Fixed GEMM example performance regression and batch GEMM fail issues. #252

	name: "Cuda Test"

	on:
	push:
	branches: [ "sycl-develop" ]
	pull_request:
	branches: [ "sycl-develop" ]
	workflow_dispatch:

	permissions: {}

	jobs:
	run-tests:
	name: Run cuda tests
	runs-on: cp-nvidia-gpu
	timeout-minutes: 120

	steps:
	- name: Checkout repository
	uses: actions/checkout@a5ac7e51b41094c92402da3b24376905380afc29 # v4.1.6

	- name: Build
	shell: bash
	run: \|
	nvidia-smi
	export CUDACXX=/usr/local/cuda/bin/nvcc
	cmake -G Ninja -DCUTLASS_NVCC_ARCHS="90a"
	cmake --build .

	- name: Unit test
	shell: bash
	run: \|
	cmake --build . --target test_unit -j

Provide feedback