amd/blis
1
0
Fork 0
mirror of https://github.com/amd/blis.git synced 2026-05-13 18:52:14 +00:00
Code Issues Packages Projects Releases Wiki Activity
1,886 Commits 10 Branches 88 Tags
810e90ee806510c57504f0cf8eeaf608d38bd9dd
Commit Graph

1886 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Devin Matthews
8945a1512d This version gets ~1550 GFLOPs on KNL wuth 16x4. 2016-08-03 11:28:24 -05:00
praveeng
cdfb3c3f29 Merge master code as on 2016_07_29 to amd-staging branch by praveeng 
Change-Id: Ic78b84d8b8d10158fb2a612f9a64bbc7b1f9b486
 2016-07-29 12:46:21 +05:30
praveeng
4bc842ca3a Merge branch 'master' of publicrepo 2016-07-28 17:32:12 +05:30
praveeng
117f883851 Revert commits 357c990bdd 
Change-Id: I12a34456d7eed93fda4369e76bcddb42ba7ccb99
 2016-07-28 17:30:53 +05:30
praveeng
2fcdc28f10 Revert commits 8aee306 
Change-Id: I3dd999c77c6779332a40dbb84371ca487216f189
 2016-07-28 17:30:53 +05:30
praveeng
1b5d104afe removed changes from readme file which are giving confilcts 
Change-Id: Ic71ad1313e1404fed444e899466043704d875af6
 2016-07-28 17:30:52 +05:30
praveeng
d81273047b first commit 
Change-Id: Ib50c81acda3b2c1583da3d421efc0ca547ef68e2
 2016-07-28 17:30:04 +05:30
praveeng
65905c3011 small modification to readme for git push test 
Change-Id: I68506a49586b07eaa907f3f85304ee40d4c92d0a
 2016-07-28 17:28:55 +05:30
praveeng
23cca231be first commit 
Change-Id: Ib50c81acda3b2c1583da3d421efc0ca547ef68e2
 2016-07-28 17:26:49 +05:30
praveeng
922e309170 small modification to readme for git push test 
Change-Id: I68506a49586b07eaa907f3f85304ee40d4c92d0a
 2016-07-28 17:26:49 +05:30
praveeng
b0d510bf0e Revert commits 357c990bdd 
Change-Id: I12a34456d7eed93fda4369e76bcddb42ba7ccb99
 2016-07-28 15:11:08 +05:30
praveeng
5ebeece5b4 Revert commits 8aee306 
Change-Id: I3dd999c77c6779332a40dbb84371ca487216f189
 2016-07-28 15:01:36 +05:30
Devin Matthews
6ce4c022eb Switch back to 24x8. I could only squeeze 24.5GFLOP out of 8x24, and scalability is not improved. 2016-07-27 16:26:36 -05:00
Field G. Van Zee
d52cb76715 Merge branch 'master' into compose 2016-07-27 16:04:55 -05:00
Field G. Van Zee
c31b1e7b9d Relax alignment restrictions for sandybridge ukrs. 
Details:
- Relaxed the base pointer and leading dimension alignment restrictions
  in the sandybridge gemm microkernels, allowing the use of vmovups/vmovupd
  instead of vmovaps/vmovapd. These change mimic those made to the haswell
  microkernels in e0d2fa0 and ee2c139.
- Updated testsuite modules as well as standalone test drivers in 'test'
  directory to use DBL_MAX as the initial time candidate. Thanks to Devin
  Matthews for suggesting this change.
- Inserted #include "float.h" into bli_system.h (to gain access to DBL_MAX).
- Minor update (vis-a-vis contexts) to driver code in test/3m4m.
 2016-07-27 15:58:07 -05:00
Devin Matthews
b8f2b55532 Try an 8x24 kernel for the hell of it. 2016-07-27 15:22:55 -05:00
Devin Matthews
7ede5863ae Allocate pack buffer on MCDRAM for KNL. 2016-07-27 13:42:32 -06:00
Devin Matthews
ad89ed2e82 Merge branch 'knl' of github.com:devinamatthews/blis into knl 2016-07-27 11:45:40 -05:00
Devin Matthews
2c9de740ed This version gets ~26GF on one core. 2016-07-27 11:44:54 -05:00
Devin Matthews
81e2b05f31 Add optimized packing kernels for KNL. 2016-07-27 11:39:05 -05:00
Devin Matthews
a7d8ca97b8 All fixed. 2016-07-25 15:15:13 -05:00
Devin Matthews
963d0393b0 Add 24xk pack kernel. 2016-07-25 14:40:53 -05:00
Devin Matthews
117b76739a In the midst of debugging. 2016-07-25 13:53:07 -05:00
Devin Matthews
8c0a4fd1d3 Fix some row/column confusion. 2016-07-25 13:09:24 -05:00
Devin Matthews
c44f9f9693 Simplify displacements -- clang assembler was badly botching EVEX compressed displacements giving false alarms for instruction length. 2016-07-25 12:02:24 -05:00
Devin Matthews
e0cce177cc Minor fixes for 8x24 KNL kernel. 2016-07-25 10:02:25 -05:00
praveeng
50a2f2efcb Merge master code as on 2016_07_25 to amd-staging branch by praveeng 
Change-Id: I84886ae241db2aac0bef6b7ef399f04aa8bca16d
 2016-07-25 17:07:38 +05:30
praveeng
cfd46c88d5 Merge remote-tracking branch 'publicrepo/master' 2016-07-25 15:38:13 +05:30
praveeng
f493bf4d70 removed changes from readme file which are giving confilcts 
Change-Id: Ic71ad1313e1404fed444e899466043704d875af6
 2016-07-25 14:14:00 +05:30
Devin Matthews
65735bbedf Switch to 24x8 kernel, unrolled by 16. 2016-07-24 21:50:32 -05:00
Devin Matthews
45d5dc9717 Add 24x8 "KNC-style" kernel for KNL. 2016-07-24 14:25:26 -05:00
Field G. Van Zee
95abea46f8 Merge branch 'master' into compose 2016-07-23 15:38:33 -05:00
Field G. Van Zee
a017062fdf Integrated "memory broker" (membrk_t) abstraction. 
Details:
- Integrated a patch originally authored and submitted by Ricardo Magana
  of HP Enterprise. The changeset inserts use of a new object type, membrk_t,
  (memory broker) that allows multiple sets of memory pools on, for example,
  separate NUMA nodes, each of which has a separate memory space.
- Added membrk field to cntx_t and defined corresponding accessor macros.
- Added membrk field to mem_t object and defined corresponding accessor macros.
- Created new bli_membrk.c file, which contains the new memory broker API,
  including:
    bli_membrk_init(), bli_membrk_finalize()
    bli_membrk_acquire_[mv](), bli_membrk_release(),
    bli_membrk_init_pools(), bli_membrk_reinit_pools(),
    bli_membrk_finalize_pools(),
    bli_membrk_pool_size()
- In bli_mem.c, changed function calls to
    bli_mem_init_pools()     -> bli_membrk_init()
    bli_mem_reinit_pools()   -> bli_membrk_reinit()
    bli_mem_finalize_pools() -> bli_membrk_finalize()
- In bli_packv_init.c, bli_packm_init.c, changed function calls to:
    bli_mem_acquire_[mv]() -> bli_membrk_acquire_[mv]()
    bli_mem_release()      -> bli_membrk_release()
- Added bli_mutex.c and related files to frame/thread. These files define
  abstract mutexes (locks) and corresponding APIs for pthreads, openmp, or
  single-threaded execution. This new API is employed within functions
  such as bli_membrk_acquire_[mv]() and bli_membrk_release().
 2016-07-22 17:02:59 -05:00
Devin Matthews
8ff2e069c4 Add 4x unrolled variant for KNL microkernel. 2016-07-22 16:22:26 -05:00
Devin Matthews
9cb2ed9b0c Git rid of one RBX update. 2016-07-22 16:10:30 -05:00
Devin Matthews
451bde076f Add some more knobs to twiddle for KNL microkernel. 2016-07-22 15:43:00 -05:00
Devin Matthews
8c6e621c09 Make knl conform to new kernel dir structure. 2016-07-22 15:05:15 -05:00
Devin Matthews
ce7214c661 Merge remote-tracking branch 'origin/master' into knl 2016-07-22 14:59:53 -05:00
Field G. Van Zee
ce59f81108 Merge pull request #88 from devinamatthews/32bit-dim_t 
Handle 32-bit dim_t in 64-bit microkernels.
 2016-07-22 14:48:14 -05:00
Devin Matthews
707a2b7fac Somehow forgot the most important microkernel. 2016-07-22 13:49:44 -05:00
Devin Matthews
47ec045056 Merge remote-tracking branch 'upstream/master' into 32bit-dim_t 2016-07-22 13:45:23 -05:00
Devin Matthews
08f1d6b6fa Use 64-bit intermediate variable for k for architectures that do 64-bit loads in case dim_t is 32-bit. 2016-07-22 13:44:37 -05:00
Field G. Van Zee
ff41153f4e Merge pull request #86 from devinamatthews/haswell-vmovups 
Remove alignment restrictions on C in haswell kernel.
 2016-07-22 13:21:03 -05:00
Devin Matthews
e0d2fa0d83 Relax alignment restrictions for haswell sgemm. 2016-07-22 12:56:51 -05:00
Field G. Van Zee
f9214ced97 Merge pull request #85 from devinamatthews/qopenmp 
Change -openmp to -fopenmp for icc.
 2016-07-22 12:16:39 -05:00
Devin Matthews
ee2c139df6 Remove alignment restrictions on C in haswell kernel. 2016-07-22 12:06:03 -05:00
Devin Matthews
08666eaa20 Change -openmp to -fopenmp for icc. 2016-07-22 11:07:34 -05:00
Devin Matthews
119d039942 Add 8x24 KNL kernel. 2016-07-22 10:23:31 -05:00
praveeng
1aa77dfc1d Merge master code as on 2016_07_21 to amd-staging branch by praveeng 
Change-Id: Ic7d0a21101358f08147736e7f1884e7409937344
 2016-07-21 14:23:41 +05:30
Devin Matthews
b58cda9eba Merge remote-tracking branch 'origin/master' into knl 
# Conflicts:
#	frame/base/bli_threading.h
#	frame/include/blis.h
#	frame/thread/bli_thread.c
 2016-07-19 14:09:09 -05:00