Stephen Heumann
6c1d703b84
Don't duplicate identical code for the last three schedule computation phases.
...
This shrinks code size and is probably faster with accelerators due to caching.
2017-06-30 12:49:51 -05:00
Stephen Heumann
d97e0c0153
Don't update index value in each unrolled loop iteration.
2017-06-30 00:52:25 -05:00
Stephen Heumann
2d8c85bd98
Unroll SHA-1 loop with variables substituted to avoid cycling values around.
...
SHA-1 has five state variables (a,b,c,d,e), and each iteration of the core loop cycles their values around (e.g. a->b), in some cases with modifications. By unrolling the loop by a factor of five and appropriately substituting the values operated on in each iteration, we can avoid actually doing the copies implied by this cycling, which gives a fairly significant performance gain.
2017-06-30 00:33:16 -05:00
Stephen Heumann
aab47e38e7
Compute SHA-1 message schedule 20 elements at a time, reusing storage.
...
The reduces the storage needed and boosts performance, since the whole schedule is contained in the direct page.
2017-06-29 22:22:06 -05:00
Stephen Heumann
7182fc5817
Use separate loops for the four parts of the SHA-1 computation.
2017-06-29 20:51:36 -05:00
Stephen Heumann
dfdb3186fd
Add macros for rotates.
2017-06-29 20:41:49 -05:00
Stephen Heumann
ad4e0a0dd5
More SHA-1 optimizations.
2017-06-29 20:16:22 -05:00
Stephen Heumann
13654c0888
Small optimizations in SHA-1 computation.
2017-06-29 19:52:52 -05:00
Stephen Heumann
a519776847
Add implementation of the basic block-processing function for SHA-1.
2017-06-29 16:54:49 -05:00
Stephen Heumann
1cb8841a46
state -> context
2017-06-29 16:52:34 -05:00
Stephen Heumann
f60585cc3e
Add note about context size.
2017-06-28 01:26:24 -05:00
Stephen Heumann
d4baf7dbe4
Usage message and indentation cleanups in test program.
2017-06-27 23:42:20 -05:00
Stephen Heumann
4ffd064204
Add comments and copyright notices.
2017-06-27 23:30:58 -05:00
Stephen Heumann
e46264f0c4
Small decryption optimization, saving a value in a register between rounds.
2017-06-27 22:35:57 -05:00
Stephen Heumann
634ff1e865
Optimize decryption code.
...
AddRoundKey is integrated into InvMixColumn and InvFinalRoundStep, rather than being done separately.
2017-06-27 20:44:14 -05:00
Stephen Heumann
18dabeb621
Allow number of test iters to be given on command line.
2017-06-27 20:07:07 -05:00
Stephen Heumann
d9b8467bb7
Rename some interfaces.
2017-06-27 19:42:53 -05:00
Stephen Heumann
14fe68de79
Time decryption as well as encryption.
2017-06-27 19:35:38 -05:00
Stephen Heumann
a522ab6d88
Optimize decryption a bit.
2017-06-27 19:24:06 -05:00
Stephen Heumann
46fcb3badc
Add support for AES-256 decryption.
2017-06-27 18:35:04 -05:00
Stephen Heumann
0e3e511ba8
Add basic decryption implementation for AES-128 and AES-192.
2017-06-27 18:17:00 -05:00
Stephen Heumann
9d9f62444d
Add timing test.
2017-06-27 12:45:58 -05:00
Stephen Heumann
4314644e75
Update to also support AES-192 and AES-256.
2017-06-27 01:36:56 -05:00
Stephen Heumann
293828584f
Optimize key expansion a bit, removing unnecessary loads.
2017-06-26 22:22:55 -05:00
Stephen Heumann
f04000c83d
Save a few cycles by not reloading the last written byte between rounds.
2017-06-26 21:49:16 -05:00
Stephen Heumann
75aac0daa9
Save a couple values on the stack rather than looking them up again.
...
This is cycle-neutral (assuming a page-aligned DP), but it reduces instruction bytes and therefore may give a speedup on accelerators with caches.
2017-06-26 21:13:25 -05:00
Stephen Heumann
301021e75e
Cut out a few instructions by reusing subexpressions in registers.
2017-06-26 20:53:17 -05:00
Stephen Heumann
3df0080371
Initial working implementation of AES-128 encryption.
2017-06-26 20:41:44 -05:00
