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.

Sign up for GitHub

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

Move AVX2 logic into Transformer class and enable SIMD on Unitary, DensityMatrix, SuperOp #961

Merged
merged 9 commits into from
Oct 2, 2020
Merged

Move AVX2 logic into Transformer class and enable SIMD on Unitary, DensityMatrix, SuperOp #961

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
824bfe8
Move AVX2 logic into QubitVector class
chriseclectic Sep 28, 2020
6f4d196
Add release note
chriseclectic Sep 28, 2020
d7954e3
Add Transformer class for running apply_matrix on container
chriseclectic Sep 29, 2020
e2cd919
Update qubitvector to use Transformer class for apply_matrix and appl…
chriseclectic Sep 29, 2020
efb82cc
Fix bug in TransformerAVX2 apply_matrix
vvilpas Sep 30, 2020
e60aa20
Remove CTRP from Transformer, move container template to functions
chriseclectic Sep 30, 2020
d007c4a
Make transformer functions static
chriseclectic Sep 30, 2020
9840717
Use virtual functions with transformer classes
vvilpas Oct 2, 2020
d00d863
Merge branch 'master' into refactor/qubitvector-avx2
chriseclectic Oct 2, 2020
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
9 changes: 9 additions & 0 deletions releasenotes/notes/simd-support-c2352ca3d639770f.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,9 @@
---
features:
- |
Extends the SIMD vectorization of the statevector simulation method to the
unitary matrix, superoperator matrix, and density matrix simulation methods.
This gives roughtly a 2x performance increase general simulation using the
:class:`~qiskit.providers.aer.UnitarySimulator`, the ``"density_matrix"``
method of the :class:`~qiskit.providers.aer.QasmSimulator`, gate
fusion, and noise simulation.
17 changes: 0 additions & 17 deletions src/controllers/qasm_controller.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,13 +16,11 @@
#define _aer_qasm_controller_hpp_

#include "controller.hpp"
#include "framework/avx2_detect.hpp"
#include "simulators/density_matrix/densitymatrix_state.hpp"
#include "simulators/extended_stabilizer/extended_stabilizer_state.hpp"
#include "simulators/matrix_product_state/matrix_product_state.hpp"
#include "simulators/stabilizer/stabilizer_state.hpp"
#include "simulators/statevector/qubitvector.hpp"
#include "simulators/statevector/qubitvector_avx2.hpp"
#include "simulators/statevector/statevector_state.hpp"
#include "simulators/superoperator/superoperator_state.hpp"
#include "transpile/delay_measure.hpp"
Expand Down Expand Up @@ -395,27 +393,12 @@ void QasmController::run_circuit(const Circuit& circ,
// Validate circuit for simulation method
switch (simulation_method(circ, noise, true)) {
case Method::statevector: {
bool avx2_enabled = is_avx2_supported();

if (simulation_precision_ == Precision::double_precision) {
if (avx2_enabled) {
return run_circuit_helper<
Statevector::State<QV::QubitVectorAvx2<double>>>(
circ, noise, config, shots, rng_seed, initial_statevector_,
Method::statevector, data);
}
// Double-precision Statevector simulation
return run_circuit_helper<Statevector::State<QV::QubitVector<double>>>(
circ, noise, config, shots, rng_seed, initial_statevector_,
Method::statevector, data);
} else {
if (avx2_enabled) {
// Single-precision Statevector simulation
return run_circuit_helper<
Statevector::State<QV::QubitVectorAvx2<float>>>(
circ, noise, config, shots, rng_seed, initial_statevector_,
Method::statevector, data);
}
// Single-precision Statevector simulation
return run_circuit_helper<Statevector::State<QV::QubitVector<float>>>(
circ, noise, config, shots, rng_seed, initial_statevector_,
Expand Down
10 changes: 0 additions & 10 deletions src/controllers/statevector_controller.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,6 @@

#include "controller.hpp"
#include "simulators/statevector/statevector_state.hpp"
#include "simulators/statevector/qubitvector_avx2.hpp"
#include "transpile/fusion.hpp"

namespace AER {
Expand Down Expand Up @@ -205,21 +204,12 @@ void StatevectorController::run_circuit(
switch (method_) {
case Method::automatic:
case Method::statevector_cpu: {
bool avx2_enabled = is_avx2_supported();
if (precision_ == Precision::double_precision) {
if(avx2_enabled){
return run_circuit_helper<Statevector::State<QV::QubitVectorAvx2<double>>>(
circ, noise, config, shots, rng_seed, data);
}
// Double-precision Statevector simulation
return run_circuit_helper<Statevector::State<QV::QubitVector<double>>>(
circ, noise, config, shots, rng_seed, data);
} else {
// Single-precision Statevector simulation
if(avx2_enabled){
return run_circuit_helper<Statevector::State<QV::QubitVectorAvx2<float>>>(
circ, noise, config, shots, rng_seed, data);
}
return run_circuit_helper<Statevector::State<QV::QubitVector<float>>>(
circ, noise, config, shots, rng_seed, data);
}
Expand Down
12 changes: 6 additions & 6 deletions src/simulators/statevector/indexes.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -177,7 +177,7 @@ inline void apply_lambda(const size_t start,
const uint_t omp_threads,
Lambda&& func) {

#pragma omp parallel if (omp_threads > 0) num_threads(omp_threads)
#pragma omp parallel if (omp_threads > 1) num_threads(omp_threads)
{
#pragma omp for
for (int_t k = int_t(start); k < int_t(stop); k++) {
Expand All @@ -197,7 +197,7 @@ inline void apply_lambda(const size_t start,
const int_t END = stop >> NUM_QUBITS;
auto qubits_sorted = qubits;
std::sort(qubits_sorted.begin(), qubits_sorted.end());
#pragma omp parallel if (omp_threads > 0) num_threads(omp_threads)
#pragma omp parallel if (omp_threads > 1) num_threads(omp_threads)
{
#pragma omp for
for (int_t k = int_t(start); k < END; k++) {
Expand All @@ -222,7 +222,7 @@ inline void apply_lambda(const size_t start,
auto qubits_sorted = qubits;
std::sort(qubits_sorted.begin(), qubits_sorted.end());

#pragma omp parallel if (omp_threads > 0) num_threads(omp_threads)
#pragma omp parallel if (omp_threads > 1) num_threads(omp_threads)
{
#pragma omp for
for (int_t k = int_t(start); k < END; k+=gap) {
Expand Down Expand Up @@ -254,7 +254,7 @@ inline std::complex<double> apply_reduction_lambda(const size_t start,
// Reduction variables
double val_re = 0.;
double val_im = 0.;
#pragma omp parallel reduction(+:val_re, val_im) if (omp_threads > 0) num_threads(omp_threads)
#pragma omp parallel reduction(+:val_re, val_im) if (omp_threads > 1) num_threads(omp_threads)
{
#pragma omp for
for (int_t k = int_t(start); k < int_t(stop); k++) {
Expand All @@ -279,7 +279,7 @@ std::complex<double> apply_reduction_lambda(const size_t start,
// Reduction variables
double val_re = 0.;
double val_im = 0.;
#pragma omp parallel reduction(+:val_re, val_im) if (omp_threads > 0) num_threads(omp_threads)
#pragma omp parallel reduction(+:val_re, val_im) if (omp_threads > 1) num_threads(omp_threads)
{
#pragma omp for
for (int_t k = int_t(start); k < END; k++) {
Expand Down Expand Up @@ -308,7 +308,7 @@ std::complex<double> apply_reduction_lambda(const size_t start,
// Reduction variables
double val_re = 0.;
double val_im = 0.;
#pragma omp parallel reduction(+:val_re, val_im) if (omp_threads > 0) num_threads(omp_threads)
#pragma omp parallel reduction(+:val_re, val_im) if (omp_threads > 1) num_threads(omp_threads)
{
#pragma omp for
for (int_t k = int_t(start); k < END; k++) {
Expand Down
Loading