Fix xetla import for wheel

Author: ZzEeKkAa

benchmarks/triton_kernels_benchmark/flash_attention_fwd_benchmark.py

@@ -4,7 +4,7 @@
 import triton.language as tl
 
 import triton_kernels_benchmark as benchmark_suit
-import xetla_kernel
+import triton_kernels_benchmark.xetla_kernel
 
 if benchmark_suit.USE_IPEX_OPTION:
     import intel_extension_for_pytorch  # type: ignore # noqa: F401

benchmarks/triton_kernels_benchmark/fused_softmax.py

@@ -13,7 +13,7 @@
 from triton.runtime import driver
 
 import triton_kernels_benchmark as benchmark_suit
-import xetla_kernel
+import triton_kernels_benchmark.xetla_kernel
 
 
 @torch.jit.script

benchmarks/triton_kernels_benchmark/gemm_benchmark.py

@@ -15,7 +15,7 @@
 import triton_kernels_benchmark as benchmark_suit
 from triton_kernels_benchmark.benchmark_testing import do_bench_elapsed_time, BENCHMARKING_METHOD
 
-import xetla_kernel
+import triton_kernels_benchmark.xetla_kernel
 
 if benchmark_suit.USE_IPEX_OPTION:
     import intel_extension_for_pytorch  # type: ignore # noqa: F401

benchmarks/triton_kernels_benchmark/gemm_splitk_benchmark.py

@@ -3,7 +3,7 @@
 import triton.language as tl
 
 import triton_kernels_benchmark as benchmark_suit
-import xetla_kernel
+import triton_kernels_benchmark.xetla_kernel
 
 if benchmark_suit.USE_IPEX_OPTION:
     import intel_extension_for_pytorch  # type: ignore # noqa: F401

benchmarks/triton_kernels_benchmark/gemm_streamk_benchmark.py

@@ -10,7 +10,7 @@
 import triton.language as tl
 
 import triton_kernels_benchmark as benchmark_suit
-import xetla_kernel
+import triton_kernels_benchmark.xetla_kernel
 
 if benchmark_suit.USE_IPEX_OPTION:
     import intel_extension_for_pytorch  # type: ignore # noqa: F401