ea3c80c949
Move Rijndael_AdvancedProcessBlocks_ARMV8 into anonymous namespace
2017-09-23 05:28:59 -04:00
26597059d9
Move to anonymous namespaces in rijndael-simd.cpp
2017-09-23 02:13:16 -04:00
12953fd0e4
Add IncrementPointerAndStore
...
This speeds up XL C/C++ by 0.1 to 0.2 cpb
2017-09-22 20:35:18 -04:00
1057f89363
Move Power8 crypto functions into ppc-crypto.h
2017-09-22 05:23:29 -04:00
3e55817819
Add C++ templates for additional Vector ops
...
Removed lower-level C-like functions such as Store8x16 and Store64x2
2017-09-22 04:15:33 -04:00
441e944a66
Switch to vec_vsx_ld, remove unaligned loads
...
Partially unroll loop Rijndael_UncheckedSetKey_POWER8 loop. It saves about another 60 cycles
2017-09-22 02:53:08 -04:00
d9592a303c
Updated comments
2017-09-21 21:45:23 -04:00
dabad4b409
Cleanup asserts and casts
2017-09-21 20:55:35 -04:00
1edea5a80f
Vectorize tail of Rijndael_UncheckedSetKey_POWER8
2017-09-21 20:02:40 -04:00
e43c0eee74
Fold ConditionalByteReverse for non-Power8 paths
2017-09-21 19:17:42 -04:00
f763bf3da6
Updated comments
2017-09-21 12:08:54 -04:00
e78464a1af
Enable little endian Rijndael_UncheckedSetKey_POWER8 using built-ins
...
The problem was vec_sld is endian sensitive. The built-in required more than us setting up arguments to ensure the vsx load resulted in a big endian value. Thanks to Paul R on Stack Overflow for sharing the information that IBM did not provide. Also see http://stackoverflow.com/q/46341923/608639
2017-09-21 09:56:37 -04:00
c6b096ddd4
Move Rijndael_UncheckedSetKey_POWER8 prior to GetUserKey call
...
Arg... GetUserKey was performing a 32-bit word reverse. It was part of the problem on little endian machines
2017-09-21 01:08:44 -04:00
9fd5d023f9
Load r5 mask once for key expansion
2017-09-20 20:27:58 -04:00
35c0fa82fd
Use <time.h> for Borland/Embarcadero (GH #512 )
2017-09-20 18:10:07 -04:00
c5a427d690
Add PowerPC VectorLoadKeyUnaligned for AES-192
...
Make internal functions static. We get better optimizations depsice using unnamed namespaces
Add PowerPC uint32x4 functions for handling 32-bit rcon and mask
2017-09-20 08:57:53 -04:00
c94d076aa1
Move r1 write to caller; remove from Rijndael_Subkey_POWER8
...
Signed-off-by: Jeffrey Walton <noloader@gmail.com>
2017-09-20 04:38:53 -04:00
5159d0803d
Add Power8 key expansion for big endian
...
This is AES-128 key expansion for big endian. Little endian has a bug in it so it can't be enabled at the moment. GDB is acting up on GCC112, so I've had trouble investigating it
2017-09-20 03:34:54 -04:00
6102333fc3
Add CRYPTOPP_NO_CPU_FEATURE_PROBES (GH #511 )
...
We determine machine capabilities by performing an os/platform *query* first, like getauxv(). If the *query* fails, we move onto a cpu *probe*. The cpu *probe* tries to exeute an instruction and then catches a SIGILL on Linux or the exception EXCEPTION_ILLEGAL_INSTRUCTION on Windows. Some OSes fail to hangle a SIGILL gracefully, like Apple OSes. Apple machines corrupt memory and variables around the probe.
2017-09-19 21:08:37 -04:00
6440921723
Add Rijndael_UncheckedSetKey_POWER8
...
We are going to attempt to perform key setup using Power8 in-core vector instructions
2017-09-19 04:55:15 -04:00
923cf95571
ByteReverseArray → ReverseByteArrayLE
2017-09-18 18:40:19 -04:00
2c18fe8af8
Refactor LoadT() and StoreT(). Add separate ReverseT() for little endian machines
...
The refactoring has no effect on little endian machines. However, on big endian GCC119 using GCC 7.1 the performance improved by 2.5x for ECB and CTR modes:
BEFORE:
<TR><TH>AES/CTR (128-bit key)<TD>2723<TD>1.4<TD>0.163<TD>670
<TR><TH>AES/CTR (192-bit key)<TD>2560<TD>1.5<TD>0.175<TD>719
<TR><TH>AES/CTR (256-bit key)<TD>2728<TD>1.4<TD>0.183<TD>749
<TR><TH>AES/CBC (128-bit key)<TD>1204<TD>3.2<TD>0.135<TD>554
<TR><TH>AES/CBC (192-bit key)<TD>1066<TD>3.7<TD>0.148<TD>605
<TR><TH>AES/CBC (256-bit key)<TD>948<TD>4.1<TD>0.155<TD>635
<TR><TH>AES/OFB (128-bit key)<TD>1019<TD>3.8<TD>0.158<TD>648
<TR><TH>AES/CFB (128-bit key)<TD>949<TD>4.1<TD>0.192<TD>787
<TR><TH>AES/ECB (128-bit key)<TD>3564<TD>1.1<TD>0.082<TD>337
AFTER:
<TR><TH>AES/CTR (128-bit key)<TD>6484<TD>0.6<TD>0.163<TD>677
<TR><TH>AES/CTR (192-bit key)<TD>5641<TD>0.7<TD>0.176<TD>728
<TR><TH>AES/CTR (256-bit key)<TD>5005<TD>0.8<TD>0.183<TD>761
<TR><TH>AES/CBC (128-bit key)<TD>1223<TD>3.2<TD>0.135<TD>559
<TR><TH>AES/CBC (192-bit key)<TD>1080<TD>3.7<TD>0.147<TD>611
<TR><TH>AES/CBC (256-bit key)<TD>966<TD>4.1<TD>0.155<TD>642
<TR><TH>AES/OFB (128-bit key)<TD>1057<TD>3.7<TD>0.158<TD>656
<TR><TH>AES/CFB (128-bit key)<TD>1217<TD>3.3<TD>0.186<TD>774
<TR><TH>AES/ECB (128-bit key)<TD>7289<TD>0.5<TD>0.082<TD>342
2017-09-18 18:15:25 -04:00
f0c2324f6b
Fix armeabi and armv7-a for Android (GH #509 )
2017-09-17 20:07:53 -04:00
adea69ab68
Avoid increment during stores of 6x blocks
...
This provides another 0.1 cpb with GCC
2017-09-14 21:06:44 -04:00
25efb7a140
Use 6x blocks for ARMv8 AES rather than 4x
...
We gain 0.1 to 0.3 cpb, depending on the mode
2017-09-14 20:32:06 -04:00
58890ff053
Use 6x blocks for Power8 AES rather than 4x
...
Perforamnce increased for all modes when performing 6x vs 4x. 8x and 12x performed worse.
Here are the numbers:
4x Blocks:
<TR><TH>AES/CTR (128-bit key)<TD>1563<TD>2.1<TD>0.409<TD>1392
<TR><TH>AES/CTR (192-bit key)<TD>1403<TD>2.3<TD>0.450<TD>1529
<TR><TH>AES/CTR (256-bit key)<TD>1280<TD>2.5<TD>0.482<TD>1639
<TR><TH>AES/CBC (128-bit key)<TD>582<TD>5.6<TD>0.359<TD>1222
<TR><TH>AES/CBC (192-bit key)<TD>517<TD>6.3<TD>0.394<TD>1339
<TR><TH>AES/CBC (256-bit key)<TD>474<TD>6.8<TD>0.432<TD>1469
<TR><TH>AES/OFB (128-bit key)<TD>533<TD>6.1<TD>0.402<TD>1368
<TR><TH>AES/CFB (128-bit key)<TD>563<TD>5.8<TD>0.461<TD>1568
<TR><TH>AES/ECB (128-bit key)<TD>1829<TD>1.8<TD>0.240<TD>817
6x Blocks:
<TR><TH>AES/CTR (128-bit key)<TD>1750<TD>1.7<TD>0.406<TD>1300
<TR><TH>AES/CTR (192-bit key)<TD>1638<TD>1.9<TD>0.447<TD>1432
<TR><TH>AES/CTR (256-bit key)<TD>1528<TD>2.0<TD>0.482<TD>1541
<TR><TH>AES/CBC (128-bit key)<TD>582<TD>5.2<TD>0.358<TD>1145
<TR><TH>AES/CBC (192-bit key)<TD>517<TD>5.9<TD>0.394<TD>1260
<TR><TH>AES/CBC (256-bit key)<TD>474<TD>6.4<TD>0.431<TD>1379
<TR><TH>AES/OFB (128-bit key)<TD>533<TD>5.7<TD>0.400<TD>1281
<TR><TH>AES/CFB (128-bit key)<TD>563<TD>5.4<TD>0.461<TD>1476
<TR><TH>AES/ECB (128-bit key)<TD>1950<TD>1.6<TD>0.238<TD>763
2017-09-14 16:07:21 -04:00
08e4ee422e
Avoid increment during stores of 4x blocks
...
This provides another 0.1 cpb with GCC
2017-09-14 15:12:07 -04:00
ddeae859d0
Use vec_xl_be and vec_xst_be for IBM XL C/C++ compiler
2017-09-14 13:27:49 -04:00
63a0af4efa
Fix endianess for s_one on ARM big-endian
2017-09-13 22:52:29 -04:00
8e52ce6dd2
Load correct value fo 1 under ARM big endian
2017-09-13 21:42:15 -04:00
c22507e38b
Clear unused variable warnings under Clang
2017-09-13 21:37:55 -04:00
8d98417306
Add Aarch64 specific defines to Android cross-compile
...
Move <arm_acle.h> logic into "sonfig.h". Detecting when we can/should include <arm_acle.h> is proving to be troublesome
2017-09-13 17:16:57 -04:00
397ccd7e49
remove commented code for Power8
2017-09-13 03:59:25 -04:00
2b24f5b9fe
VectorLoadAligned → VectorLoadKey
...
Add comments for the Load and Store functions
2017-09-12 20:38:58 -04:00
5659acb704
Cleanup vector casts
2017-09-12 19:44:34 -04:00
6899d3f8bb
Add AdvancedProcessBlocks for Power8
...
This increases performance to about 1.6 cpb. We are about 0.5 cpb behind Botan, and about 1.0 cpb behind OpenSSL. However, it beats the snot out of C/C++, which runs at 20 to 30 cpb
2017-09-12 18:15:55 -04:00
b090e5f69f
Add Power8 AES decryption
2017-09-12 05:53:17 -04:00
7fb34e9b08
Add Power8 AES encryption
...
This is the forward direction on encryption only. Crypto++ uses the "Equivalent Inverse Cipher" (FIPS-197, Section 5.3.5, p.23), and it is not compatible with IBM hardware. The library library will need to re-work the decryption key scheduling routines. (We may be able to work around it another way, but I have not investigated it).
2017-09-11 22:52:22 -04:00
37e02f9e0e
Revert AltiVec and Power8 commits
...
The strategy of "cleanup under-aligned buffers" is not scaling well. Corner cases are still turing up. The library has some corner-case breaks, like old 32-bit Intels. And it still has not solved the AltiVec and Power8 alignment problems.
For now we are backing out the changes and investigating other strategies
2017-09-05 16:28:00 -04:00
75aef9bded
Fixup under-aligned buffers when using AES on AltiVec and Power8
...
This commit supports the upcoming AltiVec and Power8 processor. This commit affects a number of classes due to the ubiquitous use of AES. The commit provides the data alignment requirements.
2017-09-04 11:21:47 -04:00
5c6a32ba0f
Support Base Implementation + SIMD implementation on Solaris (PR #461 )
2017-08-24 19:17:21 -04:00
a19f0c663b
Update asserts
...
Change 'rounds' to size_t in Rijndael_AdvancedProcessBlocks_ARMV8
2017-08-19 01:55:20 -04:00
a1b3102eab
Update comments
2017-08-19 01:35:36 -04:00
51fe8a7776
Guard use of SIGILL probes on Apple platforms
2017-08-17 18:06:57 -04:00
e2c377effd
Split source files to support Base Implementation + SIMD implementation (GH #461 )
...
Split source files to support Base Implementation + SIMD implementation
2017-08-17 12:33:43 -04:00
Jeffrey Walton