[SYCL] Set the feature macro to use the non-use matrix API (#1292)

By default, a new API that has an extra "use" argument will be used.
So for the other tests that use the old API, the macro will have to be explicitly set

Author: dkhaldi

SYCL/Matrix/element_wise_all_ops_bf16.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/element_wise_all_ops_half.cpp

@@ -9,7 +9,7 @@
 
 // Only runs on DPAS because AMX implementation does not support half data type
 // yet
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 
 #include <iostream>

SYCL/Matrix/element_wise_all_ops_int8.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/element_wise_all_ops_int8_packed.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/element_wise_irreg_sum_rows.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/element_wise_ops.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/elemwise_irreg_size_ops_bf16.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // This test is for element wise operations when matrix size does not multiply
 // SG size. This corner case only applies to AMX. Also, it tests bf16 type.
 // only run this on AMX

SYCL/Matrix/joint_matrix_bf16.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/joint_matrix_bfloat16.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/joint_matrix_half.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // Only run on the GPU because half is not supported on AMX hardware
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/joint_matrix_int8_vnni.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/joint_matrix_query_default.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 
 #include <iostream>

SYCL/Matrix/joint_matrix_ss_int8.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/joint_matrix_su_int8.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/joint_matrix_us_int8.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 

SYCL/Matrix/joint_matrix_uu_int8.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 // REQUIRES: matrix
 
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX=1
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out