Adding SME1 Convolution Kernel to convole_kleidiai.cpp

[JonathanC-ARM]

Description

• Integration of SME1 Variant of existing SME2 convolution Kernel, kai_run_imatmul_clamp_f32_f32p2vlx1_f32p2vlx1b_2vlx2vl_sme_mopa and associated packing functions
• Formatting changes in convolve_kleidiai.cpp
• Addition of proper sme2 gate for dynamic qgemm
• Updating of kleidiai version to 1.14 (first version which contains the appropriate kernel)

[hariharans29]

/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).

[hariharans29]

Can you please re-base ? @JonathanC-ARM

[hariharans29]

/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).

[hariharans29]

Can you please confirm if the corresponding tests passed on an SME1 machine or using SME1 mode on an SME2 machine ?

[JonathanC-ARM]

Hi Hari, yeah so I've one push to come fixing the sme2 checks. These are going to be changed subsequently via the gemv pr anyway but in the meantime you are correct its right to fix them now.

The sme1 tests pass also, we test on both sme2 and sme1. We typically force sme2 to be false to verify these kernels during development. Please see attached output of the tests for ort test all and mlas tests. Both of which pass and in the case of mlas tests we see that there are a few hundred skipped tests with a message "MlasDynamicQGemmBatch() requires ARM64 SME2 but it was not detected. Skipping test."
ort_mlas_test.txt
ort_test_all.txt

[hariharans29]

/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).