2641 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Frank Denis	1c2ce89326	_xgetbv() can be defined as a macro, so try an actual compilation	2017-02-28 18:20:24 +01:00
Frank Denis	e5a196a8ee	Always include intrin.h if available	2017-02-28 17:58:50 +01:00
Frank Denis	9eea164007	Check for _xgetbv() presence	2017-02-28 17:51:18 +01:00
Frank Denis	f23d7eff4d	Automatically generate MSVC projects	2017-02-28 17:33:57 +01:00
Frank Denis	bf9b322334	_mm_set_epi64x() emulation for MSVC < 2015	2017-02-28 16:27:45 +01:00
Frank Denis	72426ed9c0	Always include <intrin.h> on Visual Studio x86/x86_64	2017-02-28 15:28:59 +01:00
Frank Denis	88cc2c6ba5	scrypt_common: check for sse2 support the same way as everywhere else	2017-02-28 15:08:40 +01:00
Frank Denis	9904e95bde	salsa_ref: properly initialize variables when NULL is used as constants	2017-02-27 17:06:14 +01:00
Frank Denis	8398b47573	Remove unused variable	2017-02-27 17:04:31 +01:00
Frank Denis	fa89071261	Manual unrolling	2017-02-27 17:03:03 +01:00
Frank Denis	c6460c9e9a	Remove unreferenced variables	2017-02-27 16:57:37 +01:00
Frank Denis	03f27b4204	Use _xgetbv() on Visual Studio, when available	2017-02-27 16:54:20 +01:00
Frank Denis	4fea4d917b	With MSVC, AVX2 cause ICEs on x86 targets	2017-02-27 16:34:41 +01:00
Frank Denis	0e4a05d558	MSVC can only assembly x86 instructions	2017-02-27 16:31:46 +01:00
Frank Denis	3cd883d33e	Remove debugging leftover in tests	2017-02-27 16:25:32 +01:00
Frank Denis	47bc8df460	Avoid unsupport pragmas	2017-02-27 16:07:57 +01:00
Frank Denis	f6a032f3e6	Regen the top-level msvc project	2017-02-27 16:02:59 +01:00
Frank Denis	cc55a084f8	MSC_VER -> _MSC_VER typo	2017-02-27 09:15:37 +01:00
Frank Denis	e05d2b3d8f	C++ compat	2017-02-27 00:36:59 +01:00
Frank Denis	17b26a6e01	Missing \	2017-02-27 00:30:30 +01:00
Frank Denis	9c699ecdd4	Don't ask too much from automake	2017-02-27 00:22:35 +01:00
Frank Denis	14bebe460b	Allow compilation without a working x86_64 assembler	2017-02-27 00:19:42 +01:00
Frank Denis	e3b9907429	Use the assembly version of salsa20_xmm6 by default, if possible ... icc produces good code from the intrinsics-based translation, clang produces okay code, but gcc doesn't perform very well ATM. It's a bummer to have a 3rd implementation, but salsa20 is used quite a lot in the library, so it deserves a special attention. If the assembly code cannot be assembled, fall back to the reduced version of the intrinsics-based translation. So, in the final library, we always only get two implementations at most.	2017-02-27 00:09:48 +01:00
Frank Denis	7d29c0fbd7	Remove useless prototypes	2017-02-26 23:51:58 +01:00
Frank Denis	3db624d335	Bring the asm amd64_xmm6 implementation back	2017-02-26 23:36:54 +01:00
Frank Denis	606f569c7b	Replace the SSSE3 ChaCha20 impl with one derived from the AVX2 impl	2017-02-26 23:02:45 +01:00
Frank Denis	f60e4ae514	Capitalize macro params	2017-02-26 22:48:49 +01:00
Frank Denis	9d706253ff	+ AVX2 ChaCha20 implementation	2017-02-26 22:46:34 +01:00
Frank Denis	9dc21def26	Move declarations, for consistency	2017-02-26 22:18:09 +01:00
Frank Denis	8d256f0edd	With the full counter, the limit is comfortable enough (2^70)	2017-02-26 22:04:50 +01:00
Frank Denis	d155bfd686	Remove useless brackets	2017-02-26 21:59:52 +01:00
Frank Denis	17ebc9cb07	Import SSE2 & AVX2 salsa20 implementations	2017-02-26 21:32:00 +01:00
Frank Denis	c0c645da45	Test crypto_kdf constants, add _PRIMITIVE	2017-02-26 21:30:40 +01:00
Frank Denis	8a213f021c	Remove useless prototypes	2017-02-26 21:26:14 +01:00
Frank Denis	15f1904f00	Use the AVX2 salsa20 implementation if the CPU supports it	2017-02-26 21:23:39 +01:00
Frank Denis	d85454c3c7	No // comments	2017-02-26 21:23:17 +01:00
Frank Denis	5088ca52f6	Use <impl>/<primitive_name>*.[ch], not <impl>/<operation>_<primitive_name>*	2017-02-26 20:54:08 +01:00
Frank Denis	5ccf9c2f46	salsa20-xmm6int: don't hardcode the number of rounds	2017-02-26 19:04:34 +01:00
Frank Denis	78330822cb	Save trees	2017-02-26 19:02:23 +01:00
Frank Denis	294007a16f	Import u0.h	2017-02-26 19:00:01 +01:00
Frank Denis	e0ee69874f	xmm6int: optimize the last block	2017-02-26 18:56:27 +01:00
Frank Denis	45160a0f78	Define the HAVE_* macros for SIMD instructions on Visual Studio ... This avoids a lot of redundant preprocessor checks	2017-02-26 18:22:14 +01:00
Frank Denis	d781ec3cdf	Add missing header for _crypto_stream_salsa20_pick_best_implementation	2017-02-26 17:50:11 +01:00
Frank Denis	e08843d138	Enable the xmm6int implementation on Visual Studio, too	2017-02-26 17:48:39 +01:00
Frank Denis	0a4418a25f	Compile salsa20-xmm6int only if HAVE_EMMINTRIN_H is defined ... This is redundant with the next check, and needs to be adjusted for Visual Studio	2017-02-26 17:41:03 +01:00
Frank Denis	e4a43d5cff	Nits	2017-02-26 17:35:30 +01:00
Frank Denis	ed8277a2f6	Use a single way for stream_salsa20_ref conditional inclusion	2017-02-26 17:27:28 +01:00
Frank Denis	47a3461793	xor_salsa2012.c is not used any more	2017-02-26 17:21:13 +01:00
Frank Denis	199aa0efa2	Update AUTHORS	2017-02-26 17:21:04 +01:00
Frank Denis	3b3613d464	Add missing header files to the distfiles	2017-02-26 17:17:00 +01:00