Refactor qubit high performance functions

Signed-off-by: Vlad Gheorghiu <vsoftco@gmail.com>

Author: vsoftco

CHANGES.md

@@ -4,6 +4,8 @@
   manipulation and qubit circuit execution through extensive optimizations of
   `qpp::apply()`, `qpp::applyCTRL()`, and `qpp::measure_seq()`.
 - **pyqpp** is now feature complete, providing Python users with
+- Added `pyqpp.QEngine.get_stats_to_JSON()` to display the measurement
+  statistics in JSON format from **pyqpp**
   complete access to the full **Quantum++** API
 - Switched to MAJOR.MINOR.PATCH release versioning system:
   - MAJOR -- introduces significant changes that break backward compatibility
@@ -40,11 +42,8 @@
   `${CMAKE_BUILD_DIR}`
 - Added benchmarking suite using [Catch2](https://github.com/catchorg/Catch2)
   in ["benchmarks"](https://github.com/softwareQinc/qpp/blob/main/benchmarks)
+- Removed the `stress_tests` suite, as the new benchmarking suite supersedes it
 - Bugfixes for `qpp::QFT()` and `qpp::TFQ()`
-- Added `pyqpp.QEngine.get_stats_to_JSON()` to display the measurement
-  statistics in JSON format from **pyqpp**
-- **pyqpp** is feature-complete, now all Quantum++ functions have corresponding
-  bindings in Python
 
 # Version 6.0 - 14 April 2025
 

benchmarks/CMakeLists.txt

@@ -1,11 +1,12 @@
 # Benchmarks with Catch2
 cmake_minimum_required(VERSION 3.20)
 
-project(benchmarks)
+project(benchmarks LANGUAGES CXX)
 
 find_package(qpp REQUIRED)
 
 set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
 include(FetchContent)

benchmarks/README.md

@@ -43,7 +43,7 @@ a specific benchmark, e.g., `qft_bench`, execute
 For help and additional options, run any benchmark with the `--help` flag.
 
 > **Note:**
-> All **Quantum++**
+> **Quantum++**
 > [CMake flags and arguments](https://github.com/softwareQinc/qpp/blob/main/INSTALL.md#cmake-flags-and-optional-arguments)
 > **do not propagate** to the benchmarks. They must be defined independently if
 > they are needed.

benchmarks/include/openmp_utils.hpp

@@ -9,42 +9,42 @@
 #include <omp.h>
 
 // CLI override (default -1 -> auto-detect)
-inline int cli_core_count = -1;
+inline int cli_threads = -1;
 
 /**
  * @brief Detects/sets the number of OpenMP threads
- * @param requested Optional requested thread count; uses \a cli_core_count by
+ * @param requested Optional requested thread count; uses \a cli_threads by
  * default
  *
  * @return Final number of threads used
  */
 inline int set_openmp_threads(int requested = -1) {
-    int core_count = (requested >= 0) ? requested : cli_core_count;
+    int threads = (requested >= 0) ? requested : cli_threads;
 
     const char* env = std::getenv("OMP_NUM_THREADS");
     const bool env_set = (env && std::atoi(env) > 0);
 
     // 1. Environment variable has highest priority
     if (env_set) {
-        core_count = std::atoi(env);
-        std::cout << "OpenMP enabled, core_count=" << core_count
+        threads = std::atoi(env);
+        std::cout << "OpenMP enabled, threads=" << threads
                   << " set by OMP_NUM_THREADS\n";
     }
     // 2. If NOT set by OMP_NUM_THREADS
-    else if (core_count <= 0) {
-        core_count = omp_get_max_threads();
-        std::cout << "OpenMP enabled, core_count=" << core_count
+    else if (threads <= 0) {
+        threads = omp_get_max_threads();
+        std::cout << "OpenMP enabled, threads=" << threads
                   << " auto-detected by OpenMP\n";
     }
-    // 3. Otherwise (core_count > 0) and NOT set by OMP_NUM_THREADS
+    // 3. Otherwise (threads > 0) and NOT set by OMP_NUM_THREADS
     else {
-        std::cout << "OpenMP enabled, core_count=" << core_count
+        std::cout << "OpenMP enabled, threads=" << threads
                   << " set by the CLI\n";
     }
 
-    omp_set_num_threads(core_count);
+    omp_set_num_threads(threads);
 
-    return core_count;
+    return threads;
 }
 #endif // QPP_OPENMP
 

benchmarks/src/measure_seq_psi_bench.cpp

@@ -14,10 +14,10 @@ qpp::idx nq = 10; // default number of qubits if none provided at runtime
 int main(int argc, char* argv[]) {
     Catch::Session session;
 
-    auto cli =
-        session.cli() | Catch::Clara::Opt(nq, "nq")["--nq"]("Number of qubits");
+    auto cli = session.cli() |
+               Catch::Clara::Opt(nq, "qubits")["--nq"]("Number of qubits");
 #ifdef QPP_OPENMP
-    cli |= Catch::Clara::Opt(cli_core_count, "core_count")["--core_count"](
+    cli |= Catch::Clara::Opt(cli_threads, "threads")["--threads"](
         "Number of OpenMP threads/cores (ignored if the environment variable "
         "OMP_NUM_THREADS is set)");
 #endif // QPP_OPENMP

benchmarks/src/measure_seq_rho_bench.cpp

@@ -14,10 +14,10 @@ qpp::idx nq = 10; // default number of qubits if none provided at runtime
 int main(int argc, char* argv[]) {
     Catch::Session session;
 
-    auto cli =
-        session.cli() | Catch::Clara::Opt(nq, "nq")["--nq"]("Number of qubits");
+    auto cli = session.cli() |
+               Catch::Clara::Opt(nq, "qubits")["--nq"]("Number of qubits");
 #ifdef QPP_OPENMP
-    cli |= Catch::Clara::Opt(cli_core_count, "core_count")["--core_count"](
+    cli |= Catch::Clara::Opt(cli_threads, "threads")["--threads"](
         "Number of OpenMP threads/cores (ignored if the environment variable "
         "OMP_NUM_THREADS is set)");
 #endif // QPP_OPENMP

benchmarks/src/ptrace_psi_bench.cpp

@@ -7,17 +7,19 @@
 
 #include "openmp_utils.hpp"
 
+#include "qpp/internal/kernels/qubit/ptrace.hpp"
+
 namespace {
 qpp::idx nq = 10; // default number of qubits if none provided at runtime
 } // namespace
 
 int main(int argc, char* argv[]) {
     Catch::Session session;
 
-    auto cli =
-        session.cli() | Catch::Clara::Opt(nq, "nq")["--nq"]("Number of qubits");
+    auto cli = session.cli() |
+               Catch::Clara::Opt(nq, "qubits")["--nq"]("Number of qubits");
 #ifdef QPP_OPENMP
-    cli |= Catch::Clara::Opt(cli_core_count, "core_count")["--core_count"](
+    cli |= Catch::Clara::Opt(cli_threads, "threads")["--threads"](
         "Number of OpenMP threads/cores (ignored if the environment variable "
         "OMP_NUM_THREADS is set)");
 #endif // QPP_OPENMP
@@ -63,11 +65,23 @@ TEST_CASE("qpp::ptrace() state vector benchmark",
         // calculation away.
         return qpp::ptrace_new1(psi, subsys, dims);
     };
+    // Benchmarked portion (executed repeatedly)
+    BENCHMARK("Partial trace new2 (psi) nq=" + std::to_string(nq)) {
+        // CRITICAL: Return the result so the compiler doesn't optimize the
+        // calculation away.
+        return qpp::ptrace_new2(psi, subsys, dims);
+    };
+    // Benchmarked portion (executed repeatedly)
+    BENCHMARK("Partial trace new3 (psi) nq=" + std::to_string(nq)) {
+        // CRITICAL: Return the result so the compiler doesn't optimize the
+        // calculation away.
+        return qpp::ptrace_new3(psi, subsys, dims);
+    };
 
     // Benchmarked portion (executed repeatedly)
     BENCHMARK("Partial trace qubits (psi) nq=" + std::to_string(nq)) {
         // CRITICAL: Return the result so the compiler doesn't optimize the
         // calculation away.
-        return qpp::internal::ptrace_psi_kq(psi, subsys, nq);
+        return qpp::internal::kernels::qubit::ptrace_psi_kq(psi, subsys, nq);
     };
 }

benchmarks/src/ptrace_rho_bench.cpp

@@ -7,17 +7,19 @@
 
 #include "openmp_utils.hpp"
 
+#include "qpp/internal/kernels/qubit/ptrace.hpp"
+
 namespace {
 qpp::idx nq = 10; // default number of qubits if none provided at runtime
 } // namespace
 
 int main(int argc, char* argv[]) {
     Catch::Session session;
 
-    auto cli =
-        session.cli() | Catch::Clara::Opt(nq, "nq")["--nq"]("Number of qubits");
+    auto cli = session.cli() |
+               Catch::Clara::Opt(nq, "qubits")["--nq"]("Number of qubits");
 #ifdef QPP_OPENMP
-    cli |= Catch::Clara::Opt(cli_core_count, "core_count")["--core_count"](
+    cli |= Catch::Clara::Opt(cli_threads, "threads")["--threads"](
         "Number of OpenMP threads/cores (ignored if the environment variable "
         "OMP_NUM_THREADS is set)");
 #endif // QPP_OPENMP
@@ -63,11 +65,23 @@ TEST_CASE("qpp::ptrace() density matrix benchmark",
         // calculation away.
         return qpp::ptrace_new1(rho, subsys, dims);
     };
+    // Benchmarked portion (executed repeatedly)
+    BENCHMARK("Partial trace new2 (rho) nq=" + std::to_string(nq)) {
+        // CRITICAL: Return the result so the compiler doesn't optimize the
+        // calculation away.
+        return qpp::ptrace_new2(rho, subsys, dims);
+    };
+    // Benchmarked portion (executed repeatedly)
+    BENCHMARK("Partial trace new3 (rho) nq=" + std::to_string(nq)) {
+        // CRITICAL: Return the result so the compiler doesn't optimize the
+        // calculation away.
+        return qpp::ptrace_new3(rho, subsys, dims);
+    };
 
     // Benchmarked portion (executed repeatedly)
     BENCHMARK("Partial trace qubits (rho) nq=" + std::to_string(nq)) {
         // CRITICAL: Return the result so the compiler doesn't optimize the
         // calculation away.
-        return qpp::internal::ptrace_rho_kq(rho, subsys, nq);
+        return qpp::internal::kernels::qubit::ptrace_rho_kq(rho, subsys, nq);
     };
 }

benchmarks/src/ptranspose_psi_bench.cpp

@@ -7,17 +7,19 @@
 
 #include "openmp_utils.hpp"
 
+#include "qpp/internal/kernels/qubit/ptranspose.hpp"
+
 namespace {
 qpp::idx nq = 10; // default number of qubits if none provided at runtime
 } // namespace
 
 int main(int argc, char* argv[]) {
     Catch::Session session;
 
-    auto cli =
-        session.cli() | Catch::Clara::Opt(nq, "nq")["--nq"]("Number of qubits");
+    auto cli = session.cli() |
+               Catch::Clara::Opt(nq, "qubits")["--nq"]("Number of qubits");
 #ifdef QPP_OPENMP
-    cli |= Catch::Clara::Opt(cli_core_count, "core_count")["--core_count"](
+    cli |= Catch::Clara::Opt(cli_threads, "threads")["--threads"](
         "Number of OpenMP threads/cores (ignored if the environment variable "
         "OMP_NUM_THREADS is set)");
 #endif // QPP_OPENMP
@@ -69,6 +71,7 @@ TEST_CASE("qpp::ptranspose() state vector benchmark",
     BENCHMARK("Partial transpose qubits (psi) nq=" + std::to_string(nq)) {
         // CRITICAL: Return the result so the compiler doesn't optimize the
         // calculation away.
-        return qpp::internal::ptranspose_psi_kq(psi, subsys, nq);
+        return qpp::internal::kernels::qubit::ptranspose_psi_kq(psi, subsys,
+                                                                nq);
     };
 }

benchmarks/src/ptranspose_rho_bench.cpp

@@ -7,17 +7,19 @@
 
 #include "openmp_utils.hpp"
 
+#include "qpp/internal/kernels/qubit/ptranspose.hpp"
+
 namespace {
 qpp::idx nq = 10; // default number of qubits if none provided at runtime
 } // namespace
 
 int main(int argc, char* argv[]) {
     Catch::Session session;
 
-    auto cli =
-        session.cli() | Catch::Clara::Opt(nq, "nq")["--nq"]("Number of qubits");
+    auto cli = session.cli() |
+               Catch::Clara::Opt(nq, "qubits")["--nq"]("Number of qubits");
 #ifdef QPP_OPENMP
-    cli |= Catch::Clara::Opt(cli_core_count, "core_count")["--core_count"](
+    cli |= Catch::Clara::Opt(cli_threads, "threads")["--threads"](
         "Number of OpenMP threads/cores (ignored if the environment variable "
         "OMP_NUM_THREADS is set)");
 #endif // QPP_OPENMP
@@ -56,11 +58,19 @@ TEST_CASE("qpp::ptranspose() density matrix benchmark",
         // calculation away.
         return qpp::ptranspose(rho, subsys);
     };
+    std::vector<qpp::idx> dims(nq, 2);
+    // Benchmarked portion (executed repeatedly)
+    BENCHMARK("Partial transpose new (rho) nq=" + std::to_string(nq)) {
+        // CRITICAL: Return the result so the compiler doesn't optimize the
+        // calculation away.
+        return qpp::ptranspose_new(rho, subsys, dims);
+    };
 
     // Benchmarked portion (executed repeatedly)
     BENCHMARK("Partial transpose qubits (rho) nq=" + std::to_string(nq)) {
         // CRITICAL: Return the result so the compiler doesn't optimize the
         // calculation away.
-        return qpp::internal::ptranspose_rho_kq(rho, subsys, nq);
+        return qpp::internal::kernels::qubit::ptranspose_rho_kq(rho, subsys,
+                                                                nq);
     };
 }