[SYCL][L0] Support for >4Gb device allocations.

SYCL/Plugin/max_malloc.cpp

@@ -0,0 +1,43 @@
+// REQUIRES: gpu, level_zero, level_zero_dev_kit
+// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %level_zero_options %s -o %t.out
+// RUN: env SYCL_PROGRAM_COMPILE_OPTIONS=-ze-intel-greater-than-4GB-buffer-required %GPU_RUN_PLACEHOLDER %t.out
+
+#include <CL/sycl.hpp>
+#include <iostream>
+
+using namespace cl::sycl;
+
+const double Gb = 1024 * 1024 * 1024;
+
+int main() {
+  auto D = device(gpu_selector());
+
+  std::cout << "name = " << D.get_info<info::device::name>() << std::endl;
+
+  auto global_mem_size = D.get_info<info::device::global_mem_size>() / Gb;
+  std::cout << "global_mem_size = " << global_mem_size << std::endl;
+  std::cout << "max_mem_alloc_size = "
+            << D.get_info<info::device::max_mem_alloc_size>() / Gb << std::endl;
+
+  auto Q = queue(D);
+  for (int I = 1; I < global_mem_size; I++) {
+    void *p;
+    p = malloc_device(I * Gb, Q);
+    std::cout << "malloc_device(" << I << "Gb) = " << p << std::endl;
+    if (p == nullptr) {
+      std::cout << "FAILED" << std::endl;
+      return -1;
+    }
+    sycl::free(p, Q);
+
+    p = malloc_shared(I * Gb, Q);
+    std::cout << "malloc_shared(" << I << "Gb) = " << p << std::endl;
+    if (p == nullptr) {
+      std::cout << "FAILED" << std::endl;
+      return -1;
+    }
+    sycl::free(p, Q);
+  }
+
+  return 0;
+}

SYCL/USM/badmalloc.cpp

@@ -1,5 +1,8 @@
 // UNSUPPORTED: windows
 //
+// XFAIL: level_zero
+// TODO: enable the test when L0 driver is honoring too large sizes.
+//
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t1.out
 // RUN: %HOST_RUN_PLACEHOLDER %t1.out
 // RUN: %CPU_RUN_PLACEHOLDER %t1.out