David Testé
39d77299ed
chore(bench): harmonize dex benchmark function names
2025-06-18 09:47:57 +02:00
Arthur Meyre
c841e3be6e
chore: add new codeowners for HPU and CUDA code
2025-06-17 11:07:38 +02:00
Nicolas Sarlin
aaeb46f074
feat(shortint): add compression for squashed noise ciphertexts
2025-06-16 18:08:51 +02:00
Nicolas Sarlin
8f7281c219
fix(shortint): handle empty list in compression
2025-06-16 18:08:51 +02:00
tmontaigu
11e86e6162
chore(csprng): bump to 0.6.0
...
Some (breaking) changes were made to a trait in CSPRNG
2025-06-16 14:05:47 +02:00
David Testé
41c92e06a8
chore(ci): add missing env variable in cuda release workflow
2025-06-16 12:55:31 +02:00
Pedro Alves
f25f394763
feat(gpu): add support to GPU-accelerated expand to HL's CompactCiphertextList
...
- Drops integer's CudaCompactCiphertextList
2025-06-13 19:18:11 +02:00
pgardratzama
3230c4bb97
chore(hpu): devo was still there, tool compiled in release mode
2025-06-13 16:46:31 +02:00
Baptiste Roux
159a85fc8c
chore(hpu): Fix some lint in hpu-v80 ffi
2025-06-13 16:46:31 +02:00
Helder Campos
4cec6fb247
chore(hpu): Fixed some README.md typos.
...
Also rename pdi_mgmt to hpu_archive_mgmt to match new naming
2025-06-13 16:46:31 +02:00
Baptiste Roux
16c997d686
feat(hpu): Add support for tandem pdi
...
Add support for hpu archive and Fpga reloading.
Rely on Tandem implementation for hot-reloading of FPGA.
Add reload procedure inside ffi/v80 backend.
Now when starting application on HpuV80, a first check of version is done.
If version mismatch, a pdi reload is triggered
2025-06-13 16:46:31 +02:00
David Testé
dcd1af72d4
chore(ci): fix missing env variable for 4090 benchmark
2025-06-13 11:43:16 +02:00
Agnes Leroy
9bf9107e9e
feat(gpu): add memory tracking functions for rand/rand bounded
2025-06-12 17:50:15 +02:00
Andrei Stoian
7986e0bf1d
chore(gpu): skip packing ks test if it needs more ram than available
2025-06-12 17:47:10 +02:00
Agnes Leroy
55179c52a7
chore(gpu): fix ci on H100
2025-06-12 16:22:57 +02:00
Emmanuel Ferdman
c103f0380c
fix(hpu): modernize logger interface
...
Signed-off-by: Emmanuel Ferdman <emmanuelferdman@gmail.com >
2025-06-12 09:04:31 +02:00
Nicolas Sarlin
8024753be0
fix(zk): test failed with trivial ct equal to 0
2025-06-11 18:40:32 +02:00
Nicolas Sarlin
506fdfbdd1
chore(zk): use Shake256 XoF instead of rand to generate gamma values
2025-06-11 18:03:12 +02:00
David Testé
a0e08b80e6
chore(ci): use gh cli instead of api calls done with curl
...
This refactor simplifies reading of data_pr_close workflow.
2025-06-11 16:55:37 +02:00
Enzo Di Maria
d06c3e3926
refactor(gpu): moving sub_assign_async to backend
2025-06-11 16:34:46 +02:00
Arthur Meyre
2bf9d25402
feat: add multi bit pbs 128 with GGSW preparation in FFT domain
2025-06-11 14:47:05 +02:00
Arthur Meyre
fe73f101cc
test: add test for an adapted parameter set for std multi bit pbs 128
2025-06-11 14:47:05 +02:00
Arthur Meyre
d12de58284
feat: add std pbs 128
2025-06-11 14:47:05 +02:00
Arthur Meyre
d4ea8cd85f
feat: add parallel conversion fo Fourier 128 for LweMultiBitBootstrappigKey
2025-06-11 14:47:05 +02:00
Arthur Meyre
64f2befd6c
feat: add Fourier 128 variant of LweMultiBitBootstrappingKey
2025-06-11 14:47:05 +02:00
Arthur Meyre
ce372dcea9
refactor(core): split the LWE multi bit bsk entity file
...
- to prepare for fft128 variant addition
2025-06-11 14:47:05 +02:00
Agnes Leroy
46b4958c9c
feat(gpu): add memory tracking functions for booleans
2025-06-11 13:32:06 +02:00
Agnes Leroy
b25bcbc607
feat(gpu): add mem tracking for eq/ne
2025-06-11 13:32:06 +02:00
Agnes Leroy
5dfacc7975
feat(gpu): add memory tracking for compression and decompression
2025-06-11 11:49:09 +02:00
Guillermo Oyarzun
3d857f62cc
refactor(gpu): return trivial indexes after ms noise reduction
2025-06-11 11:28:10 +02:00
Nicolas Sarlin
54c314cd71
chore(ci): always run zk tests
2025-06-11 10:29:53 +02:00
Nicolas Sarlin
38a9853140
chore(zk): check crs conformance in backward compat test
2025-06-11 10:29:53 +02:00
Nicolas Sarlin
360097d70e
chore(zk): use random seed in tests
2025-06-11 10:29:53 +02:00
Nicolas Sarlin
c94a76a85a
fix(zk): overflow in noise tests
2025-06-11 10:29:53 +02:00
Nicolas Sarlin
be1ade6dd2
chore(zk)!: use 8 bytes dsep and 128bits SID in hash functions
...
BREAKING_CHANGE:
- PublicParams::from_vec methods have been updated to take 8 bytes dsep and an
SID. CRS generated before this PR are still supported.
2025-06-11 10:29:53 +02:00
Pedro Alves
53845b298a
fix(gpu): fix the packing keyswitch buffer not being allocated on large parameter sets
2025-06-11 08:58:09 +02:00
David Testé
11c0340eca
chore(bench): plug server-side proof in zk benchmarks
2025-06-10 18:00:39 +02:00
Baptiste Roux
5e966a3d78
chore(hpu): changes based on code review
2025-06-10 17:43:35 +02:00
Baptiste Roux
443e02215f
feat(hpu): Add recent IOp in integer benchmarks
2025-06-10 17:43:35 +02:00
Baptiste Roux
3c632c06ba
chore(hpu): Fix/Changes to be compliant with CI
2025-06-10 17:43:35 +02:00
Baptiste Roux
833b593845
feat(hpu): Add support for Lead/Trail/Count/ilog2 in high-level API
2025-06-10 17:43:35 +02:00
JJ-hw
a20c90b090
feat(hpu): Add ILOG2/COUNT0/COUNT1/LEAD0/LEAD1/TRAIL0/TRAIL1 IOp.
...
Those IOp are tested within new bitcnt category
2025-06-10 17:43:35 +02:00
Baptiste Roux
71e86f0522
feat(hpu): Add support for Shift/Rotate in high-level API
...
Scalar version not supported yet
2025-06-10 17:43:35 +02:00
Baptiste Roux
cb45f7f429
feat(hpu): Add Rot/Shift IOp
...
Proper implementation of Scalar version need update in the firmware.
And thus it wasn't done yet.
2025-06-10 17:43:35 +02:00
Baptiste Roux
05a51d47fa
feat(hpu): Add check with Pbs gid definition
...
Currently it's a runtime check, but prevent attribution of same Gid for two != lut
2025-06-10 17:43:35 +02:00
Baptiste Roux
2eb1ccd128
feat(hpu): Add support for DivRem in high-level API
2025-06-10 17:43:35 +02:00
JJ-hw
b7a518b9ee
chore(hpu): Cleanup code following clippy advices
...
Also applied cargo fmt
2025-06-10 17:43:35 +02:00
JJ-hw
39faca219f
feat(hpu): Add modulo. Note: not optimized version. Use same algo as the division.
2025-06-10 17:43:35 +02:00
JJ-hw
fe7a8915bc
feat(hpu): Add DIV/Divs IOp support
...
Thoses IOp outputs quotient and remainder for numerator and divider of same size.
2025-06-10 17:43:35 +02:00
Baptiste Roux
96c8c44c71
feat(hpu): Enable some erc20 impl
...
With the support of overflowing ops, those impl are now available to Hpu
2025-06-10 17:43:35 +02:00
