Martin Kroeker
2224bcb4f0
Update cdot_microk_sandy-2.c
7 years ago
Martin Kroeker
093a3d7d57
Update cdot_microk_haswell-2.c
7 years ago
Martin Kroeker
ba6d2c77a9
Update cdot_microk_bulldozer-2.c
7 years ago
Martin Kroeker
4673e53178
Update caxpy_microk_steamroller-2.c
7 years ago
Martin Kroeker
f6be89295f
Update caxpy_microk_sandy-2.c
7 years ago
Martin Kroeker
1878e0c95a
Update caxpy_microk_haswell-2.c
7 years ago
Martin Kroeker
300bb19b3e
Update caxpy_microk_bulldozer-2.c
7 years ago
Martin Kroeker
bbc30700e8
Update saxpy_microk_nehalem-2.c
7 years ago
Martin Kroeker
e8d835ea46
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
715b1f263d
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
b6f4ef5aea
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
f78531a9ec
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
af29c99c85
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
2b542d1036
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
0cfb647a57
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
0172c51829
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
c931bb8172
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
ba9f792e75
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
cd3a35ee79
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
d384880da5
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
922e448978
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
6fcb55b22f
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
2bd18c7b73
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
b13f3c3bcf
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
3f1719a98d
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
dc15f3b5a7
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
00aff05c40
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
c9078eb8b4
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
de207d10c1
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
c23c17163f
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
c18c2c9d9b
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
ca02ac724f
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
9d46f84f24
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
6008f65318
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
d94e7da701
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
7af8f34df4
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
bb16456fe1
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
2f5a7c1656
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
30a7bd8e15
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
47e2b4592e
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
a671e19dd2
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
663eef3b66
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
4e6f8fec31
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
8a6bbf5a5b
Tag operands 0 and 1 as both input and output
7 years ago
Martin Kroeker
f0dd058430
Tag operands 0 and 1 as both input and output
For #1964 (basically a continuation of coding problems first seen in #1292 )
7 years ago
Arjan van de Ven
795285c587
Fix thinko in skylake beta handling
casting ints is cheaper but it has a rounding, not memory casing effect, resulting in
invalid outcome
7 years ago
Arjan van de Ven
d321448a63
dgemm: use dgemm_ncopy_8_skylakex.c also for Haswell
The dgemm_ncopy_8_skylakex.c code is not avx512 specific and gives
a nice performance boost for medium sized matrices
7 years ago
Arjan van de Ven
c43331ad0a
dgemm: Use the skylakex beta function also for haswell
it's more efficient for certain tall/skinny matrices
7 years ago
Arjan van de Ven
69d206440a
Make the skylakex/haswell sgemm code compile and run even with compilers without avx2 support
7 years ago
Arjan van de Ven
0586899a10
Use sgemm_ncopy_4_skylakex.c also for Haswell
sgemm_ncopy_4_skylakex.c uses SSE transpose operations where the
real perf win happens; this also works great for Haswell.
This gives double digit percentage gains on small and skinny matrices
7 years ago