Skip to content
GitLab
Projects
Groups
Snippets
/
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
Menu
Open sidebar
elpa
elpa
Commits
de20e578
Commit
de20e578
authored
Sep 07, 2016
by
Andreas Marek
Browse files
Single precision complex block2 AVX-512 kernel
parent
d6867ff5
Changes
3
Pipelines
1
Expand all
Hide whitespace changes
Inline
Side-by-side
Makefile.am
View file @
de20e578
...
...
@@ -265,9 +265,9 @@ endif
if
WITH_COMPLEX_AVX512_BLOCK2_KERNEL
libelpa@SUFFIX@
_private_la_SOURCES
+=
src/elpa2_kernels/elpa2_kernels_complex_avx512_2hv_double_precision.c
#
if WANT_SINGLE_PRECISION_COMPLEX
#
libelpa@SUFFIX@_private_la_SOURCES += src/elpa2_kernels/elpa2_kernels_complex_avx512_2hv_single_precision.c
#
endif
if
WANT_SINGLE_PRECISION_COMPLEX
libelpa@SUFFIX@
_private_la_SOURCES
+=
src/elpa2_kernels/elpa2_kernels_complex_avx512_2hv_single_precision.c
endif
endif
.cu.lo
:
...
...
src/elpa2_kernels/elpa2_kernels_complex_avx512_2hv_single_precision.c
View file @
de20e578
This diff is collapsed.
Click to expand it.
src/mod_compute_hh_trafo_complex.F90
View file @
de20e578
...
...
@@ -568,33 +568,33 @@ module compute_hh_trafo_complex
#endif /* WITH_NO_SPECIFIC_COMPLEX_KERNEL */
#endif /* WITH_COMPLEX_AVX_BLOCK2_KERNEL || WITH_COMPLEX_AVX2_BLOCK2_KERNEL */
!
#if defined(WITH_COMPLEX_AVX512_BLOCK2_KERNEL)
!
#if defined(WITH_NO_SPECIFIC_COMPLEX_KERNEL)
!
if ( (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_AVX512_BLOCK2) ) then
!
#endif /* WITH_NO_SPECIFIC_COMPLEX_KERNEL */
!
ttt = mpi_wtime()
!
do j = ncols, 2, -2
!
w(:,1) = bcast_buffer(1:nbw,j+off)
!
w(:,2) = bcast_buffer(1:nbw,j+off-1)
!
#ifdef WITH_OPENMP
!
call double_hh_trafo_complex_avx512_2hv_single(a(1,j+off+a_off-1,istripe,my_thread), &
!
w, nbw, nl, stripe_width, nbw)
!
#else
!
call double_hh_trafo_complex_avx512_2hv_single(a(1,j+off+a_off-1,istripe), &
!
w, nbw, nl, stripe_width, nbw)
!
#endif
!
enddo
!
#ifdef WITH_OPENMP
!
if (j==1) call single_hh_trafo_complex_avx512_1hv_single(a(1,1+off+a_off,istripe,my_thread), &
!
bcast_buffer(1,off+1), nbw, nl, stripe_width)
!
#else
!
if (j==1) call single_hh_trafo_complex_avx512_1hv_single(a(1,1+off+a_off,istripe), &
!
bcast_buffer(1,off+1), nbw, nl, stripe_width)
!
#endif
!
#if defined(WITH_NO_SPECIFIC_COMPLEX_KERNEL)
!
endif ! (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_AVX512_BLOCK2)
!
#endif /* WITH_NO_SPECIFIC_COMPLEX_KERNEL */
!
#endif /* WITH_COMPLEX_AVX512_BLOCK2_KERNEL */
#if defined(WITH_COMPLEX_AVX512_BLOCK2_KERNEL)
#if defined(WITH_NO_SPECIFIC_COMPLEX_KERNEL)
if
(
(
THIS_COMPLEX_ELPA_KERNEL
.eq.
COMPLEX_ELPA_KERNEL_AVX512_BLOCK2
)
)
then
#endif /* WITH_NO_SPECIFIC_COMPLEX_KERNEL */
ttt
=
mpi_wtime
()
do
j
=
ncols
,
2
,
-2
w
(:,
1
)
=
bcast_buffer
(
1
:
nbw
,
j
+
off
)
w
(:,
2
)
=
bcast_buffer
(
1
:
nbw
,
j
+
off
-1
)
#ifdef WITH_OPENMP
call
double_hh_trafo_complex_avx512_2hv_single
(
a
(
1
,
j
+
off
+
a_off
-1
,
istripe
,
my_thread
),
&
w
,
nbw
,
nl
,
stripe_width
,
nbw
)
#else
call
double_hh_trafo_complex_avx512_2hv_single
(
a
(
1
,
j
+
off
+
a_off
-1
,
istripe
),
&
w
,
nbw
,
nl
,
stripe_width
,
nbw
)
#endif
enddo
#ifdef WITH_OPENMP
if
(
j
==
1
)
call
single_hh_trafo_complex_avx512_1hv_single
(
a
(
1
,
1
+
off
+
a_off
,
istripe
,
my_thread
),
&
bcast_buffer
(
1
,
off
+1
),
nbw
,
nl
,
stripe_width
)
#else
if
(
j
==
1
)
call
single_hh_trafo_complex_avx512_1hv_single
(
a
(
1
,
1
+
off
+
a_off
,
istripe
),
&
bcast_buffer
(
1
,
off
+1
),
nbw
,
nl
,
stripe_width
)
#endif
#if defined(WITH_NO_SPECIFIC_COMPLEX_KERNEL)
endif
! (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_AVX512_BLOCK2)
#endif /* WITH_NO_SPECIFIC_COMPLEX_KERNEL */
#endif /* WITH_COMPLEX_AVX512_BLOCK2_KERNEL */
#if defined(WITH_COMPLEX_GENERIC_SIMPLE_KERNEL)
...
...
Write
Preview
Supports
Markdown
0%
Try again
or
attach a new file
.
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment