Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
Menu
Open sidebar
pub
pi
Commits
e300c310
Commit
e300c310
authored
Feb 09, 2020
by
Neil Gershenfeld
Browse files
wip
parent
21d19cbd
Pipeline
#4933
passed with stage
in 1 second
Changes
2
Pipelines
1
Hide whitespace changes
Inline
Side-by-side
Python/numbapig.py
View file @
e300c310
#
# numbapig.py
# Neil Gershenfeld 2/9/20
# calculation of pi by a Numba
CUDA
sum
# calculation of pi by a Numba
GPU
sum
# pi = 3.14159265358979323846
#
from
numba
import
cuda
...
...
@@ -20,6 +20,7 @@ NPTS = grid_size*block_size
def
init
(
arr
):
i
=
1
+
cuda
.
grid
(
1
)
arr
[
i
-
1
]
=
0.5
/
((
i
-
0.75
)
*
(
i
-
0.25
))
#arr[i-1] = i # for testing reduction
#
@
cuda
.
reduce
def
Numba_reduce
(
a
,
b
):
...
...
@@ -39,15 +40,25 @@ def CUDA_reduce(arr,NPTS):
if
(
len
==
0
):
return
#
@
cuda
.
jit
def
CUDA_result
(
arr
,
result
):
i
=
cuda
.
grid
(
1
)
if
(
i
==
0
):
result
[
0
]
=
arr
[
0
]
#
# device array
#
arr
=
cuda
.
device_array
(
NPTS
,
np
.
float32
)
result
=
cuda
.
device_array
(
1
,
np
.
float32
)
#arr = cuda.device_array(NPTS,np.int64) # for testing reduction
#result = cuda.device_array(1,np.int64) # for testing reduction
#
# compile kernels
# compile kernels
by calling them
#
init
[
grid_size
,
block_size
](
arr
)
pi
=
Numba_reduce
(
arr
)
CUDA_reduce
(
arr
,
NPTS
)
CUDA_result
(
arr
,
result
)
#
# CUDA kernel array calculation
#
...
...
@@ -94,9 +105,11 @@ print(" time = %f, estimated MFlops = %f"%(end_time-start_time,mflops))
start_time
=
time
.
time
()
init
[
grid_size
,
block_size
](
arr
)
CUDA_reduce
(
arr
,
NPTS
)
CUDA_result
(
arr
,
result
)
end_time
=
time
.
time
()
darr
=
arr
.
copy_to_host
()
pi
=
result
.
copy_to_host
()
mflops
=
NPTS
*
5.0
/
(
1.0e6
*
(
end_time
-
start_time
))
print
(
"both with CUDA kernel reduction:"
)
print
(
" NPTS = %d, pi = %f"
%
(
NPTS
,
darr
[
0
]))
print
(
" NPTS = %d, pi = %f"
%
(
NPTS
,
pi
[
0
]))
print
(
" time = %f, estimated MFlops = %f"
%
(
end_time
-
start_time
,
mflops
))
README.md
View file @
e300c310
...
...
@@ -6,6 +6,7 @@
|88,333|
[
mpimppi.c
](
hybrid/mpimppi.c
)
|C, MPI+OpenMP, 1024 nodes, 64 cores/node, 4 threads/core
<br>
cc mpimppi.c -o mpimppi -O3 -ffast-math -fopenmp|Argonne ALCF Theta
<br>
Cray XC40|Oct 9, 2019|
|2,117|
[
mpipi2.c
](
MPI/mpipi2.c
)
|C, MPI, 10 nodes, 96 cores/node
<br>
mpicc mpipi2.c -o mpipi2 -O3 -ffast-math|Intel 2x Xeon Platinum 8175M|Oct 24, 2019|
|2,102|
[
mpipi2.py
](
Python/mpipi2.py
)
|Python, Numba, MPI
<br>
10 nodes, 96 cores/node|Intel 2x Xeon Platinum 8175M|Feb 6, 2020|
|1,919|
[
numbapig.py
](
Python/numbapig.py
)
|Python, Numba, GPU
<br>
5120 cores|NVIDIA V100|Feb 9, 2020|
|315|
[
numbapip.py
](
Python/numbapip.py
)
|Python, Numba, parallel, fastmath
<br>
96 cores|Intel 2x Xeon Platinum 8175M|Feb 7, 2020|
|272|
[
threadpi.c
](
C/threadpi.c
)
|C, 96 threads
<br>
gcc threadpi.c -o threadpi -O3 -ffast-math -pthread|Intel 2x Xeon Platinum 8175M|Jun 3, 2019|
|211|
[
mpipi2.c
](
MPI/mpipi2.c
)
|C, MPI, 1 node, 96 cores
<br>
mpicc mpipi2.c -o mpipi2 -O3 -ffast-math|Intel 2x Xeon Platinum 8175M|Oct 24, 2019|
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a 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