diff --git a/sycl/test/basic_tests/accessor/accessor_property_list_rt.cpp b/sycl/test/basic_tests/accessor/accessor_property_list_rt.cpp
index 1fc8384407043..2c28b0bc4a4cf 100644
--- a/sycl/test/basic_tests/accessor/accessor_property_list_rt.cpp
+++ b/sycl/test/basic_tests/accessor/accessor_property_list_rt.cpp
@@ -1,5 +1,5 @@
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 
 #include <sycl/sycl.hpp>
 
diff --git a/sycl/test/basic_tests/built-ins.cpp b/sycl/test/basic_tests/built-ins.cpp
deleted file mode 100644
index 9b13464678e51..0000000000000
--- a/sycl/test/basic_tests/built-ins.cpp
+++ /dev/null
@@ -1,64 +0,0 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-
-// CUDA does not support printf.
-// UNSUPPORTED: cuda
-//
-// Hits an assertion with AMD:
-// XFAIL: hip_amd
-
-#include <sycl/sycl.hpp>
-
-#include <cassert>
-
-namespace s = sycl;
-
-// According to OpenCL C spec, the format string must be in constant address
-// space
-#ifdef __SYCL_DEVICE_ONLY__
-#define CONSTANT __attribute__((opencl_constant))
-#else
-#define CONSTANT
-#endif
-
-static const CONSTANT char format[] = "Hello, World! %d %f\n";
-
-int main() {
-  s::queue q{};
-
-  // Test printf
-  q.submit([&](s::handler &CGH) {
-     CGH.single_task<class printf>([=]() {
-       s::ext::oneapi::experimental::printf(format, 123, 1.23);
-       // CHECK: {{(Hello, World! 123 1.23)?}}
-     });
-   }).wait();
-
-  s::ext::oneapi::experimental::printf(format, 321, 3.21);
-  // CHECK: {{(Hello, World! 123 1.23)?}}
-
-  // Test common
-  {
-    s::buffer<s::cl_float, 1> BufMin(s::range<1>(1));
-    s::buffer<s::cl_float2, 1> BufMax(s::range<1>(1));
-    q.submit([&](s::handler &cgh) {
-      auto AccMin = BufMin.get_access<s::access::mode::write>(cgh);
-      auto AccMax = BufMax.get_access<s::access::mode::write>(cgh);
-      cgh.single_task<class common>([=]() {
-        AccMax[0] = s::max(s::cl_float2{0.5f, 2.5}, s::cl_float2{2.3f, 2.3});
-        AccMin[0] = s::min(s::cl_float{0.5f}, s::cl_float{2.3f});
-      });
-    });
-
-    auto AccMin = BufMin.template get_access<s::access::mode::read>();
-    auto AccMax = BufMax.template get_access<s::access::mode::read>();
-
-    assert(AccMin[0] == 0.5);
-    assert(AccMax[0].x() == 2.3f && AccMax[0].y() == 2.5f);
-    assert(s::min(0.5f, 2.3f) == 0.5);
-    auto Res = s::max(s::int4{5, 2, 1, 5}, s::int4{3, 3, 4, 2});
-    assert(Res.x() == 5 && Res.y() == 3 && Res.z() == 4 && Res.w() == 5);
-  }
-
-  return 0;
-}
diff --git a/sycl/test/basic_tests/context.cpp b/sycl/test/basic_tests/context.cpp
deleted file mode 100644
index 38b8dcbab5d1f..0000000000000
--- a/sycl/test/basic_tests/context.cpp
+++ /dev/null
@@ -1,84 +0,0 @@
-// RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-
-// This test performs basic check of the SYCL context class.
-
-#include <sycl/sycl.hpp>
-#include <iostream>
-
-using namespace sycl;
-
-int main() {
-  try {
-    context c;
-  } catch (device_error e) {
-    std::cout << "Failed to create device for context" << std::endl;
-  }
-
-  auto devices = device::get_devices();
-  device &deviceA = devices[0];
-  device &deviceB = (devices.size() > 1 ? devices[1] : devices[0]);
-  {
-    std::cout << "move constructor" << std::endl;
-    context Context(deviceA);
-    size_t hash = std::hash<context>()(Context);
-    context MovedContext(std::move(Context));
-    assert(hash == std::hash<context>()(MovedContext));
-    assert(deviceA.is_host() == MovedContext.is_host());
-  }
-  {
-    std::cout << "move assignment operator" << std::endl;
-    context Context(deviceA);
-    size_t hash = std::hash<context>()(Context);
-    context WillMovedContext(deviceB);
-    WillMovedContext = std::move(Context);
-    assert(hash == std::hash<context>()(WillMovedContext));
-    assert(deviceA.is_host() == WillMovedContext.is_host());
-  }
-  {
-    std::cout << "copy constructor" << std::endl;
-    context Context(deviceA);
-    size_t hash = std::hash<context>()(Context);
-    context ContextCopy(Context);
-    assert(hash == std::hash<context>()(Context));
-    assert(hash == std::hash<context>()(ContextCopy));
-    assert(Context == ContextCopy);
-    assert(Context.is_host() == ContextCopy.is_host());
-  }
-  {
-    std::cout << "copy assignment operator" << std::endl;
-    context Context(deviceA);
-    size_t hash = std::hash<context>()(Context);
-    context WillContextCopy(deviceB);
-    WillContextCopy = Context;
-    assert(hash == std::hash<context>()(Context));
-    assert(hash == std::hash<context>()(WillContextCopy));
-    assert(Context == WillContextCopy);
-    assert(Context.is_host() == WillContextCopy.is_host());
-  }
-  {
-    auto AsyncHandler = [](const sycl::exception_list &EL) {};
-    sycl::context Context1(sycl::property_list{});
-    sycl::context Context2(AsyncHandler, sycl::property_list{});
-    sycl::context Context3(deviceA, sycl::property_list{});
-    sycl::context Context4(deviceA, AsyncHandler, sycl::property_list{});
-    sycl::context Context5(deviceA.get_platform(), sycl::property_list{});
-    sycl::context Context6(deviceA.get_platform(), AsyncHandler,
-                           sycl::property_list{});
-    sycl::context Context7(std::vector<sycl::device>{deviceA},
-                           sycl::property_list{});
-    sycl::context Context8(
-        std::vector<sycl::device>{deviceA}, AsyncHandler,
-        sycl::property_list{
-            sycl::ext::oneapi::cuda::property::context::use_primary_context{}});
-
-    if (!Context8.has_property<
-            sycl::ext::oneapi::cuda::property::context::use_primary_context>()) {
-      std::cerr << "Line " << __LINE__ << ": Property was not found"
-                << std::endl;
-      return 1;
-    }
-
-    auto Prop = Context8.get_property<sycl::ext::oneapi::cuda::property::context::use_primary_context>();
-  }
-}
diff --git a/sycl/test/basic_tests/device.cpp b/sycl/test/basic_tests/device.cpp
deleted file mode 100644
index b313fcf8c173c..0000000000000
--- a/sycl/test/basic_tests/device.cpp
+++ /dev/null
@@ -1,102 +0,0 @@
-// RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-
-// This test performs basic check of the SYCL device class.
-
-#include <cassert>
-#include <iostream>
-#include <sycl/sycl.hpp>
-#include <utility>
-
-using namespace sycl;
-
-std::string get_type(const device &dev) {
-  if (dev.is_host()) {
-    return "host";
-  } else if (dev.is_gpu()) {
-    return "OpenCL.GPU";
-  } else if (dev.is_accelerator()) {
-    return "OpenCL.ACC";
-  } else {
-    return "OpenCL.CPU";
-  }
-}
-
-int main() {
-  device d;
-  std::cout << "Default device type: " << get_type(d) << std::endl;
-
-  int i = 1;
-  std::cout << "Get all devices in the system" << std::endl;
-  for (const auto &dev : device::get_devices()) {
-    std::cout << "Device " << i++ << " is available: " << get_type(dev)
-              << std::endl;
-  }
-  i = 1;
-  std::cout << "Get host devices in the system" << std::endl;
-  for (const auto &dev : device::get_devices(info::device_type::host)) {
-    std::cout << "Device " << i++ << " is available: " << get_type(dev)
-              << std::endl;
-  }
-  i = 1;
-  std::cout << "Get OpenCL.CPU devices in the system" << std::endl;
-  for (const auto &dev : device::get_devices(info::device_type::cpu)) {
-    std::cout << "Device " << i++ << " is available: " << get_type(dev)
-              << std::endl;
-  }
-  i = 1;
-  std::cout << "Get OpenCL.GPU devices in the system" << std::endl;
-  for (const auto &dev : device::get_devices(info::device_type::gpu)) {
-    std::cout << "Device " << i++ << " is available: " << get_type(dev)
-              << std::endl;
-  }
-  i = 1;
-  std::cout << "Get OpenCL.ACC devices in the system" << std::endl;
-  for (const auto &dev : device::get_devices(info::device_type::accelerator)) {
-    std::cout << "Device " << i++ << " is available: " << get_type(dev)
-              << std::endl;
-  }
-
-  auto devices = device::get_devices();
-  device &deviceA = devices[0];
-  device &deviceB = (devices.size() > 1 ? devices[1] : devices[0]);
-  {
-    std::cout << "move constructor" << std::endl;
-    device Device(deviceA);
-    size_t hash = std::hash<device>()(Device);
-    device MovedDevice(std::move(Device));
-    assert(hash == std::hash<device>()(MovedDevice));
-    assert(deviceA.is_host() == MovedDevice.is_host());
-  }
-  {
-    std::cout << "move assignment operator" << std::endl;
-    device Device(deviceA);
-    size_t hash = std::hash<device>()(Device);
-    device WillMovedDevice(deviceB);
-    WillMovedDevice = std::move(Device);
-    assert(hash == std::hash<device>()(WillMovedDevice));
-    assert(deviceA.is_host() == WillMovedDevice.is_host());
-  }
-  {
-    std::cout << "copy constructor" << std::endl;
-    device Device(deviceA);
-    size_t hash = std::hash<device>()(Device);
-    device DeviceCopy(Device);
-    assert(hash == std::hash<device>()(Device));
-    assert(hash == std::hash<device>()(DeviceCopy));
-    assert(Device == DeviceCopy);
-    assert(Device.is_host() == DeviceCopy.is_host());
-  }
-  {
-    std::cout << "copy assignment operator" << std::endl;
-    device Device(deviceA);
-    size_t hash = std::hash<device>()(Device);
-    device WillDeviceCopy(deviceB);
-    WillDeviceCopy = Device;
-    assert(hash == std::hash<device>()(Device));
-    assert(hash == std::hash<device>()(WillDeviceCopy));
-    assert(Device == WillDeviceCopy);
-    assert(Device.is_host() == WillDeviceCopy.is_host());
-  }
-}
-
diff --git a/sycl/test/basic_tests/event_async_exception.cpp b/sycl/test/basic_tests/event_async_exception.cpp
deleted file mode 100644
index 5064952341261..0000000000000
--- a/sycl/test/basic_tests/event_async_exception.cpp
+++ /dev/null
@@ -1,40 +0,0 @@
-// RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-
-//==---- event_async_exception.cpp - Test for event async exceptions -------==//
-//
-// 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
-//
-//===----------------------------------------------------------------------===//
-
-#include <sycl/sycl.hpp>
-
-// This test checks that if there is a submit failure, the asynchronous
-// exception is associated with the returned event.
-
-using namespace sycl;
-
-class KernelName;
-
-int main() {
-  auto asyncHandler = [](exception_list el) {
-    for (auto &e : el) {
-      std::rethrow_exception(e);
-    }
-  };
-
-  queue q(asyncHandler);
-
-  try {
-    // Check that submitting a CG with no kernel or memory operation doesn't produce
-    // an async exception
-    event e = q.submit([&](handler &cgh) {});
-
-    e.wait_and_throw();
-    return 0;
-  } catch (runtime_error e) {
-    return 1;
-  }
-}
diff --git a/sycl/test/basic_tests/exceptions-SYCL-2020.cpp b/sycl/test/basic_tests/exceptions-SYCL-2020.cpp
index 6821fde06eefd..4b351880c7b63 100644
--- a/sycl/test/basic_tests/exceptions-SYCL-2020.cpp
+++ b/sycl/test/basic_tests/exceptions-SYCL-2020.cpp
@@ -1,5 +1,5 @@
 // RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 
 #include <sycl/sycl.hpp>
 
diff --git a/sycl/test/basic_tests/host_image_accessor_read.cpp b/sycl/test/basic_tests/host_image_accessor_read.cpp
index eb53b730dfdcb..b70e1a3a76761 100644
--- a/sycl/test/basic_tests/host_image_accessor_read.cpp
+++ b/sycl/test/basic_tests/host_image_accessor_read.cpp
@@ -1,5 +1,5 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %t.out
 
 //==---- host_image_accessor_read.cpp - SYCL host image accessor check ----==//
 //
@@ -22,7 +22,6 @@ int foo(float *image_data) {
   sycl::range<3> r(3, 3, 3);
   {
     sycl::buffer<int, 1> ResultBuf(result, sycl::range<1>(2));
-    sycl::queue Q;
     sycl::image<3> Image(image_data, channelOrder, channelType, r);
 
     sycl::range<2> pitch = Image.get_pitch();
diff --git a/sycl/test/basic_tests/known_identity.cpp b/sycl/test/basic_tests/known_identity.cpp
index edfc8810eb33a..e047928a286af 100644
--- a/sycl/test/basic_tests/known_identity.cpp
+++ b/sycl/test/basic_tests/known_identity.cpp
@@ -1,5 +1,5 @@
 // RUN: %clangxx -fsycl -Xclang -verify %s -Xclang -verify-ignore-unexpected=note,warning -o %t.out -std=c++17
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 // expected-no-diagnostics
 
 // This test performs basic checks of has_known_identity and known_identity
diff --git a/sycl/test/basic_tests/linear-host-dev.cpp b/sycl/test/basic_tests/linear-host-dev.cpp
deleted file mode 100644
index 75e80c3461ef6..0000000000000
--- a/sycl/test/basic_tests/linear-host-dev.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-// RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out | FileCheck %s
-
-#include <algorithm>
-#include <cstdio>
-#include <cstdlib>
-#include <sycl/sycl.hpp>
-
-// Check that linear id is monotonically increased on host device.
-// Only there we can reliable check that. Since the kernel has a restriction
-// regarding usage of global variables, use stream to log the linear id
-// and ensure that they're monotonically increased.
-//
-// Note: This test heavily relies on the current implementation of
-// host device(single-threaded ordered execution). So if the implementation
-// is somehow changed so it's no longer possible to run this test reliable
-// it can be removed.
-
-namespace s = sycl;
-
-int main(int argc, char *argv[]) {
-  s::queue q;
-
-  const size_t outer = 3;
-  const size_t inner = 2;
-  const s::range<2> rng = {outer, inner};
-
-  q.submit([&](s::handler &h) {
-    s::stream out(1024, 80, h);
-
-    h.parallel_for<class linear_id>(s::range<2>(rng), [=](s::item<2> item) {
-      // CHECK: 0
-      // CHECK-NEXT: 1
-      // CHECK-NEXT: 2
-      // CHECK-NEXT: 3
-      // CHECK-NEXT: 4
-      // CHECK-NEXT: 5
-      out << item.get_linear_id() << sycl::endl;
-    });
-  });
-  q.wait();
-
-  return 0;
-}
diff --git a/sycl/test/basic_tests/marray/marray.cpp b/sycl/test/basic_tests/marray/marray.cpp
index 3e69d98ad61a8..40905edc4e078 100755
--- a/sycl/test/basic_tests/marray/marray.cpp
+++ b/sycl/test/basic_tests/marray/marray.cpp
@@ -1,5 +1,5 @@
 // RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 
 //==--------------- marray.cpp - SYCL marray test --------------------------==//
 //
diff --git a/sycl/test/basic_tests/offset-accessor-get_pointer.cpp b/sycl/test/basic_tests/offset-accessor-get_pointer.cpp
deleted file mode 100644
index e0b6941e4787e..0000000000000
--- a/sycl/test/basic_tests/offset-accessor-get_pointer.cpp
+++ /dev/null
@@ -1,113 +0,0 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple  %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-
-// Per the SYCL 2020 spec (4.7.6.12 and others)
-// accessor::get_pointer() returns a pointer to the start of this accessor’s
-// memory. For a buffer accessor this is a pointer to the start of the
-// underlying buffer, even if this is a ranged accessor whose range does not
-// start at the beginning of the buffer.
-
-// This is a departure from how get_pointer() was interpreted with offset
-// accessors in the past. Not relevant for images, which do not support offset
-// accessors.
-
-#include <sycl/sycl.hpp>
-#include <vector>
-using namespace sycl;
-
-void test_across_ranges() {
-  constexpr auto r_w = access::mode::read_write;
-  constexpr unsigned long width = 4;
-  constexpr unsigned long count = width * width;
-  constexpr unsigned long count3D = width * width * width; // 64
-  std::vector<int> v1(count);                              // for 1D testing.
-  std::vector<int> v2(count);                              // for 2D testing.
-  std::vector<int> v3(count3D);                            // 3D
-
-  range<1> range_1D(count);
-  range<2> range_2D(width, width);
-  range<3> range_3D(width, width, width);
-
-  queue myQueue;
-  {
-    // 1D, 2D, 3D
-    buffer<int> buf_1D(v1.data(), count);
-    buffer<int, 2> buf_2D(v2.data(), range_2D);
-    buffer<int, 3> buf_3D(v3.data(), range_3D);
-
-    myQueue.submit([&](handler &cgh) {
-      auto acc_1D = buf_1D.get_access<r_w>(cgh, {2}, {10});
-      auto acc_2D = buf_2D.get_access<r_w>(cgh, {2, 2}, {1, 1});
-      auto acc_3D = buf_3D.get_access<r_w>(cgh, {2, 2, 2}, {1, 1, 1});
-      cgh.single_task<class task>([=] {
-        acc_1D.get_pointer()[0] = 5; // s.b. offset 0
-        acc_1D[0] = 15;              // s.b. offset 10
-
-        // 2D
-        acc_2D.get_pointer()[0] = 7; // s.b. offset 0
-        acc_2D[{0, 0}] = 17;         // s.b. offset {1,1} aka 5 if linear.
-
-        // 3D
-        acc_3D.get_pointer()[0] = 9; // s.b. offset 0
-        acc_3D[{0, 0, 0}] = 19;      // s.b. offset {1,1,1} aka 21 if linear.
-      });
-    });
-    myQueue.wait();
-    // now host access - we offset by one more than the device test
-    auto acc_1D = buf_1D.get_access<r_w>({2}, {11});
-    auto acc_2D = buf_2D.get_access<r_w>({2, 2}, {1, 2});
-    auto acc_3D = buf_3D.get_access<r_w>({2, 2, 2}, {1, 1, 2});
-    acc_1D.get_pointer()[1] = 4; // s.b. offset 1
-    acc_1D[0] = 14;              // s.b. offset 11
-
-    // 2D
-    acc_2D.get_pointer()[1] = 6; // s.b. offset 1
-    acc_2D[{0, 0}] = 16;         // s.b. offset {1,2} aka 6 if linear.
-
-    // 3D
-    acc_3D.get_pointer()[1] = 8; // s.b. offset 1
-    acc_3D[{0, 0, 0}] = 18;      // s.b. offset {1,1,2} aka 22 if linear.
-  }                              //~buffer
-  // always nice to have some feedback
-  std::cout << "DEVICE" << std::endl;
-  std::cout << "1D CHECK:  v1[0] should be 5: " << v1[0]
-            << ", and v1[10] s.b. 15: " << v1[10] << std::endl;
-  std::cout << "2D CHECK:  v2[0] should be 7: " << v2[0]
-            << ", and v2[5]  s.b. 17: " << v2[5] << std::endl;
-  std::cout << "3D CHECK:  v3[0] should be 9: " << v3[0]
-            << ", and v3[21] s.b. 19: " << v3[21] << std::endl
-            << std::endl;
-
-  std::cout << "HOST" << std::endl;
-  std::cout << "1D CHECK:  v1[1] should be 4: " << v1[1]
-            << ", and v1[11] s.b. 14: " << v1[11] << std::endl;
-  std::cout << "2D CHECK:  v2[1] should be 6: " << v2[1]
-            << ", and v2[6]  s.b. 16: " << v2[6] << std::endl;
-  std::cout << "3D CHECK:  v3[1] should be 8: " << v3[1]
-            << ", and v3[22] s.b. 17: " << v3[22] << std::endl
-            << std::endl;
-
-  // device
-  assert(v1[0] == 5);
-  assert(v1[10] == 15);
-  assert(v2[0] == 7);
-  assert(v2[5] == 17); // offset {1,1} in a 4x4 field is linear offset 5
-  assert(v3[0] == 9);
-  assert(v3[21] == 19); // offset {1,1,1} in a 4x4x4 field is linear offset 21
-
-  // host
-  assert(v1[1] == 4);
-  assert(v1[11] == 14);
-  assert(v2[1] == 6);
-  assert(v2[6] == 16); // offset {1,2} in a 4x4 field is linear offset 6
-  assert(v3[1] == 8);
-  assert(v3[22] == 18); // offset {1,1,2} in a 4x4x4 field is linear offset 22
-}
-
-int main() {
-  test_across_ranges();
-
-  std::cout << "OK!" << std::endl;
-
-  return 0;
-}
diff --git a/sycl/test/basic_tests/parallel_for_range_host.cpp b/sycl/test/basic_tests/parallel_for_range_host.cpp
deleted file mode 100644
index 81f14ec55920a..0000000000000
--- a/sycl/test/basic_tests/parallel_for_range_host.cpp
+++ /dev/null
@@ -1,138 +0,0 @@
-// RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-
-#include <sycl/sycl.hpp>
-
-#include <iostream>
-
-using namespace sycl;
-
-int main() {
-  auto AsyncHandler = [](exception_list ES) {
-    for (auto& E : ES) {
-      std::rethrow_exception(E);
-    }
-  };
-
-  queue Q(AsyncHandler);
-
-  // parallel_for, offset
-  try {
-    Q.submit([&](handler &CGH) {
-      CGH.parallel_for<class offset>(range<1>(1), id<1>(16),
-                                     [=](id<1> ID) { assert(ID == 16); });
-    });
-    Q.submit([&](handler &CGH) {
-      CGH.parallel_for<class offset_2D>(range<2>(1, 1), id<2>(16, 17),
-                                        [=](id<2> ID) {
-                                          assert(ID[0] == 16);
-                                          assert(ID[1] == 17);
-                                        });
-    });
-    Q.submit([&](handler &CGH) {
-      CGH.parallel_for<class offset_3D>(range<3>(1, 1, 1), id<3>(16, 17, 18),
-                                        [=](id<3> ID) {
-                                          assert(ID[0] == 16);
-                                          assert(ID[1] == 17);
-                                          assert(ID[2] == 18);
-                                        });
-    });
-    Q.wait_and_throw();
-  } catch (nd_range_error) {
-    std::cerr << "Test case 'offset' failed: exception has been thrown"
-              << std::endl;
-    return 1;
-  }
-
-  // parallel_for, 100 global, 3 local -> fail.
-  try {
-    Q.submit([&](handler &CGH) {
-        CGH.parallel_for<class a>(nd_range<1>(range<1>(100), range<1>(3)),
-            [=](nd_item<1>) {});
-    });
-    Q.wait_and_throw();
-    std::cerr << "Test case 'a' failed: no exception has been thrown"
-              << std::endl;
-    return 1;
-  } catch (nd_range_error) {
-    // We expect an error to be thrown!
-  }
-
-  // parallel_for, 100 global, 4 local -> pass.
-  try {
-    Q.submit([&](handler &CGH) {
-        CGH.parallel_for<class b>(nd_range<1>(range<1>(100), range<1>(4)),
-            [=](nd_item<1>) {});
-    });
-    Q.wait_and_throw();
-  } catch (nd_range_error) {
-    std::cerr << "Test case 'b' failed: exception has been thrown" << std::endl;
-    return 1;
-  }
-
-  // parallel_for, (100, 33, 16) global, (2, 3, 4) local -> pass.
-  try {
-    Q.submit([&](handler &CGH) {
-        CGH.parallel_for<class c>(nd_range<3>(range<3>(100, 33, 16),
-              range<3>(2, 3, 4)),
-            [=](nd_item<3>) {});
-    });
-    Q.wait_and_throw();
-  } catch (nd_range_error) {
-    std::cerr << "Test case 'c' failed: exception has been thrown" << std::endl;
-    return 1;
-  }
-
-  // parallel_for, (100, 33, 16) global, (2, 3, 5) local -> fail.
-  try {
-    Q.submit([&](handler &CGH) {
-        CGH.parallel_for<class d>(nd_range<3>(range<3>(100, 33, 16),
-              range<3>(2, 3, 5)),
-            [=](nd_item<3>) {});
-    });
-    Q.wait_and_throw();
-    std::cerr << "Test case 'd' failed: no exception has been thrown"
-              << std::endl;
-    return 1;
-  } catch (nd_range_error) {
-  }
-
-  // local size has a 0-based range -- no SIGFPEs, we hope.
-  try {
-    Q.submit([&](handler &CGH) {
-        CGH.parallel_for<class e>(nd_range<2>(range<2>(5, 33), range<2>(1, 0)),
-            [=](nd_item<2>) {});
-    });
-    Q.wait_and_throw();
-    std::cerr << "Test case 'e' failed: no exception has been thrown"
-              << std::endl;
-    return 1;
-  } catch (nd_range_error) {
-  }
-
-  // parallel_for_work_group with 0-based local range.
-  try {
-    Q.submit([&](handler &CGH) {
-        CGH.parallel_for_work_group<class f>(range<2>(5, 33), range<2>(1, 0),
-            [=](group<2>) {});
-    });
-    Q.wait_and_throw();
-    std::cerr << "Test case 'f' failed: no exception has been thrown"
-              << std::endl;
-    return 1;
-  } catch (nd_range_error) {
-  }
-
-  // parallel_for, 30 global, 1(implicit) local -> pass.
-  try {
-    Q.submit([&](handler &CGH) {
-      CGH.parallel_for<class g>(range<1>(30),
-                                [=](id<1>) {});
-    });
-    Q.wait_and_throw();
-  } catch (nd_range_error) {
-    std::cerr << "Test case 'g' failed: exception has been thrown" << std::endl;
-    return 1;
-  }
-  return 0;
-}
diff --git a/sycl/test/basic_tests/parallel_for_user_types.cpp b/sycl/test/basic_tests/parallel_for_user_types.cpp
index b050ed2b778d4..c04749f36c1db 100644
--- a/sycl/test/basic_tests/parallel_for_user_types.cpp
+++ b/sycl/test/basic_tests/parallel_for_user_types.cpp
@@ -1,5 +1,4 @@
-// RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %clangxx -fsycl %s -c -o %t.o
 
 // This test performs basic check of supporting user defined class that are
 // implicitly converted from sycl::item/sycl::nd_item in parallel_for.
diff --git a/sycl/test/basic_tests/property_list.cpp b/sycl/test/basic_tests/property_list.cpp
index d28b98cca9d2d..199870b0c8829 100644
--- a/sycl/test/basic_tests/property_list.cpp
+++ b/sycl/test/basic_tests/property_list.cpp
@@ -1,7 +1,5 @@
 // RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-//
-// CHECK: PASSED
+// RUN: %t.out
 
 // This test performs basic check of the SYCL property_list class.
 
diff --git a/sycl/test/basic_tests/reduction_ctor.cpp b/sycl/test/basic_tests/reduction_ctor.cpp
deleted file mode 100644
index 655be96e023d9..0000000000000
--- a/sycl/test/basic_tests/reduction_ctor.cpp
+++ /dev/null
@@ -1,166 +0,0 @@
-// RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-
-// This performs basic checks such as reduction creation, getIdentity() method,
-// and the combine() method of the aux class 'reducer'.
-
-#include "reduction_utils.hpp"
-#include <cassert>
-#include <sycl/sycl.hpp>
-
-using namespace sycl;
-
-bool toBool(bool V) { return V; }
-bool toBool(vec<int, 2> V) { return V.x() && V.y(); }
-bool toBool(vec<int, 4> V) { return V.x() && V.y() && V.z() && V.w(); }
-
-template <typename T, typename Reduction>
-void test_reducer(Reduction &Redu, T A, T B) {
-  typename Reduction::reducer_type Reducer;
-  Reducer.combine(A);
-  Reducer.combine(B);
-
-  typename Reduction::binary_operation BOp;
-  T ExpectedValue = BOp(A, B);
-  assert(ExpectedValue == Reducer.MValue &&
-         "Wrong result of binary operation.");
-}
-
-template <typename T, typename Reduction, typename BinaryOperation>
-void test_reducer(Reduction &Redu, T Identity, BinaryOperation BOp, T A, T B) {
-  typename Reduction::reducer_type Reducer(Identity, BOp);
-  Reducer.combine(A);
-  Reducer.combine(B);
-
-  T ExpectedValue = BOp(A, B);
-  assert(toBool(ExpectedValue == Reducer.MValue) &&
-         "Wrong result of binary operation.");
-}
-
-template <typename... Ts> class KernelNameGroup;
-
-template <typename SpecializationKernelName, typename T, int Dim,
-          class BinaryOperation>
-void testKnown(T Identity, BinaryOperation BOp, T A, T B) {
-  static_assert(has_known_identity<BinaryOperation, T>::value);
-  queue Q;
-  buffer<T, 1> ReduBuf(1);
-  T *ReduUSMPtr = malloc_host<T>(1, Q);
-
-  Q.submit([&](handler &CGH) {
-    // Reduction needs a device accessor as a parameter.
-    // This accessor is not really used in this test.
-    accessor<T, Dim, access::mode::read_write, access::target::device>
-        ReduRWAcc(ReduBuf, CGH);
-    accessor<T, Dim, access::mode::discard_write, access::target::device>
-        ReduDWAcc(ReduBuf, CGH);
-    auto Redu = sycl::reduction(ReduBuf, CGH, BOp);
-    auto ReduUSM = sycl::reduction(ReduUSMPtr, BOp);
-
-    assert(toBool(Redu.getIdentity() == Identity) &&
-           toBool(ReduUSM.getIdentity() == Identity) &&
-           toBool(known_identity<BinaryOperation, T>::value == Identity) &&
-           "Failed getIdentity() check().");
-    test_reducer(Redu, A, B);
-    test_reducer(ReduUSM, A, B);
-
-    test_reducer(Redu, Identity, BOp, A, B);
-    test_reducer(ReduUSM, Identity, BOp, A, B);
-
-    // Command group must have at least one task in it. Use an empty one.
-    CGH.single_task<SpecializationKernelName>([=]() {});
-  });
-  free(ReduUSMPtr, Q);
-}
-
-template <typename SpecializationKernelName, typename T, int Dim,
-          class BinaryOperation>
-void testUnknown(T Identity, BinaryOperation BOp, T A, T B) {
-  queue Q;
-  buffer<T, 1> ReduBuf(1);
-  T *ReduUSMPtr = malloc_host<T>(1, Q);
-  Q.submit([&](handler &CGH) {
-    // Reduction needs a device accessor as a parameter.
-    // This accessor is not really used in this test.
-    accessor<T, Dim, access::mode::read_write, access::target::device>
-        ReduRWAcc(ReduBuf, CGH);
-    accessor<T, Dim, access::mode::discard_write, access::target::device>
-        ReduDWAcc(ReduBuf, CGH);
-    auto Redu = sycl::reduction(ReduBuf, CGH, Identity, BOp);
-    auto ReduUSM = sycl::reduction(ReduUSMPtr, Identity, BOp);
-    assert(toBool(Redu.getIdentity() == Identity) &&
-           toBool(ReduUSM.getIdentity() == Identity) &&
-           "Failed getIdentity() check().");
-    test_reducer(Redu, Identity, BOp, A, B);
-    test_reducer(ReduUSM, Identity, BOp, A, B);
-
-    // Command group must have at least one task in it. Use an empty one.
-    CGH.single_task<SpecializationKernelName>([=]() {});
-  });
-  free(ReduUSMPtr, Q);
-}
-
-template <typename SpecializationKernelName, typename T, class BinaryOperation>
-void testBoth(T Identity, BinaryOperation BOp, T A, T B) {
-  testKnown<KernelNameGroup<SpecializationKernelName,
-                            class KernelName_SpronAvHpacKFL>,
-            T, 0>(Identity, BOp, A, B);
-  testKnown<
-      KernelNameGroup<SpecializationKernelName, class KernelName_XFxrYatPJlU>,
-      T, 1>(Identity, BOp, A, B);
-  testUnknown<
-      KernelNameGroup<SpecializationKernelName, class KernelName_oUFYMyQSlL>, T,
-      0>(Identity, BOp, A, B);
-  testUnknown<KernelNameGroup<SpecializationKernelName, class KernelName_Ndbp>,
-              T, 1>(Identity, BOp, A, B);
-}
-
-int main() {
-  testBoth<class KernelName_DpWavJTNjhJtrHmLWt, int>(
-      0, ext::oneapi::plus<int>(), 1, 7);
-  testBoth<class KernelName_MHRtc, int>(1, std::multiplies<int>(), 1, 7);
-  testBoth<class KernelName_eYhurMyKBZvzctmqwUZ, int>(
-      0, ext::oneapi::bit_or<int>(), 1, 8);
-  testBoth<class KernelName_DpVPIUBjUMGZEwBFHH, int>(
-      0, ext::oneapi::bit_xor<int>(), 7, 3);
-  testBoth<class KernelName_vGKFactgrkngMXd, int>(
-      ~0, ext::oneapi::bit_and<int>(), 7, 3);
-  testBoth<class KernelName_GLpknSBxclKWjm, int>(
-      (std::numeric_limits<int>::max)(), ext::oneapi::minimum<int>(), 7, 3);
-  testBoth<class KernelName_EvOaOYQ, int>((std::numeric_limits<int>::min)(),
-                                          ext::oneapi::maximum<int>(), 7, 3);
-
-  testBoth<class KernelName_iFbcoTtPeDtUEK, float>(
-      0, ext::oneapi::plus<float>(), 1, 7);
-  testBoth<class KernelName_PEMJanstdNezDSXnP, float>(
-      1, std::multiplies<float>(), 1, 7);
-  testBoth<class KernelName_wOEuftXSjCLpoTOMrYHR, float>(
-      getMaximumFPValue<float>(), ext::oneapi::minimum<float>(), 7, 3);
-  testBoth<class KernelName_HzFCIZQKeV, float>(
-      getMinimumFPValue<float>(), ext::oneapi::maximum<float>(), 7, 3);
-
-  testUnknown<class KernelName_sJOZPgFeiALyqwIWnFP, CustomVec<float>, 0,
-              CustomVecPlus<float>>(CustomVec<float>(0), CustomVecPlus<float>(),
-                                    CustomVec<float>(1), CustomVec<float>(7));
-  testUnknown<class KernelName_jMA, CustomVec<float>, 1>(
-      CustomVec<float>(0), CustomVecPlus<float>(), CustomVec<float>(1),
-      CustomVec<float>(7));
-
-  testUnknown<class KernelName_zhF, int, 0>(
-      0, [](auto a, auto b) { return a | b; }, 1, 8);
-
-  int2 IdentityI2 = {0, 0};
-  int2 AI2 = {1, 2};
-  int2 BI2 = {7, 13};
-  testUnknown<class KNI2, int2, 0>(IdentityI2, ext::oneapi::plus<int2>(), AI2,
-                                   BI2);
-
-  float4 IdentityF4 = {0, 0, 0, 0};
-  float4 AF4 = {1, 2, -1, -34};
-  float4 BF4 = {7, 13, 0, 35};
-  testUnknown<class KNF4, float4, 0>(IdentityF4, ext::oneapi::plus<>(), AF4,
-                                     BF4);
-
-  std::cout << "Test passed\n";
-  return 0;
-}
diff --git a/sycl/test/basic_tests/reduction_utils.hpp b/sycl/test/basic_tests/reduction_utils.hpp
deleted file mode 100644
index 88198d6cb146b..0000000000000
--- a/sycl/test/basic_tests/reduction_utils.hpp
+++ /dev/null
@@ -1,68 +0,0 @@
-#include <sycl/sycl.hpp>
-
-using namespace sycl;
-
-// Initializes 'InBuf' buffer with pseudo-random values, computes the reduction
-// value for the buffer and writes it to 'ExpectedOut'.
-template <typename T, class BinaryOperation>
-void initInputData(buffer<T, 1> &InBuf, T &ExpectedOut, T Identity,
-                   BinaryOperation BOp, size_t N) {
-  ExpectedOut = Identity;
-  auto In = InBuf.template get_access<access::mode::write>();
-  for (int I = 0; I < N; ++I) {
-    if (std::is_same<BinaryOperation, std::multiplies<T>>::value)
-      In[I] = 1 + (((I % 37) == 0) ? 1 : 0);
-    else
-      In[I] = ((I + 1) % 5) + 1.1;
-    ExpectedOut = BOp(ExpectedOut, In[I]);
-  }
-};
-
-// This type is needed only to check that custom types are properly handled
-// in parallel_for() with reduction. For simplicity it needs a default
-// constructor, a constructor with one argument, operators ==, != and
-// printing to a stream.
-template <typename T>
-struct CustomVec {
-  CustomVec() : X(0), Y(0) {}
-  CustomVec(T X, T Y) : X(X), Y(Y) {}
-  CustomVec(T V) : X(V), Y(V) {}
-  bool operator==(const CustomVec &V) const {
-    return V.X == X && V.Y == Y;
-  }
-  bool operator!=(const CustomVec &V) const {
-    return !(*this == V);
-  }
-  T X;
-  T Y;
-};
-template <typename T>
-bool operator==(const CustomVec<T> &A, const CustomVec<T> &B) {
-  return A.X == B.X && A.Y == B.Y;
-}
-template <typename T>
-std::ostream &operator<<(std::ostream &OS, const CustomVec<T> &V) {
-  return OS << "(" << V.X << ", " << V.Y << ")";
-}
-
-template <class T>
-struct CustomVecPlus {
-  using CV = CustomVec<T>;
-  CV operator()(const CV &A, const CV &B) const {
-    return CV(A.X + B.X, A.Y + B.Y);
-  }
-};
-
-template <typename T>
-T getMinimumFPValue() {
-  return std::numeric_limits<T>::has_infinity
-             ? static_cast<T>(-std::numeric_limits<T>::infinity())
-             : std::numeric_limits<T>::lowest();
-}
-
-template <typename T>
-T getMaximumFPValue() {
-  return std::numeric_limits<T>::has_infinity
-             ? std::numeric_limits<T>::infinity()
-             : (std::numeric_limits<T>::max)();
-}
diff --git a/sycl/test/basic_tests/vector_cross.cpp b/sycl/test/basic_tests/vector_cross.cpp
index e6938f2618528..6e2da96b81055 100644
--- a/sycl/test/basic_tests/vector_cross.cpp
+++ b/sycl/test/basic_tests/vector_cross.cpp
@@ -1,5 +1,5 @@
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 
 #include <cassert>
 #include <cmath>
@@ -10,29 +10,20 @@ bool isEqualTo(double x, double y, double epsilon = 0.001) {
 }
 
 int main(int argc, const char **argv) {
-  sycl::cl_double4 r{0};
-  {
-    sycl::buffer<sycl::cl_double4, 1> BufR(&r, sycl::range<1>(1));
-    sycl::queue myQueue;
-    myQueue.submit([&](sycl::handler &cgh) {
-      auto AccR = BufR.get_access<sycl::access::mode::write>(cgh);
-      cgh.single_task<class crossD4>([=]() {
-        AccR[0] = sycl::cross(
-            sycl::cl_double4{
-                2.5,
-                3.0,
-                4.0,
-                0.0,
-            },
-            sycl::cl_double4{
-                5.2,
-                6.0,
-                7.0,
-                0.0,
-            });
+  sycl::cl_double4 r = sycl::cross(
+      sycl::cl_double4{
+          2.5,
+          3.0,
+          4.0,
+          0.0,
+      },
+      sycl::cl_double4{
+          5.2,
+          6.0,
+          7.0,
+          0.0,
       });
-    });
-  }
+
   sycl::cl_double r1 = r.x();
   sycl::cl_double r2 = r.y();
   sycl::cl_double r3 = r.z();
diff --git a/sycl/test/esimd/vadd.cpp b/sycl/test/esimd/vadd.cpp
deleted file mode 100644
index af8c82c4020d2..0000000000000
--- a/sycl/test/esimd/vadd.cpp
+++ /dev/null
@@ -1,121 +0,0 @@
-// RUN: %clangxx -fsycl -fno-legacy-pass-manager %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-
-// Check that the code compiles with -O0 and -g
-// Managers
-// RUN: %clangxx -I %sycl_include %s -o %t.out -fsycl -fno-legacy-pass-manager -O0
-// RUN: %clangxx -I %sycl_include %s -o %t.out -fsycl -fno-legacy-pass-manager -O0 -g
-
-// Check that the code compiles with device code instrumentation enabled
-// RUN: %clangxx -I %sycl_include %s -o %t.out -fsycl -fno-legacy-pass-manager \
-// RUN: -fsycl-instrument-device-code
-
-#include <iostream>
-#include <string>
-#include <sycl/ext/intel/esimd.hpp>
-#include <sycl/sycl.hpp>
-
-using namespace sycl;
-
-class ESIMDSelector : public device_selector {
-  // Require GPU device unless HOST is requested in SYCL_DEVICE_FILTER env
-  virtual int operator()(const device &device) const {
-    if (const char *dev_filter = getenv("SYCL_DEVICE_FILTER")) {
-      std::string filter_string(dev_filter);
-      if (filter_string.find("gpu") != std::string::npos)
-        return device.is_gpu() ? 1000 : -1;
-      if (filter_string.find("host") != std::string::npos)
-        return device.is_host() ? 1000 : -1;
-      std::cerr
-          << "Supported 'SYCL_DEVICE_FILTER' env var values are 'gpu' and "
-             "'host', '"
-          << filter_string << "' does not contain such substrings.\n";
-      return -1;
-    }
-    // If "SYCL_DEVICE_FILTER" not defined, only allow gpu device
-    return device.is_gpu() ? 1000 : -1;
-  }
-};
-
-auto exception_handler = [](exception_list l) {
-  for (auto ep : l) {
-    try {
-      std::rethrow_exception(ep);
-    } catch (sycl::exception &e0) {
-      std::cout << "sycl::exception: " << e0.what() << std::endl;
-    } catch (std::exception &e) {
-      std::cout << "std::exception: " << e.what() << std::endl;
-    } catch (...) {
-      std::cout << "generic exception\n";
-    }
-  }
-};
-
-int main(void) {
-  constexpr unsigned Size = 256;
-  constexpr unsigned VL = 32;
-  constexpr unsigned GroupSize = 2;
-
-  struct Deleter {
-    queue Q;
-    void operator()(int *Ptr) {
-      if (Ptr) {
-        sycl::free(Ptr, Q);
-      }
-    }
-  };
-
-  queue q(ESIMDSelector{}, exception_handler);
-
-  std::unique_ptr<int, Deleter> BufA(sycl::malloc_shared<int>(Size, q),
-                                     Deleter{q});
-  std::unique_ptr<int, Deleter> BufB(
-      sycl::aligned_alloc_shared<int>(16u, Size, q), Deleter{q});
-  std::unique_ptr<int, Deleter> BufC(
-      sycl::aligned_alloc_shared<int>(16u, Size, q), Deleter{q});
-
-  int *A = BufA.get();
-  int *B = BufB.get();
-  int *C = BufC.get();
-
-  for (unsigned i = 0; i < Size; ++i) {
-    A[i] = B[i] = i;
-  }
-
-  {
-    // We need that many task groups
-    sycl::range<1> GroupRange{Size / VL};
-
-    // We need that many tasks in each group
-    sycl::range<1> TaskRange{GroupSize};
-
-    sycl::nd_range<1> Range{GroupRange, TaskRange};
-
-    q.submit([&](sycl::handler &cgh) {
-      cgh.parallel_for<class Test>(
-          Range, [=](nd_item<1> ndi) SYCL_ESIMD_KERNEL {
-            using namespace sycl::ext::intel::esimd;
-
-            int i = ndi.get_global_id(0);
-            constexpr int ESIZE = sizeof(int);
-            simd<uint32_t, VL> offsets(0, ESIZE);
-
-            simd<int, VL> va = gather<int, VL>(A + i * VL, offsets);
-            simd<int, VL> vb = block_load<int, VL>(B + i * VL);
-            simd<int, VL> vc = va + vb;
-
-            block_store<int, VL>(C + i * VL, vc);
-          });
-    });
-  }
-  for (unsigned i = 0; i < Size; ++i) {
-    if (A[i] + B[i] != C[i]) {
-      std::cout << "failed at index " << i << ", " << C[i] << " != " << A[i]
-                << " + " << B[i] << "\n";
-      return 1;
-    }
-  }
-
-  std::cout << "Passed\n";
-  return 0;
-}
diff --git a/sycl/test/extensions/bfloat16_host.cpp b/sycl/test/extensions/bfloat16_host.cpp
index 3304587e61b9d..acd02d2829b40 100644
--- a/sycl/test/extensions/bfloat16_host.cpp
+++ b/sycl/test/extensions/bfloat16_host.cpp
@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 
 // RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 #include <sycl/ext/oneapi/experimental/bfloat16.hpp>
 #include <sycl/sycl.hpp>
 
diff --git a/sycl/test/extensions/usm/usm_alloc_utility.cpp b/sycl/test/extensions/usm/usm_alloc_utility.cpp
deleted file mode 100644
index 7c587df097b2b..0000000000000
--- a/sycl/test/extensions/usm/usm_alloc_utility.cpp
+++ /dev/null
@@ -1,112 +0,0 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t1.out
-// RUN: %RUN_ON_HOST %t1.out
-
-//==------ usm_alloc_utility.cpp - USM malloc and aligned_alloc test -------==//
-//
-// 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
-//
-//===----------------------------------------------------------------------===//
-
-#include <sycl/sycl.hpp>
-
-#include <cassert>
-
-using namespace sycl;
-
-constexpr int N = 8;
-
-static void check_and_free(int *array, const device &dev, const context &ctxt) {
-  // host device treats all allocations as host allocations
-  assert((get_pointer_type(array, ctxt) == usm::alloc::host) &&
-         "Allocation pointer should be host type");
-  assert((get_pointer_device(array, ctxt) == dev) &&
-         "Allocation pointer should be host type");
-  free(array, ctxt);
-}
-
-int main() {
-  queue q;
-  auto dev = q.get_device();
-  auto ctxt = q.get_context();
-  int *array;
-
-  if (dev.get_info<info::device::usm_host_allocations>()) {
-    array = (int *)malloc(N * sizeof(int), q, usm::alloc::host);
-    check_and_free(array, dev, ctxt);
-
-    array =
-        (int *)malloc(N * sizeof(int), q, usm::alloc::host, property_list{});
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)aligned_alloc(alignof(long long), N * sizeof(int), q,
-                                 usm::alloc::host);
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)aligned_alloc(alignof(long long), N * sizeof(int), q,
-                                 usm::alloc::host, property_list{});
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)malloc_host(N * sizeof(int), q);
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)malloc_host(
-        N * sizeof(int), q,
-        property_list{
-            ext::intel::experimental::property::usm::buffer_location{2}});
-    check_and_free(array, dev, ctxt);
-
-    array =
-        (int *)aligned_alloc_host(alignof(long long), N * sizeof(int), ctxt);
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)aligned_alloc_host(
-        alignof(long long), N * sizeof(int), ctxt,
-        property_list{
-            ext::intel::experimental::property::usm::buffer_location{2}});
-    check_and_free(array, dev, ctxt);
-  }
-
-  if (dev.get_info<info::device::usm_shared_allocations>()) {
-    array = (int *)malloc_shared(N * sizeof(int), q);
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)malloc_shared(
-        N * sizeof(int), q,
-        property_list{
-            ext::intel::experimental::property::usm::buffer_location{2}});
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)aligned_alloc_shared(alignof(long long), N * sizeof(int),
-                                        dev, ctxt);
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)aligned_alloc_shared(
-        alignof(long long), N * sizeof(int), dev, ctxt,
-        property_list{
-            ext::intel::experimental::property::usm::buffer_location{2}});
-    check_and_free(array, dev, ctxt);
-  }
-
-  if (dev.get_info<info::device::usm_device_allocations>()) {
-    array = (int *)malloc_device(N * sizeof(int), q);
-    check_and_free(array, dev, ctxt);
-
-    array = malloc_device<int>(
-        N, q,
-        property_list{
-            ext::intel::experimental::property::usm::buffer_location(2)});
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)aligned_alloc_device(alignof(long long), N * sizeof(int),
-                                        dev, ctxt);
-    check_and_free(array, dev, ctxt);
-
-    array = (int *)aligned_alloc_device(alignof(long long), N * sizeof(int),
-                                        dev, ctxt, property_list{});
-    check_and_free(array, dev, ctxt);
-  }
-
-  return 0;
-}
diff --git a/sycl/test/extensions/usm/usm_allocator.cpp b/sycl/test/extensions/usm/usm_allocator.cpp
deleted file mode 100644
index 7890673740809..0000000000000
--- a/sycl/test/extensions/usm/usm_allocator.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t1.out
-// RUN: %RUN_ON_HOST %t1.out
-
-//==--------- usm_allocator.cpp - USM allocator construction test ----------==//
-//
-// 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
-//
-//===----------------------------------------------------------------------===//
-
-#include <sycl/sycl.hpp>
-
-#include <cassert>
-
-using namespace sycl;
-
-int main() {
-  queue q;
-  auto dev = q.get_device();
-  auto ctxt = q.get_context();
-
-  {
-    // Test usm_allocator
-    if (dev.get_info<info::device::usm_shared_allocations>() &&
-        dev.get_info<info::device::usm_host_allocations>()) {
-      usm_allocator<int, usm::alloc::shared> alloc11(ctxt, dev);
-      usm_allocator<int, usm::alloc::shared> alloc12(ctxt, dev,
-                                                     property_list{});
-      usm_allocator<int, usm::alloc::host> alloc21(q);
-      usm_allocator<int, usm::alloc::host> alloc22(alloc21);
-      usm_allocator<int, usm::alloc::host> alloc23(q, property_list{});
-
-      // usm::alloc::device is not supported by usm_allocator
-
-      assert((alloc11 != alloc22) && "Allocators should NOT be equal.");
-      assert((alloc11 == alloc12) && "Allocators should be equal.");
-      assert((alloc21 == alloc22) && "Allocators should be equal.");
-      assert((alloc21 == alloc23) && "Allocators should be equal.");
-    }
-  }
-
-  return 0;
-}
diff --git a/sycl/test/lit.cfg.py b/sycl/test/lit.cfg.py
index e90c426f5f7b5..3fa183ff86b1c 100644
--- a/sycl/test/lit.cfg.py
+++ b/sycl/test/lit.cfg.py
@@ -90,10 +90,6 @@
 
 llvm_config.add_tool_substitutions(['llvm-spirv'], [config.sycl_tools_dir])
 
-config.substitutions.append( ('%RUN_ON_HOST', "env SYCL_DEVICE_FILTER=host ") )
-
-# Every SYCL implementation provides a host implementation.
-config.available_features.add('host')
 triple=lit_config.params.get('SYCL_TRIPLE', 'spir64-unknown-unknown')
 lit_config.note("Triple: {}".format(triple))
 config.substitutions.append( ('%sycl_triple',  triple ) )
diff --git a/sycl/test/regression/builtins_vector1.cpp b/sycl/test/regression/builtins_vector1.cpp
index 7c42884d14204..9b027d939ebec 100644
--- a/sycl/test/regression/builtins_vector1.cpp
+++ b/sycl/test/regression/builtins_vector1.cpp
@@ -1,20 +1,16 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %t.out
 
 #include <sycl/sycl.hpp>
 
 int main() {
-  sycl::queue().submit([&](sycl::handler &h) {
-    h.single_task<class task>([=]() {
-      sycl::vec<float, 1> A{1}, B{2}, C{3};
-      sycl::vec<float, 1> res = sycl::mad(A, B, C);
-      assert(res.x() - 5.0f < 1e-5);
+  sycl::vec<float, 1> A{1}, B{2}, C{3};
+  sycl::vec<float, 1> res = sycl::mad(A, B, C);
+  assert(res.x() - 5.0f < 1e-5);
 
-      res = sycl::clamp(A, B, C);
-      assert(res.x() - 2.0f < 1e-5);
+  res = sycl::clamp(A, B, C);
+  assert(res.x() - 2.0f < 1e-5);
 
-      float scalarRes = sycl::length(A);
-      assert(scalarRes - 1.0f < 1e-5);
-    });
-  });
+  float scalarRes = sycl::length(A);
+  assert(scalarRes - 1.0f < 1e-5);
 }
diff --git a/sycl/test/regression/check_vector_of_opencl_event.cpp b/sycl/test/regression/check_vector_of_opencl_event.cpp
deleted file mode 100644
index f6cbaf6ad17f6..0000000000000
--- a/sycl/test/regression/check_vector_of_opencl_event.cpp
+++ /dev/null
@@ -1,30 +0,0 @@
-// RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
-//
-//===----------------------------------------------------------------------===//
-// This test verifies that sycl::get_native<backend::opencl> and
-// sycl::make_event<backend::opencl> work according to the SYCL™ 2020
-// Specification (revision 4)
-//===----------------------------------------------------------------------===//
-
-#include <sycl/sycl.hpp>
-
-int main() {
-  sycl::queue Queue;
-  if (Queue.get_backend() == sycl::backend::opencl) {
-    sycl::event event = Queue.submit([&](sycl::handler &cgh) {
-      cgh.single_task<class event_kernel>([]() {});
-    });
-    // Check that get_native function returns a vector
-    std::vector<cl_event> ClEventVec = get_native<sycl::backend::opencl>(event);
-    // Check that make_event is working properly with vector<cl_event> as a
-    // param
-    sycl::event SyclEvent = sycl::make_event<sycl::backend::opencl>(
-        ClEventVec, Queue.get_context());
-    std::vector<cl_event> ClEventVecFromMake =
-        sycl::get_native<sycl::backend::opencl>(SyclEvent);
-    if (ClEventVec[0] != ClEventVecFromMake[0])
-      throw std::runtime_error("Cl events are not the same");
-  }
-  return 0;
-}
diff --git a/sycl/test/regression/fsycl-host-compiler-win.cpp b/sycl/test/regression/fsycl-host-compiler-win.cpp
deleted file mode 100644
index bf855f1899d93..0000000000000
--- a/sycl/test/regression/fsycl-host-compiler-win.cpp
+++ /dev/null
@@ -1,46 +0,0 @@
-// RUN: %clang_cl -fsycl -fsycl-host-compiler=cl -DDEFINE_CHECK -fsycl-host-compiler-options="-DDEFINE_CHECK /std:c++17 /Zc:__cplusplus" /Fe%t1.exe %s
-// RUN: %RUN_ON_HOST %t1.exe
-// REQUIRES: system-windows
-//
-//==------- fsycl-host-compiler-win.cpp - external host compiler test ------==//
-//
-// 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
-//
-//===----------------------------------------------------------------------===//
-//
-// Uses -fsycl-host-compiler=<compiler> on a simple test, requires 'cl'
-
-#include <sycl/sycl.hpp>
-
-#ifndef DEFINE_CHECK
-#error predefined macro not set
-#endif // DEFINE_CHECK
-
-using namespace sycl;
-
-int main() {
-  int data[] = {0, 0, 0};
-
-  {
-    buffer<int, 1> b(data, range<1>(3), {property::buffer::use_host_ptr()});
-    queue q;
-    q.submit([&](handler &cgh) {
-      auto B = b.get_access<access::mode::write>(cgh);
-      cgh.parallel_for<class test>(range<1>(3), [=](id<1> idx) {
-        B[idx] = 1;
-      });
-    });
-  }
-
-  bool isSuccess = true;
-
-  for (int i = 0; i < 3; i++)
-    if (data[i] != 1) isSuccess = false;
-
-  if (!isSuccess)
-    return -1;
-
-  return 0;
-}
diff --git a/sycl/test/regression/fsycl-host-compiler.cpp b/sycl/test/regression/fsycl-host-compiler.cpp
deleted file mode 100644
index 9e82e0ccff888..0000000000000
--- a/sycl/test/regression/fsycl-host-compiler.cpp
+++ /dev/null
@@ -1,56 +0,0 @@
-// RUN: %clangxx -fsycl -fsycl-host-compiler=g++ -DDEFINE_CHECK -fsycl-host-compiler-options="-DDEFINE_CHECK -std=c++17" -o %t1.exe %s
-// RUN: %RUN_ON_HOST %t1.exe
-// REQUIRES: system-linux
-//==------- fsycl-host-compiler.cpp - external host compiler test ----------==//
-//
-// 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
-//
-//===----------------------------------------------------------------------===//
-//
-// Uses -fsycl-host-compiler=<compiler> on a simple test, requires 'g++'
-
-#include <sycl/sycl.hpp>
-
-#ifndef DEFINE_CHECK
-#error predefined macro not set
-#endif // DEFINE_CHECK
-
-using namespace sycl;
-
-int main() {
-  int data[] = {0, 0, 0};
-
-  {
-    buffer<int, 1> b(data, range<1>(3), {property::buffer::use_host_ptr()});
-    queue q;
-    q.submit([&](handler &cgh) {
-      auto B = b.get_access<access::mode::write>(cgh);
-      cgh.parallel_for<class test>(range<1>(3), [=](id<1> idx) {
-        B[idx] = 1;
-      });
-    });
-  }
-
-  bool isSuccess = true;
-
-  for (int i = 0; i < 3; i++)
-    if (data[i] != 1) isSuccess = false;
-
-  {
-    buffer<int, 1> b(1);
-    queue q;
-    q.submit([&](handler &cgh) {
-       accessor a{b, cgh};
-       cgh.single_task<class test2>([=] { a[0] = 42; });
-     }).wait();
-    host_accessor a{b};
-    isSuccess &= (a[0] == 42);
-  }
-
-  if (!isSuccess)
-    return -1;
-
-  return 0;
-}
diff --git a/sycl/test/regression/half_union.cpp b/sycl/test/regression/half_union.cpp
index 14790a3b22944..1b67f8cb76c39 100644
--- a/sycl/test/regression/half_union.cpp
+++ b/sycl/test/regression/half_union.cpp
@@ -1,5 +1,5 @@
 // RUN: %clangxx -fsycl %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 
 #include <sycl/sycl.hpp>
 
diff --git a/sycl/test/regression/sycl-include-gnu17.cpp b/sycl/test/regression/sycl-include-gnu17.cpp
index f5abc7f82dd27..db9886610bf82 100644
--- a/sycl/test/regression/sycl-include-gnu17.cpp
+++ b/sycl/test/regression/sycl-include-gnu17.cpp
@@ -1,5 +1,5 @@
 // RUN: %clangxx -std=gnu++17 -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 
 // UNSUPPORTED: system-windows
 
diff --git a/sycl/test/scheduler/BasicSchedulerTests.cpp b/sycl/test/scheduler/BasicSchedulerTests.cpp
index 9c91161851146..dc9815429e621 100644
--- a/sycl/test/scheduler/BasicSchedulerTests.cpp
+++ b/sycl/test/scheduler/BasicSchedulerTests.cpp
@@ -1,5 +1,5 @@
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple -I %sycl_source_dir %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 //==------------------- BasicSchedulerTests.cpp ----------------------------==//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
diff --git a/sycl/test/scheduler/ReleaseResourcesTest.cpp b/sycl/test/scheduler/ReleaseResourcesTest.cpp
index ffed076bccc6f..08c0da5e0cd99 100644
--- a/sycl/test/scheduler/ReleaseResourcesTest.cpp
+++ b/sycl/test/scheduler/ReleaseResourcesTest.cpp
@@ -1,5 +1,5 @@
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple -I %sycl_source_dir %s -o %t.out
-// RUN: %RUN_ON_HOST %t.out
+// RUN: %t.out
 
 //==------------------- ReleaseResourcesTests.cpp --------------------------==//
 //