fn k_s(k: usize, s: usize) -> usize {
if k == 0 {
0
} else {
let filter = 1 << s;
k.bitand(filter) >> s
}
}

#[inline(always)]
#[pyfunction]
fn a(k: usize, s: usize) -> usize {
k / 2_usize.pow(s as u32)
}

#[inline(always)]
#[pyfunction]
fn b(k: usize, s: usize) -> usize {
k - (a(k, s) * 2_usize.pow(s as u32))
}

#[pymodule]
pub fn isometry(m: &Bound<PyModule>) -> PyResult<()> {
m.add_wrapped(wrap_pyfunction!(diag_is_identity_up_to_global_phase))?;
m.add_wrapped(wrap_pyfunction!(find_squs_for_disentangling))?;
m.add_wrapped(wrap_pyfunction!(reverse_qubit_state))?;
m.add_wrapped(wrap_pyfunction!(apply_ucg))?;
m.add_wrapped(wrap_pyfunction!(apply_diagonal_gate))?;
m.add_wrapped(wrap_pyfunction!(apply_diagonal_gate_to_diag))?;
m.add_wrapped(wrap_pyfunction!(apply_multi_controlled_gate))?;
m.add_wrapped(wrap_pyfunction!(ucg_is_identity_up_to_global_phase))?;
m.add_wrapped(wrap_pyfunction!(merge_ucgate_and_diag))?;
m.add_wrapped(wrap_pyfunction!(a))?;
m.add_wrapped(wrap_pyfunction!(b))?;
m.add_wrapped(wrap_pyfunction!(k_s))?;
Ok(())
}
Loading
Loading