[SYCL][Joint Matrix][E2E] Uncomment the Joint Matrix tests for combination 32x32x16 (#16191)

Description:
The support for combination 32x32x16 on DG2 has been implemented in
commit
intel/intel-graphics-compiler@6a06c93.
This PR re-enables the Joint Matrix tests for the combination 32x32x16,
which were previously unsupported.

This PR also adds XFAIL: gpu to two tests,
joint_matrix_bf16_fill_k_cache_arg_dim.cpp and
joint_matrix_bf16_fill_k_cache_runtime_dim.cpp, because the driver has
not yet been updated to include the necessary changes in IGC to support
these tests. As a result, these tests should not be tested in CI at this
time.

Author: YixingZhang007

sycl/test-e2e/Matrix/element_wise_all_ops_impl.hpp

@@ -242,9 +242,9 @@ int main() {
       test_ewops_ab<bfloat16, 8, 16, use::a, layout::row_major, 1>();
       test_ewops_ab<bfloat16, 16, 8, use::b, layout::ext_intel_packed, 2>();
       test_ewops_c<float, 8, 8>();
-      // test_ewops_ab<bfloat16, 32, 16, use::a, layout::row_major, 1>();
-      // test_ewops_ab<bfloat16, 16, 32, use::b, layout::ext_intel_packed, 2>();
-      // test_ewops_c<float, 32, 32>();
+      test_ewops_ab<bfloat16, 32, 16, use::a, layout::row_major, 1>();
+      test_ewops_ab<bfloat16, 16, 32, use::b, layout::ext_intel_packed, 2>();
+      test_ewops_c<float, 32, 32>();
       break;
     }
   }

sycl/test-e2e/Matrix/element_wise_ops_impl.hpp

@@ -142,8 +142,7 @@ int main() {
       passed &= test<uint8_t, int32_t, 8, 8, 32, 4, class dg2_uint_8x8x32>();
       passed &= test<int8_t, int32_t, 8, 8, 32, 4, class dg2_sint_8x8x32>();
       passed &= test<bfloat16, float, 8, 8, 16, 2, class dg2_bf16_8x16x16>();
-      // passed &= test<bfloat16, float, 32, 32, 16, 2, class
-      // dg2_bf16_32x32x16>();
+      passed &= test<bfloat16, float, 32, 32, 16, 2, class dg2_bf16_32x32x16>();
       break;
     }
   }

sycl/test-e2e/Matrix/joint_matrix_bf16_fill_k_cache_arg_dim.cpp

@@ -12,5 +12,10 @@
 
 // -ffp-model=precise is added to not depend on compiler defaults.
 
+// Waiting for the commit in IGC to be pulled into the driver to resolve the
+// test.
+// XFAIL: gpu
+// XFAIL-TRACKER: CMPLRLLVM-63710
+
 #include "common.hpp"
 #include "joint_matrix_bf16_fill_k_cache_impl.hpp"

sycl/test-e2e/Matrix/joint_matrix_bf16_fill_k_cache_impl.hpp

@@ -493,8 +493,8 @@ size_t matrix_size = -1;
 
       test<bfloat16, float, VnniFactor, /*TM*/ 8, /*TN*/ 8, /*TK*/ 16, MCache1,
            NCache1, KCache1, MCache2, NCache2, KCache2>(matrix_size);
-      // test<bfloat16, float, VnniFactor, /*TM*/ 32, /*TN*/ 32, /*TK*/ 16, MCache1,
-      //      NCache1, KCache1, MCache2, NCache2, KCache2>(matrix_size);
+      test<bfloat16, float, VnniFactor, /*TM*/ 32, /*TN*/ 32, /*TK*/ 16,
+           MCache1, NCache1, KCache1, MCache2, NCache2, KCache2>(matrix_size);
       break;
     }
   }

sycl/test-e2e/Matrix/joint_matrix_bf16_fill_k_cache_runtime_dim.cpp

@@ -12,5 +12,10 @@
 
 // -ffp-model=precise is added to not depend on compiler defaults.
 
+// Waiting for the commit in IGC to be pulled into the driver to resolve the
+// test.
+// XFAIL: gpu
+// XFAIL-TRACKER: CMPLRLLVM-63710
+
 #include "common.hpp"
 #include "joint_matrix_bf16_fill_k_cache_impl.hpp"

sycl/test-e2e/Matrix/joint_matrix_rowmajorA_rowmajorB_impl.hpp

@@ -144,8 +144,8 @@ int main() {
           gemm_row_major<8, 8, 32, class su_8x8x32, int8_t, uint8_t, int32_t>();
       res += gemm_row_major<8, 8, 32, class uu_8x8x32, uint8_t, uint8_t,
                             int32_t>();
-      // res += gemm_row_major<32, 32, 16, class dg2_bf16_32x32x16, bfloat16,
-      //                       bfloat16, float>();
+      res += gemm_row_major<32, 32, 16, class dg2_bf16_32x32x16, bfloat16,
+                            bfloat16, float>();
       break;
     }
   }