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
composable_kernel_ROCM
Commits
fba3d780
"git@developer.sourcefind.cn:hehl2/torchaudio.git" did not exist on "720c36b1fad2980301f163b3835163c30ea7632b"
Commit
fba3d780
authored
Feb 13, 2025
by
mtgu0705
Browse files
fix bug, function now passes.
parent
c0ef46ff
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
9 additions
and
7 deletions
+9
-7
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle_v3_b_preshuffle.hpp
.../device/impl/device_gemm_xdl_cshuffle_v3_b_preshuffle.hpp
+6
-4
include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp
...operation/gpu/thread/threadwise_tensor_slice_transfer.hpp
+3
-3
No files found.
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle_v3_b_preshuffle.hpp
View file @
fba3d780
...
...
@@ -339,19 +339,21 @@ struct DeviceGemm_Xdl_CShuffleV3_BPreshuffle : public DeviceGemmV2BPreshuffle<AL
if(arg.KBatch > 1)
{
const auto kernel =
kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
kernel_gemm_xdl_cshuffle_v3
_b_preshuffle
<GridwiseGemm,
false,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy>;
minimum_occupancy,
TailNumber::Odd>;
Run(kernel);
}
else
{
const auto kernel =
kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
kernel_gemm_xdl_cshuffle_v3
_b_preshuffle
<GridwiseGemm,
false,
InMemoryDataOperationEnum::Set,
minimum_occupancy>;
minimum_occupancy,
TailNumber::Odd>;
Run(kernel);
}
}
...
...
include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp
View file @
fba3d780
...
...
@@ -377,14 +377,14 @@ struct ThreadwiseTensorSliceTransfer_v2
if
constexpr
(
InvalidElementAsNaN
)
{
dst_buf
(
Number
<
dst_offset
/
PackedSize
>
{})
=
dst_buf
(
Number
<
dst_offset
>
{})
=
is_src_valid
?
type_convert
<
DstData
>
(
src_vector
.
template
AsType
<
SrcData
>()[
i
])
:
NumericLimits
<
DstData
>::
QuietNaN
();
}
else
{
dst_buf
(
Number
<
dst_offset
/
PackedSize
>
{})
=
dst_buf
(
Number
<
dst_offset
>
{})
=
type_convert
<
DstData
>
(
src_vector
.
template
AsType
<
SrcData
>()[
i
]);
}
});
...
...
@@ -1619,7 +1619,7 @@ struct ThreadwiseTensorSliceTransfer_StaticToStatic
constexpr
index_t
src_offset
=
src_desc
.
CalculateOffset
(
src_slice_origin_idx
+
idx_md
+
i
*
dst_scalar_step_in_vector
);
src_tmp_vector
.
template
AsType
<
SrcData
>()(
i
)
=
src_buf
[
Number
<
src_offset
/
PackedSize
>
{}];
src_tmp_vector
.
template
AsType
<
SrcData
>()(
i
)
=
src_buf
[
Number
<
src_offset
>
{}];
});
// copy data from src_tmp_vector to dst_tmp_vector (data cast data from SrcData to
...
...
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