Fix pr 90 (#93)

* Update RL (#56) * fix:remove hashvalue when popping graphs in push_back; refactor shrink * fix: wrong error raising in graph buffer * fix: do not shrink buffer according to sys mem usage * fix: do not pop the graph to push * fix: do not pop the graph to push * feat: change mem control policy * feat: change mem control policy * Added two Rigetti ecc sets for experiment * refactor: add some scripts * fix scripts * Added rigetti rotation merging circuits * feat: add configs and scripts for rigetti set * feat: add scripts for cx on t_tdg * fix: gate set for t_tdg * Added two gate types in cython * fix: gate set for rigetti set * feat: add scripts for tdg_rm * fix:d script for tdg_rm * fix:change name for tdg_rm * fix: configs for tuning tdg * feat: add scripts for darawing full seq * feat: add scripts for darawing full seq * fix: scripts * Added rigetti 3, 6 ecc set * fix: nam scripts and actor.py * Added some rigetti circuits, a qasm converter, and modified the graph.from_qasm api * Remove debug outputs * minor changes * feat: add test_rigetti * fix: test_rigetti * feat: add test tdg rm cx * Fixed bug in graph::from_qasm * Fixed some circuits * Delete some rigetti benchmarks * Added a rigetti 3, 3, 5 ecc set * feat: run new rigetti circs * update ipynb, including vis_seq * fix: wrong circ dir in test_t_tdg_rm_cx.py * Modified tasograph.cpp * Added 4 gates ry1, ry3, rxx1, rxx3 * Add a ionq ecc set * Added ionq circs, normalized * feat: add BFS scripts for ionq * feat: add RL config for IonQ * feat: add RL config for IonQ * feat: add RL config for IonQ * Fixed a bug in Graph::from_qasm_file * fix: increase num_gate_types * New version of toffoli flip * Added 6 nam benchmark circuits * fix: set num_gate_types for Nam and TDG * add ipynb to parse log * feat: output best graph to a specified folder; close #54 Co-authored-by: zikun-li <[email protected]> * Add CP (controlled phase) gate * Add support of comment line and empty line to Graph::from_qasm_file * [qasm] Also ignore barrier and measure * [gate] Add is_sparse for sparse matrices * [Refactor] Rename "DAG" to "CircuitSeq" (#61) * [Refactor] Rename "DAG" to "CircuitSeq" * fix compilation error * [QASM] Support multiple quantum registers (#63) * [qasm] Add CircuitSeq::from_qasm_file() (#64) * [qasm] Add one kind of parameter format support in QASM parser (#65) * [qasm] Store input parameters in the context * [qasm] Add support of U gate * [gate] Add some support functions for controlled gates (#67) * Add HiGHS solver as a submodule (#70) * fix: update isort in pre-commit to resolve its issue with newer python * fix: wrong branch name * [gate] Add CU1 gate (#75) * Update Simulation (#77) * Add MQT benchmarks and code skeleton for test_simulation * Use Graph::from_qasm_file * Add support of comment line and empty line to Graph::from_qasm_file * [qasm] Also ignore barrier and measure * Use dj benchmark * Add MQT benchmarks * [gate] Add is_sparse for sparse matrices (cherry picked from commit 8a29479) * Heuristics in SnuQS * Experiment for 40-42q, 12-33k, 9 circuits * Experiment for 40-42q, 12-33k, 9 circuits * [refactor] Rename "DAG" to "CircuitSeq" in test_simulation.cpp * Add ILP * ILP for the full benchmark * Fix example k value * Fix dependencies * Add some outputs * Fix redundant dependencies * Print out schedules * Use HiGHS solver * Add tiebreaker for heuristics and output the number of iterations (number of shuffles plus one) * Use CircuitSeq::from_qasm_file in test_simulation * Add 3 more benchmark circuits * Multiple qregs for ILP * restore k * Remove initial default configuration in heuristics in test_simulation * [qasm] Add support of U gate * Update experiments for test_simulation.cpp and ilp.py * [Simulation] Add schedule and a function to compute the number of down sets * Test computing the number of down sets * debug * [Simulation] Dynamic programming to compute the kernel schedule * [simulation] Add an optimization to the DP * [simulation] Fix a bug in the DP * [simulation] Add absorbing qubits to the DP * Add 29-qubit circuits and fix absorbing qubits to be sets of sets * Add 29-qubit circuits * Not record the result kernels when not needed * Integrate sim (#68) * [simulator] initial attempt to integrate into qartz * [simulator] fix compilation error, cmake version at least 3.18 * [simulator] set using simulator to be optional * [simulator] set using simulator optional * [simulator] do matrix shuffle to guarantee the matrix is for increasing qubit order * [simulator] matrix shuffle * [simulator] virtual/physical qubit mapping when applying gates * [simulator] fix some bugs, tested on one device * [simulator] fix some bugs, tested on 2-devices * pre-commit * [Simulation] Use pybind11 to integrate ILP into the simulator (#69) * [Simulation] Use pybind11 to integrate ILP into the simulator * Add pybind11 in requirements.txt * Add pybind11 to env.yml * Fix linker error on Linux * Make pybind.cpp not dependent on the working directory * Restore circuit location in test_simulation.cpp * Fix the environment on Windows * [simulation] Add the s array to ILP and print the running time in test_simulation (#71) * [simulation] Fix a bug about ILP and DP (#72) * [bug] Amend #72 * Fix Windows environment (#74) * Fix Windows environment * Code format * Code format * Code format * [gate] Add CU1 gate * implement a distributed simulator using legion * Multi-node Simulator (mpi-based) (#76) * [simulator] init multi-node impl * [simulator][multi-node] fix some compilation error * [simulator][multi-node] fix some error * [simulator][multi-node] tested on 2 p3.2xlarge node * [simulator][multi-node] added profiling, ILP some error * [simulator][multi-node] fix some resource handle error * [simulator][multi-node] for parallelcluster * [simulator][multi-node] add more circuits * [simulator][multi-node] do not block creating event * [simulator][multi-node]fix sume bugs * add some circuits * pre-commit * comment back quartz tests --------- Co-authored-by: Ubuntu <[email protected]> --------- Co-authored-by: shiyi.c_98 <[email protected]> Co-authored-by: Zhihao Jia <[email protected]> Co-authored-by: Ubuntu <[email protected]> * Update Simulation (#81) * Add MQT benchmarks and code skeleton for test_simulation * Use Graph::from_qasm_file * Add support of comment line and empty line to Graph::from_qasm_file * [qasm] Also ignore barrier and measure * Use dj benchmark * Add MQT benchmarks * [gate] Add is_sparse for sparse matrices (cherry picked from commit 8a29479) * Heuristics in SnuQS * Experiment for 40-42q, 12-33k, 9 circuits * Experiment for 40-42q, 12-33k, 9 circuits * [refactor] Rename "DAG" to "CircuitSeq" in test_simulation.cpp * Add ILP * ILP for the full benchmark * Fix example k value * Fix dependencies * Add some outputs * Fix redundant dependencies * Print out schedules * Use HiGHS solver * Add tiebreaker for heuristics and output the number of iterations (number of shuffles plus one) * Use CircuitSeq::from_qasm_file in test_simulation * Add 3 more benchmark circuits * Multiple qregs for ILP * restore k * Remove initial default configuration in heuristics in test_simulation * [qasm] Add support of U gate * Update experiments for test_simulation.cpp and ilp.py * [Simulation] Add schedule and a function to compute the number of down sets * Test computing the number of down sets * debug * [Simulation] Dynamic programming to compute the kernel schedule * [simulation] Add an optimization to the DP * [simulation] Fix a bug in the DP * [simulation] Add absorbing qubits to the DP * Add 29-qubit circuits and fix absorbing qubits to be sets of sets * Add 29-qubit circuits * Not record the result kernels when not needed * Integrate sim (#68) * [simulator] initial attempt to integrate into qartz * [simulator] fix compilation error, cmake version at least 3.18 * [simulator] set using simulator to be optional * [simulator] set using simulator optional * [simulator] do matrix shuffle to guarantee the matrix is for increasing qubit order * [simulator] matrix shuffle * [simulator] virtual/physical qubit mapping when applying gates * [simulator] fix some bugs, tested on one device * [simulator] fix some bugs, tested on 2-devices * pre-commit * [Simulation] Use pybind11 to integrate ILP into the simulator (#69) * [Simulation] Use pybind11 to integrate ILP into the simulator * Add pybind11 in requirements.txt * Add pybind11 to env.yml * Fix linker error on Linux * Make pybind.cpp not dependent on the working directory * Restore circuit location in test_simulation.cpp * Fix the environment on Windows * [simulation] Add the s array to ILP and print the running time in test_simulation (#71) * [simulation] Fix a bug about ILP and DP (#72) * [bug] Amend #72 * Fix Windows environment (#74) * Fix Windows environment * Code format * Code format * Code format * [gate] Add CU1 gate * implement a distributed simulator using legion * Multi-node Simulator (mpi-based) (#76) * [simulator] init multi-node impl * [simulator][multi-node] fix some compilation error * [simulator][multi-node] fix some error * [simulator][multi-node] tested on 2 p3.2xlarge node * [simulator][multi-node] added profiling, ILP some error * [simulator][multi-node] fix some resource handle error * [simulator][multi-node] for parallelcluster * [simulator][multi-node] add more circuits * [simulator][multi-node] do not block creating event * [simulator][multi-node]fix sume bugs * add some circuits * pre-commit * comment back quartz tests --------- Co-authored-by: Ubuntu <[email protected]> * [Simulation] Add an option to not consider single-qubit gates in DP (#78) * [Simulation] DP for shared-memory gates (#80) * Add Kernel and KernelCost classes * Schedule::compute_end_schedule() now considers shared-memory kernels * Include kernel type in Kernel::to_string() * Fix missing single-qubit gates * DP for shared-memory kernels * Remove TODO about controlled gates * Code format --------- Co-authored-by: shiyi.c_98 <[email protected]> Co-authored-by: Zhihao Jia <[email protected]> Co-authored-by: Ubuntu <[email protected]> * [simulation] Fix CX in execution schedules when the target is global (#82) * [simulation] Fix CX being put into execution schedule when the target is global * refine criteria * edit comment * [bug] Fix clone_from() copying hash value incorrectly (#84) * [simulation] Restrict the Swap gate as local-only (#85) * [simulation] Consider cacheline qubits in DP for shared-memory kernels (#83) * Change the type of Schedule::local_qubit_ from vector<bool> to vector<int> * [simulation] Consider cacheline qubits in DP for shared-memory kernels * Apply the new algorithm and fix compilation error in debug mode * Record global qubits * Ecc pruning dev (#86) * Add a pass to eliminate ecc classes that are equivalent under qubit permutation * Add a pass to eliminate ecc classes that are equivalent under qubit permutation * Modified kNumOptimizationsToPerform * fix u3 matrix implementation (#87) * Fix bug * Eliminate u3 gate if parameters are all 0 * fix: pre-commit --------- Co-authored-by: zikun-li <[email protected]> Co-authored-by: xumingkuan <[email protected]> Co-authored-by: shiyi.c_98 <[email protected]> Co-authored-by: Zhihao Jia <[email protected]> Co-authored-by: Ubuntu <[email protected]> Co-authored-by: zikun-li <[email protected]> Co-authored-by: Shaohan Hu <[email protected]>
quantum-compiler · Mar 14, 2023 · 48926e3 · 48926e3
1 parent 3a36874
commit 48926e3
Show file tree

Hide file tree

Showing 125 changed files with 63,179 additions and 34 deletions.
diff --git a/.github/workflows/pre-commit.yml b/.github/workflows/pre-commit.yml
@@ -3,7 +3,7 @@ name: pre-commit
 on:
   pull_request:
   push:
-    branches: [main]
+    branches: [master]
 
 jobs:
   pre-commit:

diff --git a/.gitmodules b/.gitmodules
@@ -0,0 +1,6 @@
+[submodule "external/HiGHS"]
+	path = external/HiGHS
+	url = https://github.com/ERGO-Code/HiGHS.git
+[submodule "external/legion"]
+	path = external/legion
+	url = https://gitlab.com/StanfordLegion/legion.git
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -7,7 +7,7 @@ repos:
         'types_or': [c++, c, cuda]
 
 -   repo: https://github.com/pycqa/isort
-    rev: 5.10.1
+    rev: 5.12.0
     hooks:
     -   id: isort
         name: isort (python)

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -3,6 +3,33 @@ project(Quartz)
 
 SET(CMAKE_BUILD_TYPE "Release")
 
+# python
+if (MSVC)
+    # Could not find the "Development" component on Windows. Using the following commands
+    # to get Python_INCLUDE_DIRS and Python_LIBRARY_DIRS.
+    find_package(Python COMPONENTS Interpreter)
+    execute_process(COMMAND ${Python_EXECUTABLE} -c "import sysconfig; print(sysconfig.get_path('include'))"
+            OUTPUT_VARIABLE Python_INCLUDE_DIRS OUTPUT_STRIP_TRAILING_WHITESPACE)
+    execute_process(COMMAND ${Python_EXECUTABLE} -c "import distutils.command.build_ext;
+d = distutils.core.Distribution();
+b = distutils.command.build_ext.build_ext(d);
+b.finalize_options();
+print(*b.library_dirs, sep=';')"
+            OUTPUT_VARIABLE Python_LIBRARY_DIRS OUTPUT_STRIP_TRAILING_WHITESPACE)
+else()
+    find_package(Python COMPONENTS Interpreter Development)
+endif()
+message(STATUS "Python include directories: ${Python_INCLUDE_DIRS}")
+include_directories(${Python_INCLUDE_DIRS})
+message(STATUS "Python library directories: ${Python_LIBRARY_DIRS}")
+link_directories(${Python_LIBRARY_DIRS})
+
+# pybind11
+execute_process(COMMAND ${Python_EXECUTABLE} -m pybind11 --cmake
+        OUTPUT_VARIABLE pybind11_DIR OUTPUT_STRIP_TRAILING_WHITESPACE)
+find_package(pybind11 CONFIG REQUIRED)
+include_directories(${pybind11_INCLUDE_DIR})
+
 if (EXISTS ${CMAKE_CURRENT_BINARY_DIR}/config.cmake)
     include (${CMAKE_CURRENT_BINARY_DIR}/config.cmake)
 else()
@@ -32,7 +59,7 @@ include_directories("src/quartz/")
 
 #initial variables
 set(QUARTZ_LIBS "")
-set(QUARTZ_LINK_LIBS ${CMAKE_DL_LIBS})
+set(QUARTZ_LINK_LIBS ${CMAKE_DL_LIBS} pybind11::embed)
 set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
 file(GLOB_RECURSE QUARTZ_SRCS
@@ -111,3 +138,7 @@ install (DIRECTORY ${PROJECT_SOURCE_DIR}/src/quartz/
 
 add_subdirectory(src/test)
 add_subdirectory(src/benchmark)
+
+if(QUARTZ_USE_SIMULATOR)
+  add_subdirectory(src/sim)
+endif()
diff --git a/CODE_STRUCTURE.md b/CODE_STRUCTURE.md
@@ -21,44 +21,44 @@ In `src/quartz/context/rule_parser.h`
 * `class RuleParser`: define the rules to write 3-qubit gates in each of the gate sets
 
 In `src/quartz/parser/qasm_parser.h`
-* `class QASMParser`: parse an input QASM file to Quartz's DAG representation
+* `class QASMParser`: parse an input QASM file to Quartz's CircuitSeq representation
 
-In `src/dag/dag.h`
-* `class DAG`: a circuit represented in a DAG with all gates stored in an `std::vector` (which is a sequence representation)
+In `src/quartz/circuitseq/circuitseq.h`
+* `class CircuitSeq`: a circuit sequence with all gates stored in an `std::vector`
 
-In `src/dag/dagnode.h`
-* `class DAGNode`: a node in DAG corresponds to a wire in the circuit
+In `src/quartz/circuitseq/circuitwire.h`
+* `class CircuitWire`: a wire in the circuit sequence
 
-In `src/dag/daghyperedge.h`
-* `class DAGHyperEdge`: a hyperedge in DAG corresponds to a gate (or parameter expression) in the circuit
+In `src/quartz/circuitseq/circuitgate.h`
+* `class CircuitGate`: a gate (or parameter expression) in the circuit sequence
 
-In `src/dataset/dataset.h`
+In `src/quartz/dataset/dataset.h`
 * `class Dataset`: a collection of circuits grouped by fingerprints
 
-In `src/dataset/equivalence_set.h`
+In `src/quartz/dataset/equivalence_set.h`
 * `class EquivalenceClass`: an ECC
 * `class EquivalenceSet`: an ECC set
 
-In `src/generator/generator.h`
+In `src/quartz/generator/generator.h`
 * `class Generator`: the circuit generator
 * `Generator::generate`: generate circuits for an unverified ECC set (then use `src/python/verify_equivalences.py` to get the ECC set)
 
-In `src/math/matrix.h`
+In `src/quartz/math/matrix.h`
 * `class Matrix`: a complex square matrix
 
-In `src/math/vector.h`
+In `src/quartz/math/vector.h`
 * `class Vector`: a complex vector
 
-In `src/verifier/verifier.h`
+In `src/quartz/verifier/verifier.h`
 * `Verifier::redundant`: check if the circuit generated is redundant, i.e., having some slices not in the representative set
 
-In `src/tasograph/tasograph.h`
+In `src/quartz/tasograph/tasograph.h`
 * `class Graph`: the circuit to be optimized
 * `Graph::optimize`: use the search algorithm to optimize the circuit
 * `Graph::context_shift`: shift the context of the circuit, e.g., changing the gate set
 
-In `src/tasograph/substitution.h`
+In `src/quartz/tasograph/substitution.h`
 * `class GraphXfer`: a circuit transformation
 
-In `src/test/gen_ecc_set.cpp`
+In `src/quartz/test/gen_ecc_set.cpp`
 * `gen_ecc_set`: a function to generate ECC sets with given gate set and hyperparameters
diff --git a/INSTALL.md b/INSTALL.md
@@ -59,3 +59,17 @@ export LD_LIBRARY_PATH=/your/path/:$LD_LIBRARY_PATH
 ```
 
 to `~/.bashrc`.
+
+### Dependencies for the simulator
+
+For the simulator, you need to install the HiGHS solver.
+
+```shell
+cd external/HiGHS
+mkdir build
+cd build
+cmake ..
+make
+```
+
+Then add `/your/path/to/quartz/external/HiGHS/build/bin` to `PATH`.
diff --git a/README.md b/README.md
@@ -70,26 +70,26 @@ After that, you need a `QASMParser` object to parse the input `qasm` file. You c
 QASMParser qasm_parser(&src_ctx);
 ```
 
-Now you can use the `QASMParser` object to load the circuit from the `qasm` file to a `DAG` object, as below:
+Now you can use the `QASMParser` object to load the circuit from the `qasm` file to a `CircuitSeq` object, as below:
 
 ``` cpp
-DAG *dag = nullptr;
-if (!qasm_parser.load_qasm(input_fn, dag)) {
+CircuitSeq *seq = nullptr;
+if (!qasm_parser.load_qasm(input_fn, seq)) {
     std::cout << "Parser failed" << std::endl;
 }
 ```
 
-After you have the circuit loaded into the `DAG` object, you can construct a `Graph` object from it. The `Graph` object is the final circuit representation used in our optimizer. You can construct it as below:
+After you have the circuit loaded into the `CircuitSeq` object, you can construct a `Graph` object from it. The `Graph` object is the final circuit representation used in our optimizer. You can construct it as below:
 
 ``` cpp
-Graph graph(&src_ctx, dag);
+Graph graph(&src_ctx, seq);
 ```
 
 #### Context shift
 
 If the input gate set is different from your target gate set, you should consider using the `context_shift` APIs to shift the context constructed with the gate sets to a context constructed with the target gate set.
 
-To shift the context, you should create three `Contxt` objects, one for input, one for target, and one for their union as below:
+To shift the context, you should create three `Context` objects, one for input, one for target, and one for their union as below:
 
 ``` cpp
 Context src_ctx({GateType::h, GateType::ccz, GateType::x, GateType::cx,

diff --git a/circuit/MQTBench_28q/ghz_indep_qiskit_28.qasm b/circuit/MQTBench_28q/ghz_indep_qiskit_28.qasm
@@ -0,0 +1,66 @@
+// Benchmark was created by MQT Bench on 2022-12-15
+// For more information about MQT Bench, please visit https://www.cda.cit.tum.de/mqtbench/
+// MQT Bench version: 0.2.2
+// Qiskit version: {'qiskit-terra': '0.22.3', 'qiskit-aer': '0.11.1', 'qiskit-ignis': '0.7.0', 'qiskit-ibmq-provider': '0.19.2', 'qiskit': '0.39.3', 'qiskit-nature': '0.5.1', 'qiskit-finance': '0.3.4', 'qiskit-optimization': '0.4.0', 'qiskit-machine-learning': '0.5.0'}
+
+OPENQASM 2.0;
+include "qelib1.inc";
+qreg q[28];
+creg meas[28];
+h q[27];
+cx q[27],q[26];
+cx q[26],q[25];
+cx q[25],q[24];
+cx q[24],q[23];
+cx q[23],q[22];
+cx q[22],q[21];
+cx q[21],q[20];
+cx q[20],q[19];
+cx q[19],q[18];
+cx q[18],q[17];
+cx q[17],q[16];
+cx q[16],q[15];
+cx q[15],q[14];
+cx q[14],q[13];
+cx q[13],q[12];
+cx q[12],q[11];
+cx q[11],q[10];
+cx q[10],q[9];
+cx q[9],q[8];
+cx q[8],q[7];
+cx q[7],q[6];
+cx q[6],q[5];
+cx q[5],q[4];
+cx q[4],q[3];
+cx q[3],q[2];
+cx q[2],q[1];
+cx q[1],q[0];
+barrier q[0],q[1],q[2],q[3],q[4],q[5],q[6],q[7],q[8],q[9],q[10],q[11],q[12],q[13],q[14],q[15],q[16],q[17],q[18],q[19],q[20],q[21],q[22],q[23],q[24],q[25],q[26],q[27];
+measure q[0] -> meas[0];
+measure q[1] -> meas[1];
+measure q[2] -> meas[2];
+measure q[3] -> meas[3];
+measure q[4] -> meas[4];
+measure q[5] -> meas[5];
+measure q[6] -> meas[6];
+measure q[7] -> meas[7];
+measure q[8] -> meas[8];
+measure q[9] -> meas[9];
+measure q[10] -> meas[10];
+measure q[11] -> meas[11];
+measure q[12] -> meas[12];
+measure q[13] -> meas[13];
+measure q[14] -> meas[14];
+measure q[15] -> meas[15];
+measure q[16] -> meas[16];
+measure q[17] -> meas[17];
+measure q[18] -> meas[18];
+measure q[19] -> meas[19];
+measure q[20] -> meas[20];
+measure q[21] -> meas[21];
+measure q[22] -> meas[22];
+measure q[23] -> meas[23];
+measure q[24] -> meas[24];
+measure q[25] -> meas[25];
+measure q[26] -> meas[26];
+measure q[27] -> meas[27];