Introducing NXP Neutron runtime #10563
Conversation
On gcc 13.2 and the uint32_t is undeclared.
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/10563
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ❌ 4 New FailuresAs of commit 1207e08 with merge base 4559a61 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
@pytorchbot label "module: nxp" "release notes: nxp" |
pytorch-bot
bot
added
module: nxp
|
@digantdesai are you the POC for reviewing this on our end? |
| uint32_t numInputs = transpositionFlags[INPUT_TENSOR_FORMAT_LEN_POS]; | ||
| uint32_t numOutputs = transpositionFlags[OUTPUT_TENSOR_FORMAT_LEN_POS]; | ||
| cfg->inputTranspositionFlags = | ||
| INPUT_TENSOR_FORMAT_ARRAY_ADDR(transpositionFlags); |
There was a problem hiding this comment.
Why does this tranpose outside of the ET graph?
Can you explain the logic here as well?
I want to make sure I fully understand, as our application uses channel last (like Neutron) natively
| return true; | ||
| } | ||
|
|
||
| Result<DelegateHandle*> init( |
There was a problem hiding this comment.
isn't there a neutron init function as well? Should this be called within this function as well?
| } | ||
|
|
||
| // Transpose inputs. | ||
| for (int i = 0; i < cfg->numInputs; i++) { |
There was a problem hiding this comment.
@digantdesai is it the case that Executorch input tensors are always NC(H)W? Natively, we use NWC in our applications because it makes more sense for our use case. Is t
| /// - Initialize the Neutron Driver library, setting initial values, do memory | ||
| /// allocation | ||
| /// for internal data structures, do memory mapping. | ||
| NeutronError neutronInit(); |
There was a problem hiding this comment.
can these functions be in a neutron namespace?
Where are we calling this specific function to init the lib?
| using namespace std; | ||
|
|
||
| namespace torch { | ||
| namespace executor { |
|
|
||
| // Applied on outputs. | ||
| template <typename T> | ||
| void transposeToChannelFirst( |
There was a problem hiding this comment.
OK for now, but I wonder is there another way to do this, i.e. set the dim_order on the output and let the portable or someone else take care of this?
|
|
||
| // Applied on inputs. | ||
| template <typename T> | ||
| void transposeToChannelLast( |
|
|
||
| auto* cfg = allocator->allocateInstance<NeutronConfig>(); | ||
|
|
||
| // The following data is read from the "processed" data blob. |
There was a problem hiding this comment.
Nit: make this a helper function?
Summary
Introducing NXP Neutron runtime.
cc @digantdesai @JakeStevens @robert-kalmar