Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
gaoqiong
MIGraphX
Commits
b3935928
"src/git@developer.sourcefind.cn:gaoqiong/migraphx.git" did not exist on "f201285c8cd08a1140090be6bd3bffcb0746775f"
Commit
b3935928
authored
Jun 25, 2019
by
Paul
Browse files
Formatting
parent
506a73ec
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
4 additions
and
4 deletions
+4
-4
src/targets/gpu/device/include/migraphx/gpu/device/shape.hpp
src/targets/gpu/device/include/migraphx/gpu/device/shape.hpp
+3
-3
src/targets/gpu/device/reduce_sum.cpp
src/targets/gpu/device/reduce_sum.cpp
+1
-1
No files found.
src/targets/gpu/device/include/migraphx/gpu/device/shape.hpp
View file @
b3935928
...
@@ -77,11 +77,11 @@ struct hip_shape
...
@@ -77,11 +77,11 @@ struct hip_shape
MIGRAPHX_DEVICE_CONSTEXPR
hip_index
carry
(
hip_index
result
)
const
MIGRAPHX_DEVICE_CONSTEXPR
hip_index
carry
(
hip_index
result
)
const
{
{
std
::
ptrdiff_t
rem
=
0
;
std
::
ptrdiff_t
rem
=
0
;
for
(
std
::
ptrdiff_t
i
=
result
.
size
()
-
1
;
i
>=
0
;
i
--
)
for
(
std
::
ptrdiff_t
i
=
result
.
size
()
-
1
;
i
>=
0
;
i
--
)
{
{
auto
z
=
result
[
i
]
+
rem
;
auto
z
=
result
[
i
]
+
rem
;
rem
=
z
-
std
::
ptrdiff_t
(
lens
[
i
])
+
1
;
rem
=
z
-
std
::
ptrdiff_t
(
lens
[
i
])
+
1
;
if
(
rem
>
0
)
if
(
rem
>
0
)
z
-=
rem
;
z
-=
rem
;
else
else
rem
=
0
;
rem
=
0
;
...
...
src/targets/gpu/device/reduce_sum.cpp
View file @
b3935928
...
@@ -70,7 +70,7 @@ void reduce_sum(hipStream_t stream, const argument& result, const argument& arg)
...
@@ -70,7 +70,7 @@ void reduce_sum(hipStream_t stream, const argument& result, const argument& arg)
const
std
::
size_t
block_size
=
compute_block_size
(
relements
,
max_block_size
);
const
std
::
size_t
block_size
=
compute_block_size
(
relements
,
max_block_size
);
gs_launch
(
stream
,
nelements
*
block_size
,
block_size
)([
=
](
auto
i
,
auto
idx
)
__device__
{
gs_launch
(
stream
,
nelements
*
block_size
,
block_size
)([
=
](
auto
i
,
auto
idx
)
__device__
{
const
auto
out_idx
=
i
/
block_size
;
const
auto
out_idx
=
i
/
block_size
;
auto
base_idx
=
output
.
get_shape
().
multi
(
out_idx
);
auto
base_idx
=
output
.
get_shape
().
multi
(
out_idx
);
auto
r
=
block_reduce
<
max_block_size
>
(
idx
,
sum
{},
0
,
relements
,
[
&
](
auto
j
)
__device__
{
auto
r
=
block_reduce
<
max_block_size
>
(
idx
,
sum
{},
0
,
relements
,
[
&
](
auto
j
)
__device__
{
auto
reduce_idx
=
reduce_shape
.
multi
(
j
);
auto
reduce_idx
=
reduce_shape
.
multi
(
j
);
return
input
[
reduce_idx
+
base_idx
];
return
input
[
reduce_idx
+
base_idx
];
...
...
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