Commit Graph

372 Commits

Author SHA1 Message Date
youben11
f4166a4973 docs: use consistent style for comment blocks
prefix comment blocks with ///
2022-07-07 16:11:19 +01:00
Antoniu Pop
261e59d2b4 feat(dfr): add more operations to be converted to tasks. 2022-06-28 17:11:12 +01:00
Antoniu Pop
26084a68aa fix(compiler): delay all memref deallocation calls introduced by the bufferizer and that are made into futures until after the synchronization point. 2022-06-28 17:11:12 +01:00
Antoniu Pop
fb219f0fbf feat(runtime): increase the limit of task input parameters to 16. 2022-06-28 17:11:12 +01:00
Umut
b3a2671dc7 feat: implement all kinds of subtractions 2022-06-28 11:13:03 +03:00
Umut
8f8a57d220 feat: implement MANP calculation for tensor.insert op 2022-06-28 11:13:03 +03:00
Mayeul@Zama
bcc5c750a1 fix(client): fix assumption that glweDimension = 1 2022-06-28 09:24:39 +02:00
Quentin Bourgerie
2de76e9c4e fix(compiler): Fix tfhe global parametrization to handle k>1
Co-authored-by: Mayeul@Zama <mayeul.debellabre@zama.ai>
2022-06-24 15:35:51 +02:00
Quentin Bourgerie
ef9d11c16f cleanup(compiler): Remove dead code and usless pass 2022-06-24 11:00:08 +02:00
Quentin Bourgerie
ab64532b7b cleanup(compiler): Remove old bufferization passes 2022-06-24 11:00:08 +02:00
Quentin Bourgerie
9baee39351 cleanup(compiler): Cleanup comments on ConcreteToBConcrete 2022-06-24 11:00:08 +02:00
Quentin Bourgerie
acbb5bf17a cleanup(compiler): Remove custom print and parse TFHE types 2022-06-24 11:00:08 +02:00
Quentin Bourgerie
9dd776533a cleanup(compiler): Remove useless concrete types, simplify print and parse, and remove BConcreteToBConcreteCAPI pass 2022-06-24 11:00:08 +02:00
Mayeul@Zama
928a44d243 chore(ci): enforce ending files with a newline 2022-06-16 17:04:25 +02:00
Andi Drebes
3b8ea630a1 refactor(compiler): Rename SubIntGLWEOp to SubGLWEIntOp
Rename `SubIntGLWEOp` to `SubGLWEIntOp` for a consistent naming scheme
of all TFHE ops.
2022-06-14 14:36:28 +02:00
Andi Drebes
45577fb79e Rebase onto llvm-project f69328049e9e with local changes
This commit rebases the compiler onto commit f69328049e9e from
llvm-project.

Changes:

* Use of the one-shot bufferizer for improved memory management

* A new pass `OneShotBufferizeDPSWrapper` that converts functions
  returning tensors to destination-passing-style as required by the
  one-shot bufferizer

* A new pass `LinalgGenericOpWithTensorsToLoopsPass` that converts
  `linalg.generic` operations with value semantics to loop nests

* Rebase onto a fork of llvm-project at f69328049e9e with local
  modifications to enable bufferization of `linalg.generic` operations
  with value semantics

* Workaround for the absence of type propagation after type conversion
  via extra patterns in all dialect conversion passes

* Printer, parser and verifier definitions moved from inline
  declarations in ODS to the respective source files as required by
  upstream changes

* New tests for functions with a large number of inputs

* Increase the number of allowed task inputs as required by new tests

* Use upstream function `mlir_configure_python_dev_packages()` to
  locate Python development files for compatibility with various CMake
  versions

Co-authored-by: Quentin Bourgerie <quentin.bourgerie@zama.ai>
Co-authored-by: Ayoub Benaissa <ayoub.benaissa@zama.ai>
Co-authored-by: Antoniu Pop <antoniu.pop@zama.ai>
2022-06-14 14:35:25 +02:00
youben11
670af02112 fix: remove subinteint folder
it actually requires to negate the result which can't be done via
standard folder, so we remove it can cause erroneous computation
2022-06-14 10:00:01 +02:00
youben11
1354759884 chore: rename refs to main branch 2022-06-09 09:19:06 +01:00
youben11
5636342097 feat: add option to enable/disable optimization 2022-06-01 14:35:12 +01:00
youben11
8d0f20390c feat: optimize concrete mul with constant values 2022-06-01 14:35:12 +01:00
Quentin Bourgerie
dd28a7f132 fix(compiler): Concrete.ZeroOp is a 0 value tensor in BConcrete Dialect 2022-05-30 14:22:32 +01:00
youben11
8d04c1e4af feat: fold tensor ops with constant values
add/sub with constant tensors of 0s
mul with constant tensors of 1s
2022-05-30 08:35:14 +01:00
youben11
8aa6f3e809 feat: fold add/sub/mul with specific constant values
Remove add/sub with constant zero values
Remove mul with constant one values
2022-05-30 08:35:14 +01:00
Umut
b052157fae refactor: separate runtime context from public arguments 2022-05-30 10:19:14 +03:00
Quentin Bourgerie
ae9a04cd56 fix(compiler): Use the absolute value when computing the square of a constant in MANP analysis
The computation of the norm2 should take care of the sign of the constant to compute the square of this constant.
2022-05-24 14:12:28 +02:00
Quentin Bourgerie
6532c8f449 fix(compiler): Canonicalize before MANP analysis 2022-05-24 14:12:28 +02:00
Quentin Bourgerie
6bb1e05658 fix(compiler): Compiler // options are badly setted 2022-05-23 13:53:23 +02:00
Mayeul@Zama
f5faf4c421 feat(optimizer): update concrete_optimizer dependency
fix breaking changes
2022-05-18 14:04:48 +02:00
rudy
6f02d1248d feat(optimizer): update to use the new cpp brige 2022-05-10 15:36:40 +02:00
youben11
5f1a539505 feat(python): get path to diff artifacts
- path to client parameters file
- path to shared library
2022-05-09 16:18:21 +01:00
youben11
211241fcb2 refactor: store compil artifacts under a single dir 2022-05-09 16:18:21 +01:00
youben11
f223f02ab7 feat: parameterize artifact generation in lib compilation 2022-05-09 16:18:21 +01:00
rudy
1b70a38754 feat(optimizer): expose the p_error parameter
simplify the handling of options by relying more on CompilationOptions
2022-05-06 13:06:08 +02:00
rudy
eb1f5b1842 fix(keysetcache): use a portable hash function
will help to make macos faster on the ci
2022-05-06 10:00:11 +02:00
rudy
8c25a97133 fix(KeySetCache): remove a race condition
thanks to Umut for pointing it
2022-05-04 08:42:57 +02:00
rudy
b45e469a44 feat(ci): macos, fix keysetcache use 2022-05-02 17:20:22 +02:00
youben11
843dd0eb5b feat(python): seriliaze client parameters 2022-05-02 10:16:39 +02:00
rudy
923a1b58e1 feat(optimizer): plug online parameter estimation 2022-04-29 15:11:42 +02:00
youben11
e9345b3859 feat: serialize public result w/ python bindings 2022-04-22 15:53:26 +01:00
youben11
0a5881096c feat(python): serialize public arguments 2022-04-15 15:25:25 +01:00
youben11
d9d35c1023 refactor: use unique_ptr consistently for PublicArguments 2022-04-15 15:25:25 +01:00
rudy
1fca8f4b91 feat(KeySetCache): detect corrupted keysetcache 2022-04-15 09:13:55 +02:00
youben11
c8ca318fde feat: throw clear error if DF jit-exec is not supported
was previously getting an undefined symbol error as the runtime lib
didn't support dataflow execution
2022-04-14 11:07:39 +01:00
Quentin Bourgerie
247d60503d fix(compiler): Custom copy op from 1D tensor to avoir stack allocation from mlir memref to llvm lowering 2022-04-13 10:16:53 +02:00
youben11
af300055a7 fix(macos): update dependency after compiling dylib
when dellocate is used to include dependencies in python wheels, the
runtime library will have an id that is prefixed with /DLC, and that path
doesn't exist. So when generated libraries won't be able to find it
during load time. To solve this, we change the dep in the generated
library to be relative to the rpath which should be set correctly during
linking. This shouldn't have an impact when /DLC/concrete/.dylibs/* isn't
a dependecy in the first place (when not using python).

also set rpath when linking to RT lib
2022-04-12 13:56:18 +01:00
youben11
d7440f15c0 fix(python): lookup RTLib at a diff location in Mac 2022-04-12 13:56:18 +01:00
Quentin Bourgerie
35d8489b9c fix(compiler/TFHE): Introduce bootstrap and keyswitch at the TFHE level 2022-04-12 09:37:56 +02:00
Andi Drebes
7cc0c01326 fix(compiler): Do not use C-style casts in multi_arity_call_dynamic_rank
C-style casts between function pointers result in an error using g++
8.3.0. This patch introduces a workaround based on `reinterpret_cast`,
in which the original function pointer is first cast into an unsigned
integer of sufficient size and then cast into the target function
pointer type.

TODO: Check that this is actually valid C++ with defined behavior on
all implementations / platforms.
2022-04-11 17:24:23 +02:00
Andi Drebes
e83c421bce enhance(compiler): genDynamicRankCall.py: Match clang-format formatting rules 2022-04-11 17:24:23 +02:00
Andi Drebes
b6cec58549 fix(compiler): genDynamicRankCall.py use renamed tensorDataFromMemRef 2022-04-11 17:24:23 +02:00