Nicolas Sarlin
57cbab9fe1
chore(backward): integrate backward compat data
...
Code is taken from
59a6179831
2025-07-07 13:10:55 +02:00
Andrei Stoian
97ce0f6ecf
feat(gpu): update GPU documentation
2025-07-07 09:44:43 +02:00
Nicolas Sarlin
b6c21ef1fe
docs: describe noise squashed compression
2025-07-07 09:32:51 +02:00
Nicolas Sarlin
e599608831
chore(shortint): make decrypt_no_decode public
2025-07-07 09:30:14 +02:00
Arthur Meyre
f243491442
chore(docs): add features to the rust_configuration page
2025-07-04 17:06:15 +02:00
Arthur Meyre
b5248930a2
chore(docs): add handbook in explanation section
2025-07-04 17:06:15 +02:00
Arthur Meyre
2d280d98d2
chore(docs): add handbook in the security and cryptography section
2025-07-04 17:06:15 +02:00
Arthur Meyre
10b57f8a8e
chore(docs): add link to GPU and HPU backend docs in the installation page
2025-07-04 17:06:15 +02:00
Arthur Meyre
242df05eb2
chore(docs): add links to GPU and HPU backend on front page
2025-07-04 17:06:15 +02:00
Arthur Meyre
899d4a7750
docs: add noise squashing documentation
2025-07-04 16:08:25 +02:00
Agnes Leroy
48dfeb21dc
chore(gpu): refactor size tracker to avoid future bugs
2025-07-04 14:37:02 +01:00
Skylar Ray
a46ce3fb51
chore: fix typo in classic.rs
2025-07-04 13:33:15 +02:00
Arthur Meyre
192777bde6
chore(ci): handle unverified PRs to autoclose
2025-07-04 13:18:35 +02:00
Dmitry
3aa198311c
fix: broken GPU arg due to typo
2025-07-04 11:04:14 +01:00
David Testé
7034d4ceb4
doc(bench): update benchmark results tables
...
All the results are using parameters set with p-fail of 2**-128.
CPU tables using parameters set with p-fail 2**-64 are removed.
GPU tables for 1xH100 and 2xH100 are now replace with the new
hardware standard: 8xH100-SXM5.
HPU results are added to the backend comparison table and integrate
latest operations available.
2025-07-04 10:06:14 +02:00
Arthur Meyre
799ae92f59
chore: remove dead link from docs
2025-07-04 10:04:22 +02:00
Arthur Meyre
36e9371fdf
test: use hamming weight = 1/2 for core noise tests
...
- allows to have less variability and matches exactly what the noise
formulas expect for uniform binary secret keys
2025-07-04 09:55:35 +02:00
Pedro Alves
8c88678ee8
feat(gpu): implement 128-bit multi-bit PBS
2025-07-03 20:34:32 -03:00
leopardracer
e1beea5ecb
chore: Update test_user_docs.rs
2025-07-03 20:08:13 +02:00
Agnes Leroy
701411044b
chore(gpu): update SXM5 cost
2025-07-03 17:00:02 +01:00
JJ-hw
405fdec6b9
fix(hpu): Fix iop_propagate_msb_to_lsb_blockv: propagation in application was not done correctly
2025-07-03 14:31:59 +02:00
Agnes Leroy
b3355e2b2f
chore(gpu): remove template from sum ciphertexts, add two missing delete
2025-07-03 12:51:29 +01:00
Agnes Leroy
e4d856afdf
chore(gpu): update noise squashing parameters
2025-07-03 12:51:19 +01:00
Pedro Alves
22ddba7145
fix(gpu): refactor the (128-bit and regular) classical PBS entry point to remove the num_samples parameter
...
- fixes the throughput for those PBSs
- also fixes the throughput benchmark for regular PBSs
2025-07-03 08:23:09 -03:00
David Testé
d955696fe0
chore(bench): reduce number of bit sizes to benchmark
...
This is done to reduce execution time since 4 bits precision is not useful to measure.
2025-07-03 12:45:02 +02:00
Baptiste Roux
eb0b9643bb
fix(hpu): Fix clippy_hpu_mockup makefile entry
2025-07-03 10:28:52 +02:00
Arthur Meyre
d68305e984
chore: change link to point to the FHE.org discord for support
2025-07-03 10:28:10 +02:00
Enzo Di Maria
3d64316c66
refactor(gpu): moving signed_scalar_div_async and get_signed_scalar_div to the backend
2025-07-03 08:52:04 +01:00
Agnes Leroy
4bba35e926
chore(gpu): remove m3_c3 & gf 3 params from multi-gpu tests to reduce ci time
2025-07-02 17:18:26 +01:00
Baptiste Roux
187159d9f9
chore(hpu): bump backend version
2025-07-02 17:31:45 +02:00
Nicolas Sarlin
0cf9f9f3bd
chore(zk): bump tfhe-zk-pok to 0.7.0
2025-07-02 17:31:02 +02:00
tmontaigu
dcb6049441
chore: backward data test for CompressedSquashedNoiseCiphertextList
2025-07-02 16:51:05 +02:00
tmontaigu
7203cc3564
feat(hlapi): add CompressedSquashedNoiseCiphertextList
2025-07-02 16:51:05 +02:00
Agnes Leroy
b198c18498
chore(gpu): bump backend version
2025-07-02 15:34:10 +01:00
pgardratzama
916e6e6a61
chore(hpu): fix typo in comment of Event implementation
...
Co-authored-by: emmmm <155267286+eeemmmmmm@users.noreply.github.com >
2025-07-02 15:32:57 +02:00
pgardratzama
9ac776185a
doc(hpu): fix spelling issue in data_versioning.md
...
Co-authored-by: futreall <86553580+futreall@users.noreply.github.com >
2025-07-02 15:32:57 +02:00
pgardratzama
28e44ca237
doc(hpu): Fix link to FPGA repository in the README
...
Co-authored-by: MozirDmitriy <dmitriymozir@gmail.com >
2025-07-02 15:32:57 +02:00
Baptiste Roux
6432b98591
chore(mockup): Add clippy target for tfhe_hpu_mockup
...
Also fix all clippy lint
2025-07-02 14:41:41 +02:00
Helder Campos
15cce9f641
fix(hpu): Fixing the llt scheduler
...
In RTL simulations, it is possible that a very strange HPU with huge
amount of batches and very little registers is randomized. In this case,
if the scheduler was configured to fill the batch before flushing, it
would run out of registers. The solution is to force flush in this
scenario.
2025-07-02 14:41:41 +02:00
Baptiste Roux
5090e9152b
chore: Revert "chore: allow to not perform the half case correction for mean compensation"
...
This reverts commit 00ffa3efdc
2025-07-02 14:41:41 +02:00
Baptiste Roux
24572edb1c
feat(hpu): Add support for centered modswitch.
...
Add new field in HpuPBSParameters (log2_pfail and modulus_switch_type).
Also add new parameters set definition in shortint for benchmark matching.
Remove the used of use_mean_compensation register, this information is now embedded inside the parameters set definition.
Update psi64.hpu archive with newest bitstream
2025-07-02 14:41:41 +02:00
Helder Campos
303f67fe11
fix(hpu): Fixing the multiplication algorithm in LLT
...
It was failing before for nu > 5. Also corrected the initial degree
after the partial products, which decreases the number of PBSs to do
with nu > 5.
2025-07-02 14:41:41 +02:00
Arthur Meyre
86a40bcea9
chore: move gated import to section with feature gate in HL erc20 bench
2025-07-02 13:14:31 +02:00
Agnes Leroy
97c0290ff7
fix(gpu): revert avoid copy to host in sum ciphertexts
...
This reverts commit 2b57fc7bd8
2025-07-02 08:30:12 +01:00
Agnes Leroy
3ba6a72166
chore(gpu): move sum ctxt lut allocation to host to save memory
2025-07-02 08:30:12 +01:00
tmontaigu
dbd158c641
feat(integer): add CompressedSquashedNoiseCiphertextList
2025-07-02 08:51:26 +02:00
Nicolas Sarlin
0a738c368a
chore(backward): update backward data repo branch
2025-07-01 14:18:10 +02:00
Arthur Meyre
4325da72cf
chore: allow to not perform the half case correction for mean compensation
2025-07-01 14:18:10 +02:00
Mayeul@Zama
e1620d4087
feat(shortint): add support for centered modulus switch in parameters
2025-07-01 14:18:10 +02:00
Mayeul@Zama
6805778cb8
feat: add centered modulus switch
2025-07-01 14:18:10 +02:00
