Merge branch 'release-v24.08' into 'main'

Release v24.08

See merge request cuda-hpc-libraries/cuquantum-sdk/cuquantum-public!28

Author: mtjrider

python/README.md

@@ -5,7 +5,7 @@
 Please visit the [NVIDIA cuQuantum Python documentation](https://docs.nvidia.com/cuda/cuquantum/latest/python).
 
 For instructions on installing *cuQuantum Python*, refer to our 
-[getting started section](../getting_started/getting_started.rst)
+[getting started section](../getting-started/index.rst)
 
 ## Building and installing cuQuantum Python from source
 
@@ -15,8 +15,8 @@ The build-time dependencies of the cuQuantum Python package include:
 
 * CUDA Toolkit 11.x or 12.x
 * cuStateVec 1.4.0+
-* cuTensorNet 2.4.0+
-* Python 3.9+
+* cuTensorNet 2.5.0+
+* Python 3.10+
 * Cython >=0.29.22,<3
 * pip 21.3.1+
 * [packaging](https://packaging.pypa.io/en/latest/)
@@ -57,8 +57,8 @@ Runtime dependencies of the cuQuantum Python package include:
 * Driver: Linux (450.80.02+ for CUDA 11, 525.60.13+ for CUDA 12)
 * CUDA Toolkit 11.x or 12.x
 * cuStateVec 1.4.0+
-* cuTensorNet 2.4.0+
-* Python 3.9+
+* cuTensorNet 2.5.0+
+* Python 3.10+
 * NumPy v1.21+
 * CuPy v13.0.0+ (see [installation guide](https://docs.cupy.dev/en/stable/install.html))
 * PyTorch v1.10+ (optional, see [installation guide](https://pytorch.org/get-started/locally/))
@@ -96,4 +96,4 @@ variable `CUDA_PATH` is not set.
 
 ## Citing cuQuantum
 
-Pleae click this Zenodo badge to see the citation format: [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.6385574.svg)](https://doi.org/10.5281/zenodo.6385574)
+Please click this Zenodo badge to see the citation format: [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.6385574.svg)](https://doi.org/10.5281/zenodo.6385574)

python/builder/pep517.py

@@ -31,7 +31,7 @@ def get_requires_for_build_wheel(config_settings=None):
     # 22.3 by setuptools, so we must follow the same practice in the constraints;
     # also, we don't need the patch number here
     cuqnt_require = [f'custatevec-cu{utils.cuda_major_ver}~=1.6',   # ">=1.6.0,<2"
-                     f'cutensornet-cu{utils.cuda_major_ver}~=2.4',  # ">=2.4.0,<3"
+                     f'cutensornet-cu{utils.cuda_major_ver}~=2.5',   # ">=2.5.0,<3"
                     ]
 
     return _build_meta.get_requires_for_build_wheel(config_settings) + cuqnt_require

python/cuquantum/__init__.py

@@ -5,8 +5,9 @@
 from cuquantum import custatevec
 from cuquantum import cutensornet
 from cuquantum.cutensornet import (
-    contract, contract_path, einsum, einsum_path, tensor, tensor_qualifiers_dtype, Network, BaseCUDAMemoryManager, MemoryPointer,
-    NetworkOptions, OptimizerInfo, OptimizerOptions, PathFinderOptions, ReconfigOptions, SlicerOptions, CircuitToEinsum)
+    contract, contract_path, einsum, einsum_path, tensor, tensor_qualifiers_dtype, BaseCUDAMemoryManager, CircuitToEinsum, MemoryPointer, 
+    Network, NetworkOptions, OptimizerInfo, OptimizerOptions, PathFinderOptions, 
+    ReconfigOptions, SlicerOptions)
 from cuquantum._utils import ComputeType, cudaDataType, libraryPropertyType
 from cuquantum._version import __version__
 

python/cuquantum/_utils.pyx

@@ -111,6 +111,7 @@ class ComputeType(IntEnum):
     COMPUTE_32I     = 1 << 9
     COMPUTE_16BF    = 1 << 10
     COMPUTE_TF32    = 1 << 12
+    COMPUTE_3XTF32  = 1 << 13
 
 
 # TODO: use those exposed by CUDA Python instead, but before removing these

python/cuquantum/_version.py

@@ -5,4 +5,4 @@
 # Note: cuQuantum Python follows the cuQuantum SDK version, which is now
 # switched to YY.MM and is different from individual libraries' (semantic)
 # versioning scheme.
-__version__ = '24.03.0'
+__version__ = '24.08.0'

python/cuquantum/custatevec/_internal/custatevec.pxd

@@ -79,8 +79,8 @@ cdef custatevecStatus_t _custatevecAbs2SumArrayBatched(custatevecHandle_t handle
 cdef custatevecStatus_t _custatevecCollapseByBitStringBatchedGetWorkspaceSize(custatevecHandle_t handle, const uint32_t nSVs, const custatevecIndex_t* bitStrings, const double* norms, size_t* extraWorkspaceSizeInBytes) except* nogil
 cdef custatevecStatus_t _custatevecCollapseByBitStringBatched(custatevecHandle_t handle, void* batchedSv, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, const custatevecIndex_t svStride, const custatevecIndex_t* bitStrings, const int32_t* bitOrdering, const uint32_t bitStringLen, const double* norms, void* extraWorkspace, size_t extraWorkspaceSizeInBytes) except* nogil
 cdef custatevecStatus_t _custatevecMeasureBatched(custatevecHandle_t handle, void* batchedSv, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, const custatevecIndex_t svStride, custatevecIndex_t* bitStrings, const int32_t* bitOrdering, const uint32_t bitStringLen, const double* randnums, custatevecCollapseOp_t collapse) except* nogil
-cdef custatevecStatus_t _custatevecComputeExpectationBatchedGetWorkspaceSize(custatevecHandle_t handle, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, const custatevecIndex_t svStride, const void* matrices, cudaDataType_t matrixDataType, custatevecMatrixLayout_t layout, const uint32_t nMatrices, const uint32_t nBasisBits, custatevecComputeType_t computeType, size_t* extraWorkspaceSizeInBytes) except* nogil
-cdef custatevecStatus_t _custatevecComputeExpectationBatched(custatevecHandle_t handle, const void* batchedSv, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, custatevecIndex_t svStride, double2* expectationValues, const void* matrices, cudaDataType_t matrixDataType, custatevecMatrixLayout_t layout, const uint32_t nMatrices, const int32_t* basisBits, const uint32_t nBasisBits, custatevecComputeType_t computeType, void* extraWorkspace, size_t extraWorkspaceSizeInBytes) except* nogil
 cdef custatevecStatus_t _custatevecSubSVMigratorCreate(custatevecHandle_t handle, custatevecSubSVMigratorDescriptor_t* migrator, void* deviceSlots, cudaDataType_t svDataType, int nDeviceSlots, int nLocalIndexBits) except* nogil
 cdef custatevecStatus_t _custatevecSubSVMigratorDestroy(custatevecHandle_t handle, custatevecSubSVMigratorDescriptor_t migrator) except* nogil
 cdef custatevecStatus_t _custatevecSubSVMigratorMigrate(custatevecHandle_t handle, custatevecSubSVMigratorDescriptor_t migrator, int deviceSlotIndex, const void* srcSubSV, void* dstSubSV, custatevecIndex_t begin, custatevecIndex_t end) except* nogil
+cdef custatevecStatus_t _custatevecComputeExpectationBatchedGetWorkspaceSize(custatevecHandle_t handle, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, const custatevecIndex_t svStride, const void* matrices, cudaDataType_t matrixDataType, custatevecMatrixLayout_t layout, const uint32_t nMatrices, const uint32_t nBasisBits, custatevecComputeType_t computeType, size_t* extraWorkspaceSizeInBytes) except* nogil
+cdef custatevecStatus_t _custatevecComputeExpectationBatched(custatevecHandle_t handle, const void* batchedSv, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, custatevecIndex_t svStride, double2* expectationValues, const void* matrices, cudaDataType_t matrixDataType, custatevecMatrixLayout_t layout, const uint32_t nMatrices, const int32_t* basisBits, const uint32_t nBasisBits, custatevecComputeType_t computeType, void* extraWorkspace, size_t extraWorkspaceSizeInBytes) except* nogil

python/cuquantum/custatevec/_internal/custatevec_linux.pyx

@@ -102,11 +102,11 @@ cdef void* __custatevecAbs2SumArrayBatched = NULL
 cdef void* __custatevecCollapseByBitStringBatchedGetWorkspaceSize = NULL
 cdef void* __custatevecCollapseByBitStringBatched = NULL
 cdef void* __custatevecMeasureBatched = NULL
-cdef void* __custatevecComputeExpectationBatchedGetWorkspaceSize = NULL
-cdef void* __custatevecComputeExpectationBatched = NULL
 cdef void* __custatevecSubSVMigratorCreate = NULL
 cdef void* __custatevecSubSVMigratorDestroy = NULL
 cdef void* __custatevecSubSVMigratorMigrate = NULL
+cdef void* __custatevecComputeExpectationBatchedGetWorkspaceSize = NULL
+cdef void* __custatevecComputeExpectationBatched = NULL
 
 
 cdef void* load_library() except* nogil:
@@ -602,20 +602,6 @@ cdef int _check_or_init_custatevec() except -1 nogil:
             handle = load_library()
         __custatevecMeasureBatched = dlsym(handle, 'custatevecMeasureBatched')
 
-    global __custatevecComputeExpectationBatchedGetWorkspaceSize
-    __custatevecComputeExpectationBatchedGetWorkspaceSize = dlsym(RTLD_DEFAULT, 'custatevecComputeExpectationBatchedGetWorkspaceSize')
-    if __custatevecComputeExpectationBatchedGetWorkspaceSize == NULL:
-        if handle == NULL:
-            handle = load_library()
-        __custatevecComputeExpectationBatchedGetWorkspaceSize = dlsym(handle, 'custatevecComputeExpectationBatchedGetWorkspaceSize')
-    
-    global __custatevecComputeExpectationBatched
-    __custatevecComputeExpectationBatched = dlsym(RTLD_DEFAULT, 'custatevecComputeExpectationBatched')
-    if __custatevecComputeExpectationBatched == NULL:
-        if handle == NULL:
-            handle = load_library()
-        __custatevecComputeExpectationBatched = dlsym(handle, 'custatevecComputeExpectationBatched')
-    
     global __custatevecSubSVMigratorCreate
     __custatevecSubSVMigratorCreate = dlsym(RTLD_DEFAULT, 'custatevecSubSVMigratorCreate')
     if __custatevecSubSVMigratorCreate == NULL:
@@ -636,6 +622,20 @@ cdef int _check_or_init_custatevec() except -1 nogil:
         if handle == NULL:
             handle = load_library()
         __custatevecSubSVMigratorMigrate = dlsym(handle, 'custatevecSubSVMigratorMigrate')
+    
+    global __custatevecComputeExpectationBatchedGetWorkspaceSize
+    __custatevecComputeExpectationBatchedGetWorkspaceSize = dlsym(RTLD_DEFAULT, 'custatevecComputeExpectationBatchedGetWorkspaceSize')
+    if __custatevecComputeExpectationBatchedGetWorkspaceSize == NULL:
+        if handle == NULL:
+            handle = load_library()
+        __custatevecComputeExpectationBatchedGetWorkspaceSize = dlsym(handle, 'custatevecComputeExpectationBatchedGetWorkspaceSize')
+    
+    global __custatevecComputeExpectationBatched
+    __custatevecComputeExpectationBatched = dlsym(RTLD_DEFAULT, 'custatevecComputeExpectationBatched')
+    if __custatevecComputeExpectationBatched == NULL:
+        if handle == NULL:
+            handle = load_library()
+        __custatevecComputeExpectationBatched = dlsym(handle, 'custatevecComputeExpectationBatched')
 
     __py_custatevec_init = True
     return 0
@@ -849,12 +849,6 @@ cpdef dict _inspect_function_pointers():
     global __custatevecMeasureBatched
     data["__custatevecMeasureBatched"] = <intptr_t>__custatevecMeasureBatched
 
-    global __custatevecComputeExpectationBatchedGetWorkspaceSize
-    data["__custatevecComputeExpectationBatchedGetWorkspaceSize"] = <intptr_t>__custatevecComputeExpectationBatchedGetWorkspaceSize
-    
-    global __custatevecComputeExpectationBatched
-    data["__custatevecComputeExpectationBatched"] = <intptr_t>__custatevecComputeExpectationBatched
-
     global __custatevecSubSVMigratorCreate
     data["__custatevecSubSVMigratorCreate"] = <intptr_t>__custatevecSubSVMigratorCreate
 
@@ -863,6 +857,12 @@ cpdef dict _inspect_function_pointers():
 
     global __custatevecSubSVMigratorMigrate
     data["__custatevecSubSVMigratorMigrate"] = <intptr_t>__custatevecSubSVMigratorMigrate
+    
+    global __custatevecComputeExpectationBatchedGetWorkspaceSize
+    data["__custatevecComputeExpectationBatchedGetWorkspaceSize"] = <intptr_t>__custatevecComputeExpectationBatchedGetWorkspaceSize
+    
+    global __custatevecComputeExpectationBatched
+    data["__custatevecComputeExpectationBatched"] = <intptr_t>__custatevecComputeExpectationBatched
 
     return data
 
@@ -1551,26 +1551,6 @@ cdef custatevecStatus_t _custatevecMeasureBatched(custatevecHandle_t handle, voi
         handle, batchedSv, svDataType, nIndexBits, nSVs, svStride, bitStrings, bitOrdering, bitStringLen, randnums, collapse)
 
 
-cdef custatevecStatus_t _custatevecComputeExpectationBatchedGetWorkspaceSize(custatevecHandle_t handle, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, const custatevecIndex_t svStride, const void* matrices, cudaDataType_t matrixDataType, custatevecMatrixLayout_t layout, const uint32_t nMatrices, const uint32_t nBasisBits, custatevecComputeType_t computeType, size_t* extraWorkspaceSizeInBytes) except* nogil:
-    global __custatevecComputeExpectationBatchedGetWorkspaceSize
-    _check_or_init_custatevec()
-    if __custatevecComputeExpectationBatchedGetWorkspaceSize == NULL:
-        with gil:
-            raise FunctionNotFoundError("function custatevecComputeExpectationBatchedGetWorkspaceSize is not found")
-    return (<custatevecStatus_t (*)(custatevecHandle_t, cudaDataType_t, const uint32_t, const uint32_t, const custatevecIndex_t, const void*, cudaDataType_t, custatevecMatrixLayout_t, const uint32_t, const uint32_t, custatevecComputeType_t, size_t*) nogil>__custatevecComputeExpectationBatchedGetWorkspaceSize)(
-        handle, svDataType, nIndexBits, nSVs, svStride, matrices, matrixDataType, layout, nMatrices, nBasisBits, computeType, extraWorkspaceSizeInBytes)
-
-
-cdef custatevecStatus_t _custatevecComputeExpectationBatched(custatevecHandle_t handle, const void* batchedSv, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, custatevecIndex_t svStride, double2* expectationValues, const void* matrices, cudaDataType_t matrixDataType, custatevecMatrixLayout_t layout, const uint32_t nMatrices, const int32_t* basisBits, const uint32_t nBasisBits, custatevecComputeType_t computeType, void* extraWorkspace, size_t extraWorkspaceSizeInBytes) except* nogil:
-    global __custatevecComputeExpectationBatched
-    _check_or_init_custatevec()
-    if __custatevecComputeExpectationBatched == NULL:
-        with gil:
-            raise FunctionNotFoundError("function custatevecComputeExpectationBatched is not found")
-    return (<custatevecStatus_t (*)(custatevecHandle_t, const void*, cudaDataType_t, const uint32_t, const uint32_t, custatevecIndex_t, double2*, const void*, cudaDataType_t, custatevecMatrixLayout_t, const uint32_t, const int32_t*, const uint32_t, custatevecComputeType_t, void*, size_t) nogil>__custatevecComputeExpectationBatched)(
-        handle, batchedSv, svDataType, nIndexBits, nSVs, svStride, expectationValues, matrices, matrixDataType, layout, nMatrices, basisBits, nBasisBits, computeType, extraWorkspace, extraWorkspaceSizeInBytes)
-
-
 cdef custatevecStatus_t _custatevecSubSVMigratorCreate(custatevecHandle_t handle, custatevecSubSVMigratorDescriptor_t* migrator, void* deviceSlots, cudaDataType_t svDataType, int nDeviceSlots, int nLocalIndexBits) except* nogil:
     global __custatevecSubSVMigratorCreate
     _check_or_init_custatevec()
@@ -1599,3 +1579,23 @@ cdef custatevecStatus_t _custatevecSubSVMigratorMigrate(custatevecHandle_t handl
             raise FunctionNotFoundError("function custatevecSubSVMigratorMigrate is not found")
     return (<custatevecStatus_t (*)(custatevecHandle_t, custatevecSubSVMigratorDescriptor_t, int, const void*, void*, custatevecIndex_t, custatevecIndex_t) nogil>__custatevecSubSVMigratorMigrate)(
         handle, migrator, deviceSlotIndex, srcSubSV, dstSubSV, begin, end)
+
+
+cdef custatevecStatus_t _custatevecComputeExpectationBatchedGetWorkspaceSize(custatevecHandle_t handle, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, const custatevecIndex_t svStride, const void* matrices, cudaDataType_t matrixDataType, custatevecMatrixLayout_t layout, const uint32_t nMatrices, const uint32_t nBasisBits, custatevecComputeType_t computeType, size_t* extraWorkspaceSizeInBytes) except* nogil:
+    global __custatevecComputeExpectationBatchedGetWorkspaceSize
+    _check_or_init_custatevec()
+    if __custatevecComputeExpectationBatchedGetWorkspaceSize == NULL:
+        with gil:
+            raise FunctionNotFoundError("function custatevecComputeExpectationBatchedGetWorkspaceSize is not found")
+    return (<custatevecStatus_t (*)(custatevecHandle_t, cudaDataType_t, const uint32_t, const uint32_t, const custatevecIndex_t, const void*, cudaDataType_t, custatevecMatrixLayout_t, const uint32_t, const uint32_t, custatevecComputeType_t, size_t*) nogil>__custatevecComputeExpectationBatchedGetWorkspaceSize)(
+        handle, svDataType, nIndexBits, nSVs, svStride, matrices, matrixDataType, layout, nMatrices, nBasisBits, computeType, extraWorkspaceSizeInBytes)
+
+
+cdef custatevecStatus_t _custatevecComputeExpectationBatched(custatevecHandle_t handle, const void* batchedSv, cudaDataType_t svDataType, const uint32_t nIndexBits, const uint32_t nSVs, custatevecIndex_t svStride, double2* expectationValues, const void* matrices, cudaDataType_t matrixDataType, custatevecMatrixLayout_t layout, const uint32_t nMatrices, const int32_t* basisBits, const uint32_t nBasisBits, custatevecComputeType_t computeType, void* extraWorkspace, size_t extraWorkspaceSizeInBytes) except* nogil:
+    global __custatevecComputeExpectationBatched
+    _check_or_init_custatevec()
+    if __custatevecComputeExpectationBatched == NULL:
+        with gil:
+            raise FunctionNotFoundError("function custatevecComputeExpectationBatched is not found")
+    return (<custatevecStatus_t (*)(custatevecHandle_t, const void*, cudaDataType_t, const uint32_t, const uint32_t, custatevecIndex_t, double2*, const void*, cudaDataType_t, custatevecMatrixLayout_t, const uint32_t, const int32_t*, const uint32_t, custatevecComputeType_t, void*, size_t) nogil>__custatevecComputeExpectationBatched)(
+        handle, batchedSv, svDataType, nIndexBits, nSVs, svStride, expectationValues, matrices, matrixDataType, layout, nMatrices, basisBits, nBasisBits, computeType, extraWorkspace, extraWorkspaceSizeInBytes)

python/cuquantum/custatevec/custatevec.pxd

@@ -112,8 +112,8 @@ cpdef abs2sum_array_batched(intptr_t handle, intptr_t batched_sv, int sv_data_ty
 cpdef size_t collapse_by_bit_string_batched_get_workspace_size(intptr_t handle, uint32_t n_svs, bit_strings, norms) except? 0
 cpdef collapse_by_bit_string_batched(intptr_t handle, intptr_t batched_sv, int sv_data_type, uint32_t n_index_bits, uint32_t n_svs, int64_t sv_stride, bit_strings, bit_ordering, uint32_t bit_string_len, norms, intptr_t extra_workspace, size_t extra_workspace_size_in_bytes)
 cpdef measure_batched(intptr_t handle, intptr_t batched_sv, int sv_data_type, uint32_t n_index_bits, uint32_t n_svs, int64_t sv_stride, intptr_t bit_strings, bit_ordering, uint32_t bit_string_len, randnums, int collapse)
-cpdef size_t compute_expectation_batched_get_workspace_size(intptr_t handle, int sv_data_type, uint32_t n_index_bits, uint32_t n_svs, int64_t sv_stride, intptr_t matrices, int matrix_data_type, int layout, uint32_t n_matrices, uint32_t n_basis_bits, int compute_type) except? 0
-cpdef compute_expectation_batched(intptr_t handle, intptr_t batched_sv, int sv_data_type, uint32_t n_index_bits, uint32_t n_svs, int64_t sv_stride, intptr_t expectation_values, intptr_t matrices, int matrix_data_type, int layout, uint32_t n_matrices, basis_bits, uint32_t n_basis_bits, int compute_type, intptr_t extra_workspace, size_t extra_workspace_size_in_bytes)
 cpdef intptr_t sub_sv_migrator_create(intptr_t handle, intptr_t device_slots, int sv_data_type, int n_device_slots, int n_local_index_bits) except? 0
 cpdef sub_sv_migrator_destroy(intptr_t handle, intptr_t migrator)
 cpdef sub_sv_migrator_migrate(intptr_t handle, intptr_t migrator, int device_slot_ind_ex, intptr_t src_sub_sv, intptr_t dst_sub_sv, int64_t begin, int64_t end)
+cpdef size_t compute_expectation_batched_get_workspace_size(intptr_t handle, int sv_data_type, uint32_t n_index_bits, uint32_t n_svs, int64_t sv_stride, intptr_t matrices, int matrix_data_type, int layout, uint32_t n_matrices, uint32_t n_basis_bits, int compute_type) except? 0
+cpdef compute_expectation_batched(intptr_t handle, intptr_t batched_sv, int sv_data_type, uint32_t n_index_bits, uint32_t n_svs, int64_t sv_stride, intptr_t expectation_values, intptr_t matrices, int matrix_data_type, int layout, uint32_t n_matrices, basis_bits, uint32_t n_basis_bits, int compute_type, intptr_t extra_workspace, size_t extra_workspace_size_in_bytes)