mlkem_native.h: Introduce MLK_TOTAL_ALLOC constants
#1468
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
Author
|
I have included the extension as a separate commit. |
mkannwischer
force-pushed
the
mlk-max-alloc
branch
from
2f4200d to
920f9dd
Compare
mkannwischer
force-pushed
the
mlk-max-alloc
branch
from
c37874d to
a1aac27
Compare
hanno-becker
requested changes
Jan 12, 2026
Contributor
hanno-becker
left a comment
hanno-becker
left a comment
There was a problem hiding this comment.
I think it is slightly confusing to expose MLK_MAX_TOTAL_ALLOC as the maximum across all security levels, even in a single-level build. I'd prefer to only export constants explicitly labeled by a security level.
For consumers implementing custom allocations (in cases where the stack is
not large enough), it may be important to know how much space has to be
reserved for allocations of large structures.
This commit adds per-operation allocation constants:
- MLK_TOTAL_ALLOC_{512,768,1024}_{KEYPAIR,ENCAPS,DECAPS} for each parameter
set and operation, indicating peak memory consumption.
- MLK_MAX_TOTAL_ALLOC_{KEYPAIR,ENCAPS,DECAPS} for maximum across parameter
sets per operation.
- MLK_TOTAL_ALLOC_{512,768,1024} for maximum across operations per parameter
set.
- MLK_MAX_TOTAL_ALLOC for maximum across all parameter sets and operations.
The alloc_test is extended to check that the constants indeed match the peak
memory consumption of the corresponding functions.
- Ported from pq-code-package/mldsa-native#850
- Resolves #1442
Signed-off-by: Matthias J. Kannwischer <[email protected]>
Multilevel builds require MLK_CONFIG_NO_SUPERCOP since the SUPERCOP naming convention (crypto_kem_xxx) does not disambiguate between parameter sets. Add MLK_CONFIG_NO_SUPERCOP to all multilevel build configs and propagate it to the API layer in the legacy config path. This also fixes problem with mlkem_native.h: On multiple inclusions of mlkem_native.h without NO_SUPERCOP set, mlkem_native.h would not undef the API_CONFIG macros in the end of the file causing the second inclusion of mlkem_native.h to set the LEGACY_CONFIG option. - Ported from pq-code-package/mldsa-native#869 Signed-off-by: Matthias J. Kannwischer <[email protected]>
Split KEYPAIR allocation constants into _NO_PCT and _PCT variants to accurately reflect memory usage with and without pairwise consistency testing (PCT). MLK_TOTAL_ALLOC_*_KEYPAIR now automatically selects the appropriate value based on MLK_CONFIG_KEYGEN_PCT. For legacy configs, assume PCT is enabled (conservative). The default alloc test config no longer enables PCT; PCT is tested via config-variations. - Ported from pq-code-package/mldsa-native#869 Signed-off-by: Matthias J. Kannwischer <[email protected]>
Remove MLK_MAX_TOTAL_ALLOC and MLK_MAX_TOTAL_ALLOC_{KEYPAIR,ENCAPS,DECAPS}
constants. These exposed the maximum allocation across all security
levels even in single-level builds, which was confusing. Users should
instead use the explicitly security-level-labeled constants such as
MLK_TOTAL_ALLOC_1024_KEYPAIR.
The per-level MLK_TOTAL_ALLOC_{512,768,1024}_{KEYPAIR,ENCAPS,DECAPS} and
aggregate MLK_TOTAL_ALLOC_{512,768,1024} constants remain available.
Signed-off-by: Matthias J. Kannwischer <[email protected]>
mkannwischer
force-pushed
the
mlk-max-alloc
branch
from
a1aac27 to
26cdf45
Compare
Contributor
Author
Okay, yes, makes sense. I have removed the macros. Also removed it in mldsa-native in pq-code-package/mldsa-native#869. |
hanno-becker
approved these changes
Jan 12, 2026
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
For consumers implementing custom allocations (in cases where the stack is not large enough), it may be important to know how much space has to be reserved for allocations of large structures.
This commit adds per-operation allocation constants:
The alloc_test is extended to check that the constants indeed match the peak memory consumption of the corresponding functions.
I included the extension from #869 that differentiates between keygen w/ and w/o PCT. The commits are kept separate though to follow the same structure as in mldsa-native.
MLD_TOTAL_ALLOCconstants mldsa-native#850