Jens Steube
9a128ce667
Fix force_jit_compilation usage
2016-02-21 13:30:28 +01:00
Jens Steube
7190dcf855
Prepare NEW_SIMD_MODE for -a 1 kernels
2016-02-20 16:13:06 +01:00
Jens Steube
4338d71d62
Merge pull request #233 from Fist0urs/format_krb5tgs
...
Fix test.pl verify -m 13100
2016-02-20 12:47:01 +01:00
Fist0urs
233cdd009c
Fix test.pl verify
2016-02-20 01:39:58 +01:00
Jens Steube
990f973c48
Add hard-wired Device-Name for Tuning-Database which matches all Device-Types:
...
- DEVICE_TYPE_CPU
- DEVICE_TYPE_GPU
- DEVICE_TYPE_ACCELERATOR
Optimized IRIS GPU workaround settings
2016-02-20 00:58:22 +01:00
Jens Steube
b996858751
Merge pull request #232 from Fist0urs/format_krb5tgs
...
Add verification of checksum for -m 13100
2016-02-20 00:05:30 +01:00
Fist0urs
62bed36638
Add verification of checksum for -m 13100
...
This avoid collisions by arc4'ing all data then hmac-md5
when valid ASN1 structures headers are found.
Performance should not be impacted.
2016-02-19 23:12:46 +01:00
Jens Steube
96bde85378
Allow mixed settings auto and fixed values for parameter and inside tuningdb for loops and accel
2016-02-19 18:58:03 +01:00
Jens Steube
14fe45b1f0
Simplify steps for autotune
2016-02-18 14:44:44 +01:00
Jens Steube
a83622faaf
Fix final autotune balancing
2016-02-18 13:04:25 +01:00
Jens Steube
f05944395a
Move fake word generation to try_run() to avoid unwanted caching
2016-02-18 10:49:23 +01:00
Jens Steube
9ff7718f59
Merge pull request #230 from magnumripper/master
...
Support vector width 16
2016-02-18 09:02:01 +01:00
magnum
a5be8a75ed
Allow and support vector-width 16, which is current maximum for
...
OpenCL. Closes #226 .
2016-02-18 08:51:45 +01:00
magnum
998605ef89
Simplify code. No change in functionality at all.
2016-02-17 18:24:35 +01:00
Jens Steube
2790bc7baf
Merge pull request #229 from Fist0urs/format_krb5tgs
...
Cosmetic change -m 13100
2016-02-17 15:11:43 +01:00
Fist0urs
223ab0b3db
Cosmetic change -m 13100
...
Removed commented part (the commented part was there to implement full last arc4'ing + hmac-md5)
We will see if some people find collision. In this case we will add this last check
2016-02-17 13:32:07 +01:00
Jens Steube
814b18dd7e
Merge pull request #228 from philsmd/master
...
minor: some formatting fixes re: #225 (-m 13100 = Kerberos 5 TGS-REP etype 23)
2016-02-17 11:33:40 +01:00
philsmd
8298afa462
minor: some formatting fixes re: #225 (-m 13100 = Kerberos 5 TGS-REP etype 23)
2016-02-17 11:02:40 +01:00
Jens Steube
e6900d82f6
Add missing changes entry for -m 13100
2016-02-17 08:50:12 +01:00
Jens Steube
0891e39b62
Merge pull request #225 from Fist0urs/format_krb5tgs
...
Initial commit
2016-02-17 08:23:27 +01:00
Fist0urs
c3dabdd69e
Initial commit
2016-02-16 16:34:46 +01:00
jsteube
7645a6f00f
Fix warp bug on -m 8500
2016-02-16 13:24:36 +01:00
Jens Steube
138ea7ed11
Fix Truecrypt Whirlpool speed
2016-02-16 10:47:40 +01:00
Jens Steube
cc3e152f21
Fix old NV Titan OpenCL device name
2016-02-16 10:34:28 +01:00
Jens Steube
4bc02e5ac8
Replace OpenCL event's with own timers, this saves us a thread-pair spawned on nvidia's OpenCL runtime and doesn't affect others
2016-02-15 22:27:30 +01:00
Jens Steube
c44b50c2e2
Use CL_DEVICE_NATIVE_VECTOR_WIDTH_* instead of CL_DEVICE_PREFERRED_VECTOR_WIDTH_*
2016-02-15 21:19:41 +01:00
Jens Steube
1af0f9c67b
Fix autotune in -i mode
2016-02-15 21:08:16 +01:00
jsteube
bbadabe278
Fix another source of 0H/s in benchmark mode
2016-02-15 20:32:01 +01:00
Jens Steube
0cbe2347cd
Merge pull request #223 from gm4tr1x/clEnqueueNDRangeKernelMod
...
Get rid of hc_clEnqueueNDRangeKernel double call
2016-02-15 18:56:03 +01:00
Jens Steube
709f79a7da
Add option to repeat the kernel on the device NUM times to increase benchmark accuracy
2016-02-15 17:42:09 +01:00
Jens Steube
a5603086d5
Add autotune engine to changes
2016-02-15 15:09:22 +01:00
Jens Steube
eaf7a7cb2f
Final balance meassurements for autotune
2016-02-15 14:57:03 +01:00
Jens Steube
a9e3ef0576
In benchmark-mode, do not depend any longer on a fixed time, better use a single iteration instead
2016-02-15 12:38:54 +01:00
Jens Steube
4b241b42ea
Remove measurement section from tuning database and create a forum-thread or wiki-entry for it
2016-02-15 11:11:20 +01:00
Jens Steube
8af57d55f4
Disable GPU-Temp monitoring during benchmark
2016-02-14 20:15:14 +01:00
jsteube
f1e0cd6d56
Autotune: 200 loops should be fine for a start
2016-02-14 20:07:16 +01:00
Jens Steube
515385c57d
Add dedicated steps for loops and accel
2016-02-14 18:23:21 +01:00
Jens Steube
72e0553e44
Fixed a missing condition
2016-02-14 15:50:29 +01:00
Jens Steube
c09bc848f7
Autotuning engine prototype
2016-02-14 15:45:52 +01:00
Jens Steube
c0293928dd
Prepare for better autotuning capabilities
2016-02-13 16:07:58 +01:00
Gabriele 'matrix' Gristina
2a1d0d21ff
Get rid of hc_clEnqueueNDRangeKernel double call
2016-02-12 17:11:23 +01:00
Jens Steube
9ed2bcc9b9
Merge pull request #221 from gm4tr1x/8900tuning
...
Updated 8900 osx kernel accel
2016-02-12 16:35:16 +01:00
Jens Steube
a736c741a8
Merge pull request #220 from gm4tr1x/weak_hash_check
...
Fix bug in weak_hash_check(), wrong kernel_loops assigned
2016-02-12 16:35:09 +01:00
Gabriele 'matrix' Gristina
38ce0b7427
Updated 8900 osx kernel accel
2016-02-12 15:23:35 +01:00
Gabriele 'matrix' Gristina
c97dbf56cb
Fix bug in weak_hash_check(), wrong kernel_loops assigned
2016-02-12 14:04:13 +01:00
Jens Steube
e593c5fcd8
Update tuningdb help text
2016-02-11 14:26:18 +01:00
Jens Steube
2dfe9200d6
Remove Wordload-Profiles from tuningdb, we can simply calculate it
2016-02-11 14:16:08 +01:00
Jens Steube
02a6f2b4a6
Updated Workload finding strategy, should be easy to automate now
2016-02-11 13:38:59 +01:00
Jens Steube
3e25dc0a08
Updated Workload finding strategy, should be easy to automate now
2016-02-11 13:31:50 +01:00
Jens Steube
75fdf018b5
Merge pull request #219 from gm4tr1x/updateTuningDB
...
Added two aliases for osx
2016-02-11 13:01:45 +01:00
