-
Notifications
You must be signed in to change notification settings - Fork 2.9k
[GPU] Support fsv16 Shape agnostic convolution. #25020
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
36192b8 to
d5da6d5
Compare
Signed-off-by: hyunback <[email protected]>
Find the root cause and fixing.. Signed-off-by: hyunback <[email protected]>
Signed-off-by: hyunback <[email protected]>
Signed-off-by: hyunback <[email protected]>
d5da6d5 to
01e85b5
Compare
Signed-off-by: hyunback <[email protected]>
Signed-off-by: hyunback <[email protected]>
1100ab5 to
f3fa3d3
Compare
...s/intel_gpu/src/kernel_selector/kernels/convolution/convolution_kernel_b_fs_yx_fsv16_1x1.cpp
Show resolved
Hide resolved
Signed-off-by: hyunback <[email protected]>
src/plugins/intel_gpu/src/graph/graph_optimizer/compile_graph.cpp
Outdated
Show resolved
Hide resolved
...s/intel_gpu/src/kernel_selector/kernels/convolution/convolution_kernel_b_fs_yx_fsv16_1x1.cpp
Outdated
Show resolved
Hide resolved
...s/intel_gpu/src/kernel_selector/kernels/convolution/convolution_kernel_b_fs_yx_fsv16_1x1.cpp
Outdated
Show resolved
Hide resolved
...s/intel_gpu/src/kernel_selector/kernels/convolution/convolution_kernel_b_fs_yx_fsv16_1x1.cpp
Outdated
Show resolved
Hide resolved
Signed-off-by: hyunback <[email protected]>
Signed-off-by: hyunback <[email protected]>
Signed-off-by: hyunback <[email protected]>
e-ddykim
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good to me
...s/intel_gpu/src/kernel_selector/kernels/convolution/convolution_kernel_b_fs_yx_fsv16_1x1.cpp
Outdated
Show resolved
Hide resolved
...s/intel_gpu/src/kernel_selector/kernels/convolution/convolution_kernel_b_fs_yx_fsv16_1x1.cpp
Outdated
Show resolved
Hide resolved
Signed-off-by: hyunback <[email protected]>
| kd.internalBufferSizes.clear(); | ||
| kd.internalBufferSizes.push_back(prim_params.inputs[0].PhysicalSizeInBytes()); | ||
| kd.internalBufferDataType = prim_params.inputs[0].GetDType(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why do we need this internal buffer?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Applied, frankly no need, it came from convolution_kernel_base.
Signed-off-by: hyunback <[email protected]>
### Details: - Stable Diffusion in dpas has bad first inference latency because all onednn convolutions are compiled at first inference. We can resolve this bottleneck with shape agnostic kernel. Target kernel is convolution_fsv16_1x1 ### Tickets: - *143317* --------- Signed-off-by: hyunback <[email protected]>
### Details: - Stable Diffusion in dpas has bad first inference latency because all onednn convolutions are compiled at first inference. We can resolve this bottleneck with shape agnostic kernel. Target kernel is convolution_fsv16_1x1 ### Tickets: - *143317* --------- Signed-off-by: hyunback <[email protected]>
Details:
Tickets: