Add ppc64le backend (supports p8 and above architectures) [full CI]#1677

Draft

mkannwischer wants to merge 29 commits into

mainfrom

ppc64le

Commits on May 19, 2026

CBMC: Introduce separate proofs for Keccak XXX_c() functions

willieyz
authored and
dannytsen
committed
CI: Update AWS-LC integration to current HEAD

mkannwischer
authored and
dannytsen
committed
MAINTAINERS.md: Remove affiliations and chat

mkannwischer
authored and
dannytsen
committed
CI: Add clang22 tests and constant-time tests

mkannwischer
authored and
dannytsen
committed
Updated patch for ML_KEM for ppc64le supports p8 and above architectures and tested.
dannytsen
committed
1. Update supporting system to p9 and above or
dannytsen
committed
Re-run autogen. Here is the output.
dannytsen
committed
Fixed more typos.
dannytsen
committed
Fixed monolithic_build_multilevel_native functional test by manually added
dannytsen
committed
Fixed markdown heading in README files.
dannytsen
committed
Add autogen support for ppc64le

bhess
authored and
dannytsen
committed
Merged with Basil's autogen and run autogen.
dannytsen
committed
Have capability checked to support PPC_FEATURE2_ARCH_2_07 and above
dannytsen
committed
Fixed parameter passed to getauxval and remove machine directive.
dannytsen
committed
Remove capability check PPC_FEATURE2_ARCH_2_07.
dannytsen
committed
This patch fixed the following items,
dannytsen
committed
Fixed missing balign in intt_ppc.S.
dannytsen
committed
Fixed CFI issue by re-running simpasm without cfify. Not sure if it is correct.
dannytsen
committed
This patch fixed the following,
dannytsen
committed
extend autogen/simpasm for ppc64le backend. Auto-generate zeta values

bhess
authored and
dannytsen
committed
rerun autogen

bhess
authored and
dannytsen
committed
Use Barrett for NTT twiddle products, and calculate the corresponding twisted zetas.

bhess
authored and
dannytsen
committed

Commits on May 21, 2026

ppc64le: fix Python and magic-constant linting
bhess
committed
ppc64le: fix power7 fallback by compiling for power7 but running for power8 (nix libc is compiled for power8 and otherwise causes illegal instructions.
bhess
committed
ppc64le: workaround to avoid hol-light failures
bhess
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add ppc64le backend (supports p8 and above architectures) [full CI]#1677

Add ppc64le backend (supports p8 and above architectures) [full CI]#1677
mkannwischer wants to merge 29 commits into
mainfrom
ppc64le

Commits on May 19, 2026

CBMC: Introduce separate proofs for Keccak XXX_c() functions

CI: Update AWS-LC integration to current HEAD

MAINTAINERS.md: Remove affiliations and chat

CI: Add clang22 tests and constant-time tests

Updated patch for ML_KEM for ppc64le supports p8 and above architectures and tested.

1. Update supporting system to p9 and above or

Re-run autogen. Here is the output.

Fixed more typos.

Fixed monolithic_build_multilevel_native functional test by manually added

Fixed markdown heading in README files.

Add autogen support for ppc64le

Merged with Basil's autogen and run autogen.

Have capability checked to support PPC_FEATURE2_ARCH_2_07 and above

Fixed parameter passed to getauxval and remove machine directive.

Remove capability check PPC_FEATURE2_ARCH_2_07.

This patch fixed the following items,

Fixed missing balign in intt_ppc.S.

Fixed CFI issue by re-running simpasm without cfify. Not sure if it is correct.

This patch fixed the following,

extend autogen/simpasm for ppc64le backend. Auto-generate zeta values

rerun autogen

Use Barrett for NTT twiddle products, and calculate the corresponding twisted zetas.

Commits on May 20, 2026

ppc64le: address review comments from mkannwischer

Merge pull request #4 from bhess/bhe-review

Merge branch 'main' into new_main

Updates after merge

Commits on May 21, 2026

ppc64le: fix Python and magic-constant linting

ppc64le: fix power7 fallback by compiling for power7 but running for power8 (nix libc is compiled for power8 and otherwise causes illegal instructions.

ppc64le: workaround to avoid hol-light failures