nmeyer-ur
|
5cb3530c34
|
enable counters in Benchmark_wilson
|
2020-05-29 15:44:52 +02:00 |
|
nmeyer-ur
|
250008372f
|
update SVE readme
|
2020-05-29 15:44:25 +02:00 |
|
nmeyer-ur
|
4fedd8d29f
|
switch to MPI_THREAD_SERIALIZED instead of SINGLE
|
2020-05-27 14:08:34 +02:00 |
|
nmeyer-ur
|
6ddcef1bca
|
fix build error enabling fcmla/mac in vector types for VLA
|
2020-05-21 21:21:03 +02:00 |
|
nmeyer-ur
|
8c5a5fdfce
|
disable fcmla in vector type building for VLA
|
2020-05-21 19:41:42 +02:00 |
|
nmeyer-ur
|
046b1cbbc0
|
enable fcmla in tensor arithmetics; fixed-size works, VLA does not compile
|
2020-05-21 19:39:07 +02:00 |
|
nmeyer-ur
|
a65ce237c1
|
clean up; Exch1 VLA sp+dp integrate, tested, working
|
2020-05-21 09:48:06 +02:00 |
|
nmeyer-ur
|
cd27f1005d
|
clean up; Exch1 sp integrate, tested, working
|
2020-05-21 08:45:43 +02:00 |
|
nmeyer-ur
|
f8c0a59221
|
clean up; Exch1 dp integrate, tested, working
|
2020-05-21 02:48:14 +02:00 |
|
nmeyer-ur
|
832485699f
|
save some cycles in HtoD and DtoH by direct instead of multi-pass conversion
|
2020-05-20 23:04:35 +02:00 |
|
nmeyer-ur
|
81484a4760
|
symmetrize Mult and MultAddComplex
|
2020-05-20 22:36:45 +02:00 |
|
nmeyer-ur
|
9a86059761
|
symmetrize VLA and fixed size build messages
|
2020-05-20 20:05:42 +02:00 |
|
nmeyer-ur
|
b780b7b7a0
|
guard prevents multiple TOFU messages
|
2020-05-20 19:20:59 +02:00 |
|
nmeyer-ur
|
9e085bd04e
|
guard prevents multiple A64FX build messages
|
2020-05-20 19:16:30 +02:00 |
|
nmeyer-ur
|
6b6bf537d3
|
comment out mac in vector types
|
2020-05-18 20:36:16 +02:00 |
|
nmeyer-ur
|
323a651c71
|
correct typo
|
2020-05-18 19:58:27 +02:00 |
|
nmeyer-ur
|
9f212679f1
|
support fcmla in vector_types, untested
|
2020-05-18 19:55:18 +02:00 |
|
nmeyer-ur
|
032f7dde1a
|
update SVE readme, asm generator
|
2020-05-18 19:10:36 +02:00 |
|
nmeyer-ur
|
50b1db1e8b
|
implemented correct _m form (using 3 operands instead of 2)
|
2020-05-15 10:01:05 +02:00 |
|
nmeyer-ur
|
015d8bb38a
|
introduced assertions in Benchmark_wilson, removed data output from Benchmark_dwf
|
2020-05-15 09:15:50 +02:00 |
|
nmeyer-ur
|
10a34312dc
|
some fixed-size code clean up
|
2020-05-14 23:20:16 +02:00 |
|
nmeyer-ur
|
db8c0e7584
|
replaced _x form with _m form when using even/odd predication
|
2020-05-14 23:17:35 +02:00 |
|
nmeyer-ur
|
d15ccad8a7
|
switched to vec* in Reduce
|
2020-05-12 20:41:14 +02:00 |
|
nmeyer-ur
|
0009b5cee8
|
updated SVE_README
|
2020-05-12 19:02:33 +02:00 |
|
nmeyer-ur
|
20d1941a45
|
enabled asm kernels for fixed-size A64FXFIXEDSIZE
|
2020-05-12 19:01:12 +02:00 |
|
nmeyer-ur
|
b7c76ede29
|
Removed some assertions in Test_simd and removed exit() in Reduce
|
2020-05-11 22:43:00 +02:00 |
|
nmeyer-ur
|
05edf803bd
|
corrected typo
|
2020-05-12 03:59:59 +09:00 |
|
nmeyer-ur
|
78b8e40f83
|
switched to gcc's internal data types
|
2020-05-11 18:11:23 +02:00 |
|
nmeyer-ur
|
fc2e9850d3
|
temporarily enable TOFU by default when using A64FX or A64FXFIXEDSIZE
|
2020-05-11 13:25:02 +02:00 |
|
nmeyer-ur
|
ffaaed679e
|
MPI_THREAD_SINGLE hack for Fugaku, enabled by -DTOFU
|
2020-05-11 13:21:39 +02:00 |
|
nmeyer-ur
|
b2fd8b993a
|
fixed-size clean up
|
2020-05-09 22:53:42 +02:00 |
|
nmeyer-ur
|
291ee8c3d0
|
updated fixed-size implementation; only Exch1 and prefetches missing
|
2020-05-09 22:18:02 +02:00 |
|
nmeyer-ur
|
e1a5b3ea49
|
unions for tables eliminate explicit loads, gcc does not complain
|
2020-05-09 21:21:57 +02:00 |
|
nmeyer-ur
|
55a55660cb
|
reverted changes
|
2020-05-09 12:48:42 +02:00 |
|
nmeyer-ur
|
ceb8b374da
|
API change v3
|
2020-05-08 15:04:44 +02:00 |
|
nmeyer-ur
|
4bc2ad2894
|
API change v2
|
2020-05-08 15:00:25 +02:00 |
|
nmeyer-ur
|
798af3e68f
|
retry changing StoD API
|
2020-05-08 14:34:59 +02:00 |
|
nmeyer-ur
|
b0ef2367f3
|
testing alternate call to PrecisionChange
|
2020-05-08 14:22:44 +02:00 |
|
nmeyer-ur
|
71a7350a85
|
changed 2nd argument in Reduce to native vector type
|
2020-05-08 12:26:51 +02:00 |
|
nmeyer-ur
|
6f79369955
|
trying to get rid of macro definition error
|
2020-05-08 12:19:24 +02:00 |
|
nmeyer-ur
|
f9cb6b979f
|
corrected more typos
|
2020-05-08 12:11:01 +02:00 |
|
nmeyer-ur
|
ed4d9d17f8
|
corrected type
|
2020-05-08 12:09:22 +02:00 |
|
nmeyer-ur
|
fbed02690d
|
some changes in breaking out A64FX: use -DA64FXFIXEDSIZE for fixed size, but also define GEN
|
2020-05-08 12:05:31 +02:00 |
|
nmeyer-ur
|
39f3ae5b1d
|
corrected more types
|
2020-05-08 11:07:14 +02:00 |
|
nmeyer-ur
|
e64bec8c8e
|
pulled SVE typedefs out of Optimization
|
2020-05-08 11:04:21 +02:00 |
|
nmeyer-ur
|
0893b4e552
|
fixed typos in PrecisionChange
|
2020-05-08 10:59:07 +02:00 |
|
nmeyer-ur
|
92f0f29670
|
fixed double overloading vecf in Div, corrected typos
|
2020-05-08 10:57:23 +02:00 |
|
nmeyer-ur
|
48a340a9d1
|
GEN seems to defined by default -> some fixes applied
|
2020-05-08 10:47:49 +02:00 |
|
nmeyer-ur
|
f45621109b
|
placed typedefs in Optimization
|
2020-05-08 10:41:52 +02:00 |
|
nmeyer-ur
|
32d1a0bbea
|
added even more debug output
|
2020-05-08 10:39:26 +02:00 |
|
nmeyer-ur
|
267cce66a1
|
added more debug output
|
2020-05-08 10:29:28 +02:00 |
|
nmeyer-ur
|
3417147b11
|
added real fma, corrected typos in tbls; integrated, must supply A64FXGCC with GEN in configure
|
2020-05-08 10:20:19 +02:00 |
|
nmeyer-ur
|
b338719bc8
|
first transition to fixed-size done, excl. Exch; next step: integration
|
2020-05-07 22:33:28 +02:00 |
|
nmeyer-ur
|
2b81cbe2c2
|
first attempt to introduce tables using fixed-size; still incomplete
|
2020-05-07 22:01:19 +02:00 |
|
nmeyer-ur
|
acff9d6ed2
|
transition to fixed size data types almost done; still incomplete
|
2020-05-07 21:24:07 +02:00 |
|
nmeyer-ur
|
a306a49788
|
first mods for fixed size; still incomplete
|
2020-05-07 19:07:49 +02:00 |
|
nmeyer-ur
|
7ef03c5368
|
updated SVE readme
|
2020-05-06 16:30:37 +02:00 |
|
nmeyer-ur
|
5abec5b8a9
|
SVE_readme update, update Grid_vector_types.h
|
2020-04-25 13:48:26 +02:00 |
|
nmeyer-ur
|
499edc0636
|
updated SVE_README.txt; defined ARMCLANGCOMPAT macro
|
2020-04-25 13:41:24 +02:00 |
|
nmeyer-ur
|
d990e61be3
|
armclang 20.1 settings in SVE readme
|
2020-04-25 12:11:43 +02:00 |
|
nmeyer-ur
|
3edb2dc2da
|
removed -static from gcc CXXFLAGS
|
2020-04-24 13:04:34 +02:00 |
|
nils meyer
|
345721220e
|
resolved merge conflict
|
2020-04-24 10:14:21 +02:00 |
|
nils meyer
|
6db68d6ecb
|
added SVE configure for armclang and gcc
|
2020-04-24 10:10:47 +02:00 |
|
nmeyer-ur
|
09f0963d1f
|
changes in configure.ac ; to be verified
|
2020-04-23 11:27:03 +02:00 |
|
nils meyer
|
6f44e3c192
|
reverted changes in configure.ac ; included SVE configure readme
|
2020-04-23 11:18:50 +02:00 |
|
nils meyer
|
5893888f87
|
removed default no-strict-aliasing for gcc-10.0.1 exclusively
|
2020-04-22 19:29:55 +02:00 |
|
nmeyer-ur
|
39b448affb
|
Merge remote-tracking branch 'origin/develop' into feature/a64fx-2
|
2020-04-22 17:34:12 +02:00 |
|
nils meyer
|
e54a8f05a9
|
Exchange1 with generic version for now, should use svtbl2 in final version
|
2020-04-20 22:45:27 +02:00 |
|
nils meyer
|
64b72fc17f
|
testing gcc 10.0.1: build errors in Exchange1 using -DA64FX and in Lattice_base.h building Dslash only
|
2020-04-19 01:25:40 +02:00 |
|
nils meyer
|
6fdce60492
|
revised BodyA64FX; 990 GiB/s Wilson, 687 GiB/s DW using intrinsics (armclang 20.0)
|
2020-04-16 22:43:32 +02:00 |
|
nils meyer
|
852db4626a
|
re-introduced HOTFIX cause Grid binaries give wrong results otherwise; checked in good gridverter.py
|
2020-04-15 18:22:19 +02:00 |
|
nils meyer
|
6504a098cc
|
999 GiB/s Wilson; 694 GiB/s DW (DP)
|
2020-04-15 15:06:52 +02:00 |
|
nils meyer
|
79a385faca
|
disabled armclang hotfix cause armclang 20.0 performance gets a little
|
2020-04-15 11:46:55 +02:00 |
|
nils meyer
|
c12a67030a
|
980 GiB/s Wilson; 680 GiB/s DW (DP)
|
2020-04-15 10:55:06 +02:00 |
|
nils meyer
|
581392f2f2
|
now with pf, best results so far using intrinsics+pf
|
2020-04-12 22:06:14 +02:00 |
|
nils meyer
|
113f277b6a
|
enable dslash asm using -DA64FXASM, additionaly -DDSLASHINTRIN for intrinsics impl
|
2020-04-11 04:55:01 +02:00 |
|
nils meyer
|
974586bedc
|
Dslash finally works; cleaned up; uses MOVPRFX in assembly
|
2020-04-10 22:26:40 +02:00 |
|
nmeyer-ur
|
160f78c1e4
|
changed debug output to variable direct 3
|
2020-04-10 12:23:07 +02:00 |
|
nmeyer-ur
|
7e4e1bbbc2
|
changed debug output to variable direct 2
|
2020-04-10 12:22:04 +02:00 |
|
nmeyer-ur
|
e699b7e9f9
|
changed debug output to variable direct
|
2020-04-10 12:18:30 +02:00 |
|
nmeyer-ur
|
a28bc0de90
|
debug register address test in WilsonHand
|
2020-04-10 12:07:45 +02:00 |
|
nmeyer-ur
|
14d0fe4d6c
|
added predication in WilsonHand
|
2020-04-10 12:04:00 +02:00 |
|
nmeyer-ur
|
0ad2e0815c
|
debug output in WilsonHand
|
2020-04-10 11:56:29 +02:00 |
|
nils meyer
|
1c8ca05e16
|
Merge branch 'feature/a64fx-2' of https://github.com/nmeyer-ur/Grid into feature/a64fx-2
|
2020-04-09 23:32:19 +02:00 |
|
nils meyer
|
dc9c8340bb
|
switched to DSLASHINTRIN for A64FX Dslash intrinsics
|
2020-04-09 23:30:23 +02:00 |
|
nils meyer
|
19eef97503
|
specialized A64FX Dslash kernels
|
2020-04-09 23:25:25 +02:00 |
|
nmeyer-ur
|
635246ce50
|
corrected typo
|
2020-04-09 21:42:50 +02:00 |
|
nils meyer
|
5cdbb7e71e
|
fixed A64FX Dslash; compiles, but does not specialize -> assertion
|
2020-04-09 21:23:39 +02:00 |
|
nmeyer-ur
|
8123590a1b
|
changes
|
2020-04-09 16:45:47 +02:00 |
|
nmeyer-ur
|
86c9c4da8b
|
changes
|
2020-04-09 16:40:06 +02:00 |
|
nmeyer-ur
|
cd1efee866
|
changes
|
2020-04-09 16:35:13 +02:00 |
|
nmeyer-ur
|
bd310932f7
|
changes
|
2020-04-09 16:32:31 +02:00 |
|
nmeyer-ur
|
304762e7ac
|
changes
|
2020-04-09 16:26:01 +02:00 |
|
nmeyer-ur
|
d79ab03a6c
|
changes
|
2020-04-09 16:19:25 +02:00 |
|
nmeyer-ur
|
d5708e0eb2
|
more changes
|
2020-04-09 15:43:34 +02:00 |
|
nmeyer-ur
|
123f6b7a61
|
more changes
|
2020-04-09 15:17:19 +02:00 |
|
nmeyer-ur
|
2b6457dd9a
|
added xp/xm recon accum
|
2020-04-09 15:13:19 +02:00 |
|
nmeyer-ur
|
b367cbd422
|
defined ADD_RESULT
|
2020-04-09 15:08:45 +02:00 |
|
nmeyer-ur
|
e252c1aca3
|
addressing
|
2020-04-09 15:03:12 +02:00 |
|
nmeyer-ur
|
b140c6a4f9
|
addressing
|
2020-04-09 15:01:15 +02:00 |
|