[SYCL][Joint Matrix][E2E] Add few load/store tests for different JointMatrixLoadStoreOpt flag setting (#13873)

This PR adds the following tests:
- element_wise_all_ops verification for
IGC_JointMatrixLoadStoreOpt=2,1,0. It is split to 3 different tests,
since test runtime became too big, so splitting allows to run tests
faster in several threads
- big combinations, B row-major and annotated_ptr tests with different
flag setting (IGC_JointMatrixLoadStoreOpt=0,1) in accordance with the
loads/stores implemented in IGC for those flag settings.

Author: YuriPlyakhin

sycl/test-e2e/Matrix/element_wise_all_ops_1d.cpp

@@ -0,0 +1,14 @@
+//==-------- element_wise_all_ops_1d.cpp - DPC++ joint_matrix ---------==//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// REQUIRES: aspect-ext_intel_matrix, gpu
+
+// RUN: %{build} -o %t.out
+// RUN: env IGC_JointMatrixLoadStoreOpt=1 %{run} %t.out
+
+#include "common.hpp"
+#include "element_wise_all_ops_impl.hpp"

sycl/test-e2e/Matrix/element_wise_all_ops_1d_cont.cpp

@@ -0,0 +1,14 @@
+//==-------- element_wise_all_ops_1d_cont.cpp - DPC++ joint_matrix ---------==//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// REQUIRES: aspect-ext_intel_matrix, gpu
+
+// RUN: %{build} -o %t.out
+// RUN: env IGC_JointMatrixLoadStoreOpt=2 %{run} %t.out
+
+#include "common.hpp"
+#include "element_wise_all_ops_impl.hpp"

sycl/test-e2e/Matrix/element_wise_all_ops_scalar.cpp

@@ -0,0 +1,14 @@
+//==---- element_wise_all_ops_scalar.cpp - DPC++ joint_matrix ------==//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// REQUIRES: aspect-ext_intel_matrix, gpu
+
+// RUN: %{build} -o %t.out
+// RUN: env IGC_JointMatrixLoadStoreOpt=0 %{run} %t.out
+
+#include "common.hpp"
+#include "element_wise_all_ops_impl.hpp"

sycl/test-e2e/Matrix/joint_matrix_annotated_ptr.cpp

@@ -9,6 +9,8 @@
 
 // RUN: %{build} -o %t.out
 // RUN: %{run} %t.out
+// RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=0 %{run} %t.out %}
+// RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=1 %{run} %t.out %}
 
 #include "common.hpp"
 

sycl/test-e2e/Matrix/joint_matrix_bfloat16_packedB.cpp

@@ -11,6 +11,8 @@
 // RUN: %{build} -o %t.out
 // RUN: %{run} %t.out
 // RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=2 %{run} %t.out %}
+// RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=1 %{run} %t.out %}
+// RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=0 %{run} %t.out %}
 
 #include "common.hpp"
 #include "joint_matrix_bfloat16_packedB_impl.hpp"

sycl/test-e2e/Matrix/joint_matrix_int8_vnni.cpp

@@ -9,6 +9,8 @@
 
 // RUN: %{build} -o %t.out
 // RUN: %{run} %t.out
+// RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=0 %{run} %t.out %}
+// RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=1 %{run} %t.out %}
 
 #include "common.hpp"
 

sycl/test-e2e/Matrix/joint_matrix_rowmajorA_rowmajorB.cpp

@@ -12,6 +12,8 @@
 // RUN: %{build} -o %t.out
 // RUN: %{run} %t.out
 // RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=2 %{run} %t.out %}
+// RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=1 %{run} %t.out %}
+// RUN: %if gpu %{ env IGC_JointMatrixLoadStoreOpt=0 %{run} %t.out %}
 
 // This tests support of row major layout for matrix B which does automatic VNNI
 // transform. This is currently only available on AMX and XMX of PVC