da4ml 0.1.0__py3-none-any.whl → 0.1.2__py3-none-any.whl

da4ml/_version.py

@@ -1,8 +1,13 @@
-# file generated by setuptools_scm
+# file generated by setuptools-scm
 # don't change, don't track in version control
+
+__all__ = ["__version__", "__version_tuple__", "version", "version_tuple"]
+
 TYPE_CHECKING = False
 if TYPE_CHECKING:
-    from typing import Tuple, Union
+    from typing import Tuple
+    from typing import Union
+
     VERSION_TUPLE = Tuple[Union[int, str], ...]
 else:
     VERSION_TUPLE = object
@@ -12,5 +17,5 @@ __version__: str
 __version_tuple__: VERSION_TUPLE
 version_tuple: VERSION_TUPLE
 
-__version__ = version = '0.1.0'
-__version_tuple__ = version_tuple = (0, 1, 0)
+__version__ = version = '0.1.2'
+__version_tuple__ = version_tuple = (0, 1, 2)

da4ml/cmvm/__init__.py

@@ -24,7 +24,7 @@ _ = fn_from_kernel(
     symmetrics=symmetrics,
     depths=depths,
     n_beams=1,
-    dc=0,
+    dc=None,
     n_inp_max=-1,
     n_out_max=-1,
     codegen_backend=PyCodegenBackend(),

da4ml/cmvm/api.py

@@ -45,7 +45,7 @@ def fn_from_kernel(
     n_beams : int, optional
         Number of beams to use in beam search. Defaults to 1. (Currently disabled!)
     dc : int | None, optional
-        Delay constraint. Not implemented yet. Defaults to None.
+        Delay constraint. Not (properly) implemented yet. Defaults to None.
     n_inp_max : int, optional
         Number of inputs to process in one block. Defaults to -1 (no limit). Decrease to improve performance, but result will be less optimal.
     n_out_max : int, optional

da4ml/cmvm/cmvm.py

@@ -1,4 +1,5 @@
 import heapq
+from collections.abc import Sequence
 from math import ceil
 
 import numpy as np
@@ -22,16 +23,6 @@ def extract_pairs(
     _stat = np.zeros((d_in, d_in, n_bit, 2), dtype=np.int64)
     process_locs = np.zeros((d_in, d_out), dtype=np.bool_)
 
-    if dc is not None:
-        depths = np.zeros(d_in, dtype=np.int64)
-        for pos in range(d_in):
-            depths[pos] = precisions[pos]._depth
-        depth_min = np.min(depths)
-        mask = depths <= depth_min + dc
-        if np.count_nonzero(mask) >= 2:
-            for n in range(d_out):
-                process_locs[:, n] = mask
-
     if updated is not None:
         for i in range(len(updated)):
             pos = updated[i]
@@ -43,6 +34,16 @@ def extract_pairs(
                 if np.any(csd[pos][n]) and precisions[pos].b != 0:
                     process_locs[pos, n] = True
 
+    if dc is not None:
+        depths = np.zeros(d_in, dtype=np.int64)
+        for pos in range(d_in):
+            depths[pos] = precisions[pos]._depth
+        depth_min = np.min(depths)
+        mask = depths <= depth_min + dc
+        if np.count_nonzero(mask) >= 2:
+            for n in range(d_out):
+                process_locs[:, n] &= mask
+
     args: list[tuple[int, int, int]] = []
     for pos0 in range(d_in):
         for pos1 in range(d_in):
@@ -100,11 +101,11 @@ def init_var(
 
 @njit
 def init_vars(
-    ks: list[bool],
-    bs: list[int],
-    is_: list[int],
-    symmetrics: list[bool],
-    depths: list[int],
+    ks: tuple[bool, ...],
+    bs: tuple[int, ...],
+    is_: tuple[int, ...],
+    symmetrics: tuple[bool, ...],
+    depths: tuple[int, ...],
 ):
     n = len(ks)
     vars_ = []
@@ -116,11 +117,11 @@ def init_vars(
 @njit
 def init_state(
     kernel: np.ndarray,
-    signs: list[bool],
-    bits: list[int],
-    int_bits: list[int],
-    symmetrics: list[bool],
-    depths: list[int],
+    signs: tuple[bool, ...],
+    bits: tuple[int, ...],
+    int_bits: tuple[int, ...],
+    symmetrics: tuple[bool, ...],
+    depths: tuple[int, ...],
 ):
     assert kernel.ndim == 2
     assert len(signs) == len(bits) == len(int_bits) == len(symmetrics) == len(depths) == kernel.shape[0]
@@ -261,11 +262,11 @@ def cmvm_cse(state: DAState, progress=None, beams: int = 1, dc=None):
 @njit
 def compile_kernel_mono(
     kernel: np.ndarray,
-    signs: list[bool],
-    bits: list[int],
-    int_bits: list[int],
-    symmetrics: list[bool],
-    depths: list[int],
+    signs: tuple[bool, ...],
+    bits: tuple[int, ...],
+    int_bits: tuple[int, ...],
+    symmetrics: tuple[bool, ...],
+    depths: tuple[int, ...],
     n_beams: int = 1,
     dc: int | None = None,
 ):
@@ -274,14 +275,13 @@ def compile_kernel_mono(
     return _state
 
 
-# @njit(cache=True)
 def compile_kernel(
     kernel: np.ndarray,
-    signs: list[bool],
-    bits: list[int],
-    int_bits: list[int],
-    symmetrics: list[bool],
-    depths: list[int],
+    signs: Sequence[bool],
+    bits: Sequence[int],
+    int_bits: Sequence[int],
+    symmetrics: Sequence[bool],
+    depths: Sequence[int],
     n_beams: int = 1,
     dc: int | None = None,
     n_inp_max: int = -1,
@@ -315,11 +315,11 @@ def compile_kernel(
 
         # unify input type to prevent recompilation
         _kernel = np.ascontiguousarray(_kernel)
-        _signs = [bool(v) for v in _signs]
-        _bits = [int(v) for v in _bits]
-        _int_bits = [int(v) for v in _int_bits]
-        _symmetrics = [bool(v) for v in _symmetrics]
-        _depths = [int(v) for v in _depths]
+        _signs = tuple(bool(v) for v in _signs)
+        _bits = tuple(int(v) for v in _bits)
+        _int_bits = tuple(int(v) for v in _int_bits)
+        _symmetrics = tuple(bool(v) for v in _symmetrics)
+        _depths = tuple(int(v) for v in _depths)
         try:
             states[i][j] = compile_kernel_mono(_kernel, _signs, _bits, _int_bits, _symmetrics, _depths, n_beams, dc)
         except AssertionError:

{da4ml-0.1.0.dist-info → da4ml-0.1.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
-Metadata-Version: 2.2
+Metadata-Version: 2.4
 Name: da4ml
-Version: 0.1.0
+Version: 0.1.2
 Summary: Digital Arithmetic for Machine Learning
 Author-email: Chang Sun <chsun@cern.ch>
 License: GNU Lesser General Public License v3 (LGPLv3)
@@ -17,6 +17,9 @@ Classifier: Programming Language :: Python :: 3.13
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
+Requires-Dist: llvmlite>=0.43
+Requires-Dist: numba>=0.60
+Dynamic: license-file
 
 # da4ml: Distributed Arithmetic for Machine Learning
 

{da4ml-0.1.0.dist-info → da4ml-0.1.2.dist-info}/RECORD

@@ -1,9 +1,9 @@
 da4ml/__init__.py,sha256=kkpf91aU4n9MMsX63Me90mo_0JSQyU3N-vhgTIrjkT0,437
-da4ml/_version.py,sha256=IMl2Pr_Sy4LVRKy_Sm4CdwUl1Gryous6ncL96EMYsnM,411
-da4ml/cmvm/__init__.py,sha256=fgeQr2Tbg4rqEix3OJwm5z0o6WJxkvLCkqJKcjBxBSI,845
-da4ml/cmvm/api.py,sha256=jA1PBkZqi12Wa8G7V3dIyGzuNePdCpf9juJoOb-dutM,3450
+da4ml/_version.py,sha256=bSmADqydH8nBu-J4lG8UVuR7hnU_zcwhnSav2oQ0W0A,511
+da4ml/cmvm/__init__.py,sha256=UUnJxtZGQpqVWKT_djuPr7CqyiWlUrmKoyAvH7f4-q8,848
+da4ml/cmvm/api.py,sha256=G7YZ1n32EaYzsahSUwB-0vvRZt1Q11idShhn735vDCM,3461
 da4ml/cmvm/balanced_reduction.py,sha256=fDKaRIY4WkRo3s5dGWeWmylDgjMD-hr1-tpdwu7EL6Y,1533
-da4ml/cmvm/cmvm.py,sha256=s2hMPaDPk_1bjmoPnnjjnMXmxvlSZUoiubDQpMc8wyE,10536
+da4ml/cmvm/cmvm.py,sha256=4RN6sKLNxybHZcBoUkw4MBluLZoEhAArutt8BE5NCNE,10689
 da4ml/cmvm/codegen.py,sha256=w_1xR36Oxwb-1XivFwjQ_n2uRw-abjwzjOhwARg_93k,6088
 da4ml/cmvm/csd.py,sha256=k-9k0CigqnvyrgtXzBvKLZ32FdwWUE_EhAlXkK6Mlxk,1988
 da4ml/cmvm/fixed_variable.py,sha256=APPT2EN7hOwjHIaD5JdOpd7riGUpmtXUJ6QaAnE1cjw,5753
@@ -11,8 +11,8 @@ da4ml/cmvm/graph_compile.py,sha256=bODuY7DAi-9L-ZleAzK0szZDrEH4Ljq2ocINUHXD588,3
 da4ml/cmvm/nb_fixed_precision.py,sha256=M2bnflNNOGLgUsI0QGlF2noT5ZJD-2bU2dZ3lw8GZvM,2457
 da4ml/cmvm/scoring.py,sha256=EUREpoyQHUe-vpRat_JjRD73uQhD0dOhnG2D3opazwQ,1375
 da4ml/cmvm/utils.py,sha256=Svp82TOtpNOtKfXlCkijq7DBSbMEqyWoY2PAYVhXXQY,288
-da4ml-0.1.0.dist-info/LICENSE,sha256=46mU2C5kSwOnkqkw9XQAJlhBL2JAf1_uCD8lVcXyMRg,7652
-da4ml-0.1.0.dist-info/METADATA,sha256=r6-o6o3W3Qr8IXjj1roJNWeLGghIuhdxBC6aKHqIUdo,5428
-da4ml-0.1.0.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-da4ml-0.1.0.dist-info/top_level.txt,sha256=N0tnKVwRqFiffFdeAzCgFq71hUNySh5-ITbNd6-R58Q,6
-da4ml-0.1.0.dist-info/RECORD,,
+da4ml-0.1.2.dist-info/licenses/LICENSE,sha256=46mU2C5kSwOnkqkw9XQAJlhBL2JAf1_uCD8lVcXyMRg,7652
+da4ml-0.1.2.dist-info/METADATA,sha256=YirD1SJdRbYNKdbY68TWtacjfs-bUrcWIbQBdoIIaYw,5507
+da4ml-0.1.2.dist-info/WHEEL,sha256=1tXe9gY0PYatrMPMDd6jXqjfpz_B-Wqm32CPfRC58XU,91
+da4ml-0.1.2.dist-info/top_level.txt,sha256=N0tnKVwRqFiffFdeAzCgFq71hUNySh5-ITbNd6-R58Q,6
+da4ml-0.1.2.dist-info/RECORD,,

{da4ml-0.1.0.dist-info → da4ml-0.1.2.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.8.0)
+Generator: setuptools (77.0.3)
 Root-Is-Purelib: true
 Tag: py3-none-any