Add CRC64NVME to smithy-checksums

packages/smithy-checksums/.changes/next-release/smithy-checksums-feature-6c167ff7220c4033b9eb3299d4627564.json

@@ -0,0 +1,4 @@
+{
+  "type": "feature",
+  "description": "Added CRC64NVME checksums."
+}

packages/smithy-checksums/README.md

@@ -1,7 +1,7 @@
 # smithy-checksums
 
-This package provides the checksum algorithms used by Smithy clients: CRC32 and
-CRC32C, and, through `new()`, SHA-1, SHA-256, SHA-512, and MD5.
+This package provides the checksum algorithms used by Smithy clients: CRC32,
+CRC32C, CRC64NVME, and, through `new()`, SHA-1, SHA-256, SHA-512, and MD5.
 
 ```python
 import smithy_checksums

packages/smithy-checksums/build_backend/smithy_checksums_build.py

@@ -36,7 +36,7 @@
 _CARGO_TOML = Path(__file__).parent.parent / "Cargo.toml"
 _PURE_PYTHON_HINT = (
     f"To install the pure-Python implementation instead, set {_ENV_VAR}=never. It "
-    "produces identical checksums, but CRC32C is much slower."
+    "produces identical checksums, but CRC32C and CRC64NVME are much slower."
 )
 
 

packages/smithy-checksums/pyproject.toml

@@ -9,7 +9,7 @@ authors = [
 description = "Checksum algorithms for Smithy tooling, with native acceleration."
 readme = "README.md"
 license = {text = "Apache License 2.0"}
-keywords = ["smithy", "sdk", "checksum", "crc32c"]
+keywords = ["smithy", "sdk", "checksum", "crc32c", "crc64nvme"]
 classifiers = [
   "Development Status :: 2 - Pre-Alpha",
   "Intended Audience :: Developers",

packages/smithy-checksums/rust/lib.rs

@@ -48,6 +48,7 @@ macro_rules! crc_state {
 
 crc_state!(Crc32State, Crc32IsoHdlc, 4);
 crc_state!(Crc32cState, Crc32Iscsi, 4);
+crc_state!(Crc64NvmeState, Crc64Nvme, 8);
 
 fn lock<S>(state: &Mutex<S>) -> MutexGuard<'_, S> {
     // A panic while holding the lock can't leave a checksum half-updated in a
@@ -127,9 +128,10 @@ macro_rules! checksum_class {
 
 checksum_class!(Crc32, Crc32State);
 checksum_class!(Crc32C, Crc32cState);
+checksum_class!(Crc64Nvme, Crc64NvmeState);
 
 #[pymodule]
 mod _native {
     #[pymodule_export]
-    use super::{Crc32, Crc32C};
+    use super::{Crc32, Crc32C, Crc64Nvme};
 }

packages/smithy-checksums/src/smithy_checksums/__init__.py

@@ -2,10 +2,11 @@
 #  SPDX-License-Identifier: Apache-2.0
 """Checksum algorithms for Smithy clients.
 
-CRC32 and CRC32C use a native extension when one is available for the platform
-and fall back to pure Python otherwise. The fallback produces identical results
-but is orders of magnitude slower, except for CRC32, which uses :mod:`zlib`. SHA-1, SHA-256, SHA-512, and MD5 are available
-through :func:`new` and always use :mod:`hashlib`.
+CRC32, CRC32C, and CRC64NVME use a native extension when one is available for the
+platform and fall back to pure Python otherwise. The fallback produces identical
+results but is orders of magnitude slower, except for CRC32, which uses :mod:`zlib`.
+SHA-1, SHA-256, SHA-512, and MD5 are available through :func:`new` and always use
+:mod:`hashlib`.
 """
 
 import hashlib
@@ -22,6 +23,7 @@
 __all__ = (
     "CRC32",
     "CRC32C",
+    "CRC64NVME",
     "HAS_NATIVE_EXTENSION",
     "Checksum",
     "new",
@@ -30,7 +32,7 @@
 HAS_NATIVE_EXTENSION: Final[bool] = _NATIVE
 """Whether the native extension is in use.
 
-When it isn't, CRC32C is computed in pure Python.
+When it isn't, CRC32C and CRC64NVME are computed in pure Python.
 """
 
 
@@ -121,18 +123,27 @@ class CRC32C(_BackendChecksum):
     digest_size = 4
 
 
-_CHECKSUMS: Final = {cls.name: cls for cls in (CRC32, CRC32C)}
+class CRC64NVME(_BackendChecksum):
+    """CRC-64/NVME."""
+
+    __slots__ = ()
+
+    name = "crc64nvme"
+    digest_size = 8
+
+
+_CHECKSUMS: Final = {cls.name: cls for cls in (CRC32, CRC32C, CRC64NVME)}
 _HASHLIB_ALGORITHMS: Final = frozenset(("sha1", "sha256", "sha512"))
 
 
 def new(name: str) -> Checksum:
     """Creates a checksum from the name of its algorithm.
 
     Names are case-insensitive, so the values of a checksum algorithm enum such as
-    ``CRC32C`` or ``SHA256`` can be passed directly.
+    ``CRC64NVME`` or ``SHA256`` can be passed directly.
 
-    :param name: One of ``crc32``, ``crc32c``, ``sha1``, ``sha256``, ``sha512``, or
-        ``md5``.
+    :param name: One of ``crc32``, ``crc32c``, ``crc64nvme``, ``sha1``, ``sha256``,
+        ``sha512``, or ``md5``.
     :raises ValueError: If the algorithm isn't supported.
     """
     key = name.lower()

packages/smithy-checksums/src/smithy_checksums/_native.pyi

@@ -15,3 +15,8 @@ class Crc32C:
     def update(self, data: _Bytes, /) -> None: ...
     def digest(self) -> bytes: ...
     def copy(self) -> Self: ...
+
+class Crc64Nvme:
+    def update(self, data: _Bytes, /) -> None: ...
+    def digest(self) -> bytes: ...
+    def copy(self) -> Self: ...

packages/smithy-checksums/src/smithy_checksums/_private/backend.py

@@ -31,11 +31,13 @@ def _load() -> tuple[bool, Mapping[str, ImplFactory]]:
         return False, {
             "crc32": crc.Crc32,
             "crc32c": crc.Crc32C,
+            "crc64nvme": crc.Crc64Nvme,
         }
 
     return True, {
         "crc32": _native.Crc32,
         "crc32c": _native.Crc32C,
+        "crc64nvme": _native.Crc64Nvme,
     }
 
 

packages/smithy-checksums/src/smithy_checksums/_private/crc.py

@@ -76,3 +76,13 @@ class Crc32C(_ReflectedCrc):
     _table = _reflected_table(0x82F63B78)
     _mask = 0xFFFFFFFF
     _size = 4
+
+
+class Crc64Nvme(_ReflectedCrc):
+    """CRC-64/NVME."""
+
+    __slots__ = ()
+
+    _table = _reflected_table(0x9A6C9329AC4BC9B5)
+    _mask = 0xFFFFFFFFFFFFFFFF
+    _size = 8

packages/smithy-checksums/tests/unit/test_checksums.py

@@ -10,6 +10,7 @@
 from smithy_checksums import (
     CRC32,
     CRC32C,
+    CRC64NVME,
     Checksum,
     new,
 )
@@ -19,6 +20,7 @@
 CLASSES = {
     "crc32": CRC32,
     "crc32c": CRC32C,
+    "crc64nvme": CRC64NVME,
 }
 
 
@@ -90,6 +92,7 @@ def compute(_: int) -> str:
     [
         ("crc32", CRC32),
         ("CRC32C", CRC32C),
+        ("CRC64NVME", CRC64NVME),
     ],
 )
 def test_new_returns_checksum_classes(name: str, expected: type[Checksum]) -> None:

packages/smithy-checksums/tests/unit/test_implementations.py

@@ -23,6 +23,7 @@
 PURE: dict[str, Factory] = {
     "crc32": crc.Crc32,
     "crc32c": crc.Crc32C,
+    "crc64nvme": crc.Crc64Nvme,
 }
 
 
@@ -41,6 +42,7 @@ def _load_native() -> ModuleType | None:
     else {
         "crc32": _NATIVE_MODULE.Crc32,
         "crc32c": _NATIVE_MODULE.Crc32C,
+        "crc64nvme": _NATIVE_MODULE.Crc64Nvme,
     }
 )
 requires_native = pytest.mark.skipif(
@@ -173,6 +175,7 @@ def test_matches_crt(impls: dict[str, Factory], algorithm: str) -> None:
     compute: dict[str, Callable[[bytes], bytes]] = {
         "crc32": lambda data: crt.crc32(data).to_bytes(4, "big"),
         "crc32c": lambda data: crt.crc32c(data).to_bytes(4, "big"),
+        "crc64nvme": lambda data: crt.crc64nvme(data).to_bytes(8, "big"),
     }
     data = os.urandom(3000)
     for length in [*range(0, 300, 7), 1023, 1024, 1025, 3000]:

packages/smithy-checksums/tests/unit/vectors.py

@@ -2,11 +2,12 @@
 # SPDX-License-Identifier: Apache-2.0
 """Expected digests shared by the checksum tests."""
 
-ALGORITHMS = ("crc32", "crc32c")
+ALGORITHMS = ("crc32", "crc32c", "crc64nvme")
 
 # Digests from independent sources: the standard check value of each CRC (the
-# digest of "123456789"), the CRC32C test vectors from RFC 3720, and "Hello world"
-# digests that match the ones the CRT produces.
+# digest of "123456789"), the CRC32C test vectors from RFC 3720, the CRC64NVME test
+# cases from the NVMe NVM Command Set specification, and "Hello world" digests that
+# match the ones the CRT produces.
 KNOWN_ANSWERS: dict[str, dict[bytes, str]] = {
     "crc32": {b"123456789": "cbf43926", b"Hello world": "8bd69e52"},
     "crc32c": {
@@ -19,6 +20,15 @@
         bytes(range(32)): "46dd794e",
         bytes(range(31, -1, -1)): "113fdb5c",
     },
+    "crc64nvme": {
+        b"123456789": "ae8b14860a799888",
+        b"Hello world": "38e259d03f312adb",
+        # The 64b CRC test cases in the NVMe NVM Command Set specification.
+        bytes(4096): "6482d367eb22b64e",
+        b"\xff" * 4096: "c0ddba7302eca3ac",
+        bytes(i % 256 for i in range(4096)): "3e729f5f6750449c",
+        bytes(255 - i % 256 for i in range(4096)): "9a2df64b8e9e517e",
+    },
 }
 
 
@@ -64,4 +74,21 @@ def pattern(length: int) -> bytes:
         1025: "c8d03add",
         16385: "896b2de8",
     },
+    "crc64nvme": {
+        0: "0000000000000000",
+        1: "d5da5047efec8728",
+        3: "3f3990291e6454d7",
+        4: "29513a0b0028dea8",
+        8: "7703cc98ac9209d7",
+        9: "0eb6fdbb89d04d0b",
+        16: "cf6454c75df95620",
+        17: "5c34355fb173cefd",
+        128: "a5296136625de4f7",
+        129: "c3d556fcf23d6a24",
+        240: "717e94e4fd87236d",
+        241: "2737bbceef8467bc",
+        1024: "fee95dc7adac063f",
+        1025: "034aba81017791d2",
+        16385: "9c0531959bc1eacf",
+    },
 }

scripts/api-snapshot/api.json

@@ -608,6 +608,7 @@
   "smithy_checksums": {
     "CRC32": "class",
     "CRC32C": "class",
+    "CRC64NVME": "class",
     "Checksum": "protocol",
     "HAS_NATIVE_EXTENSION": "bool",
     "new": "function"