diff --git a/SYCL/AOT/gpu.cpp b/SYCL/AOT/gpu.cpp
index 8bf9a07886..587afecc99 100644
--- a/SYCL/AOT/gpu.cpp
+++ b/SYCL/AOT/gpu.cpp
@@ -10,5 +10,5 @@
 // UNSUPPORTED: cuda
 // CUDA is not compatible with SPIR.
 //
-// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend=spir64_gen "-device *" %S/Inputs/aot.cpp -o %t.out
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend=spir64_gen %gpu_aot_target_opts %S/Inputs/aot.cpp -o %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
diff --git a/SYCL/AOT/multiple-devices.cpp b/SYCL/AOT/multiple-devices.cpp
index 90667ceea1..fd36cd030c 100644
--- a/SYCL/AOT/multiple-devices.cpp
+++ b/SYCL/AOT/multiple-devices.cpp
@@ -12,7 +12,7 @@
 
 // 1-command compilation case
 // Targeting CPU, GPU, FPGA
-// RUN: %clangxx -fsycl -fsycl-targets=spir64_x86_64,spir64_gen,spir64_fpga -Xsycl-target-backend=spir64_gen "-device *" %S/Inputs/aot.cpp -o %t_all.out
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_x86_64,spir64_gen,spir64_fpga -Xsycl-target-backend=spir64_gen %gpu_aot_target_opts %S/Inputs/aot.cpp -o %t_all.out
 // RUN: %HOST_RUN_PLACEHOLDER %t_all.out
 // RUN: %CPU_RUN_PLACEHOLDER %t_all.out
 // RUN: %GPU_RUN_PLACEHOLDER %t_all.out
diff --git a/SYCL/DeviceCodeSplit/aot-gpu.cpp b/SYCL/DeviceCodeSplit/aot-gpu.cpp
index f11a4ce7c9..0364ed88de 100644
--- a/SYCL/DeviceCodeSplit/aot-gpu.cpp
+++ b/SYCL/DeviceCodeSplit/aot-gpu.cpp
@@ -5,7 +5,7 @@
 // RUN: %clangxx -fsycl -fsycl-device-code-split=per_source \
 // RUN:   -fsycl-targets=spir64_gen \
 // RUN:   -Xsycl-target-backend=spir64_gen \
-// RUN:   "-device *" -I %S/Inputs -o %t.out \
+// RUN:   %gpu_aot_target_opts -I %S/Inputs -o %t.out \
 // RUN:   %S/split-per-source-main.cpp \
 // RUN:   %S/Inputs/split-per-source-second-file.cpp \
 // RUN:   -fsycl-dead-args-optimization
diff --git a/SYCL/ESIMD/aot_mixed.cpp b/SYCL/ESIMD/aot_mixed.cpp
index 0a4079209e..3fe0013a0a 100644
--- a/SYCL/ESIMD/aot_mixed.cpp
+++ b/SYCL/ESIMD/aot_mixed.cpp
@@ -13,9 +13,9 @@
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 // XFAIL: *
 // TODO: remove XFAIL once ocloc support for automatic scalar/vector SPIRV
-// module dispatching is available in public drivers. Also change 'gen9' (safe
-// variant to reliably get unexpected PASS when ocloc is fixed) to '*' (which
-// stresses ocloc).
+// module dispatching is available in public drivers. Also change '-device gen9'
+// (safe variant to reliably get unexpected PASS when ocloc is fixed) to
+// %gpu_aot_target_opts aka '-device *' (which stresses ocloc).
 
 // This test checks ESIMD ahead-of-time compilation scenarios:
 // 1) When the application contains both SYCL and ESIMD kernel, thus requiring
diff --git a/SYCL/README.md b/SYCL/README.md
index 8bab44e134..22ed8d159a 100644
--- a/SYCL/README.md
+++ b/SYCL/README.md
@@ -62,20 +62,20 @@ ninja check-sycl-all
 
 These parameters can be used to configure tests:
 
-***CMAKE_CXX_COMPILER*** path DPCPP compiler
+***CMAKE_CXX_COMPILER*** - path to DPCPP compiler
 
-***TEST_SUITE_LLVM_SIZE*** path to llvm-size tool, required for code size
+***TEST_SUITE_LLVM_SIZE*** - path to llvm-size tool, required for code size
 collection
 
-***TEST_SUITE_COLLECT_COMPILE_TIME=OFF*** can be used to turn off compile time
-collection
+***TEST_SUITE_COLLECT_COMPILE_TIME=OFF*** - can be used to turn off compile
+time collection
 
-***TEST_SUITE_COLLECT_CODE_SIZE=OFF*** can be used to turn off code size
+***TEST_SUITE_COLLECT_CODE_SIZE=OFF*** - can be used to turn off code size
 collection
 
-***TEST_SUITE_LIT*** path to llvm-lit tool
+***TEST_SUITE_LIT*** - path to llvm-lit tool
 
-***CHECK_SYCL_ALL*** defines selection of multiple SYCL backends with set of
+***CHECK_SYCL_ALL*** - defines selection of multiple SYCL backends with set of
 target devices per each to be tested iteratively. Value is semicolon-separated
 list of configurations. Each configuration includes backend separated
 from comma-separated list of target devices with colon. Example:
@@ -84,36 +84,39 @@ from comma-separated list of target devices with colon. Example:
 -DCHECK_SYCL_ALL="opencl:cpu,host;level_zero:gpu,host;cuda:gpu;hip:gpu"
 ```
 
-***SYCL_BE*** SYCL backend to be used for testing. Supported values are:
+***SYCL_BE*** - SYCL backend to be used for testing. Supported values are:
  - **opencl** - for OpenCL backend;
  - **cuda** - for CUDA backend;
  - **hip** - for HIP backend;
  - **level_zero** - Level Zero backend.
 
-***SYCL_TARGET_DEVICES*** comma separated list of target devices for testing.
+***SYCL_TARGET_DEVICES*** - comma separated list of target devices for testing.
 Default value is cpu,gpu,acc,host. Supported values are:
  - **cpu**  - CPU device available in OpenCL backend only;
  - **gpu**  - GPU device available in OpenCL, Level Zero, CUDA, and HIP backends;
  - **acc**  - FPGA emulator device available in OpenCL backend only;
  - **host** - SYCL Host device available with all backends.
 
-***OpenCL_LIBRARY*** path to OpenCL ICD loader library. OpenCL interoperability
-tests require OpenCL ICD loader to be linked with. For such tests OpenCL ICD
-loader library should be installed in the system or available at the full path
-specified by this variable.
+***OpenCL_LIBRARY*** - path to OpenCL ICD loader library. OpenCL
+interoperability tests require OpenCL ICD loader to be linked with. For such
+tests OpenCL ICD loader library should be installed in the system or available
+at the full path specified by this variable.
 
-***LEVEL_ZERO_INCLUDE*** path to Level Zero headers.
+***LEVEL_ZERO_INCLUDE*** - path to Level Zero headers.
 
-***LEVEL_ZERO_LIBS_DIR*** path to Level Zero libraries.
+***LEVEL_ZERO_LIBS_DIR*** - path to Level Zero libraries.
 
-***HIP_PLATFORM*** platform selection for HIP targeted devices.
+***HIP_PLATFORM*** - platform selection for HIP targeted devices.
 Defaults to AMD if no value is given. Supported values are:
  - **AMD**    - for HIP to target AMD GPUs
  - **NVIDIA** - for HIP to target NVIDIA GPUs
  
- ***AMD_ARCH*** Flag must be set for when using HIP AMD triple.
+ ***AMD_ARCH*** - flag must be set for when using HIP AMD triple.
  For example it may be set to "gfx906".
 
+***GPU_AOT_TARGET_OPTS*** - defines additional options which are passed to AOT
+compilation command line for GPU device. If not specified "-device *" value
+is used.
 
 # Special test categories
 
@@ -154,6 +157,9 @@ configure specific single test execution in the command line:
    host);
  * **sycl_be** - SYCL backend to be used (opencl, level_zero, cuda, hip);
  * **dump_ir** - if IR dumping is supported for compiler (True, False);
+ * **gpu_aot_target_opts** - defines additional options which are passed to AOT
+   compilation command line for GPU device. It can be also set by CMake variable
+   GPU_AOT_TARGET_OPTS. If not specified "-device *" value is used.
  * **gpu-intel-dg1** - tells LIT infra that Intel GPU DG1 is present in the
    system. It is developer / CI infra responsibility to make sure that the
    device is available in the system. Tests requiring DG1 to run must use proper
diff --git a/SYCL/SpecConstants/2020/non_native/gpu.cpp b/SYCL/SpecConstants/2020/non_native/gpu.cpp
index e45431ca55..27a2911b88 100644
--- a/SYCL/SpecConstants/2020/non_native/gpu.cpp
+++ b/SYCL/SpecConstants/2020/non_native/gpu.cpp
@@ -5,7 +5,7 @@
 // UNSUPPORTED: hip
 // HIP is not compatible with SPIR.
 
-// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend=spir64_gen "-device *" %S/Inputs/common.cpp -o %t.out -fsycl-dead-args-optimization
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend=spir64_gen %gpu_aot_target_opts %S/Inputs/common.cpp -o %t.out -fsycl-dead-args-optimization
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 
 // This test checks correctness of SYCL2020 non-native specialization constants
diff --git a/SYCL/lit.cfg.py b/SYCL/lit.cfg.py
index 4e28715879..1a51880d0c 100644
--- a/SYCL/lit.cfg.py
+++ b/SYCL/lit.cfg.py
@@ -149,6 +149,11 @@
     config.substitutions.append( ('%debug_option',  '-g' ) )
     config.substitutions.append( ('%cxx_std_option',  '-std=' ) )
 
+if not config.gpu_aot_target_opts:
+    config.gpu_aot_target_opts = '"-device *"'
+
+config.substitutions.append( ('%gpu_aot_target_opts',  config.gpu_aot_target_opts ) )
+
 if not config.sycl_be:
      lit_config.error("SYCL backend is not specified")
 
diff --git a/SYCL/lit.site.cfg.py.in b/SYCL/lit.site.cfg.py.in
index b0ee4db3ed..bd85b79b0a 100644
--- a/SYCL/lit.site.cfg.py.in
+++ b/SYCL/lit.site.cfg.py.in
@@ -30,6 +30,7 @@ config.cxx_flags = "@CMAKE_CXX_FLAGS@"
 config.c_flags = "@CMAKE_C_FLAGS@"
 config.external_tests = "@SYCL_EXTERNAL_TESTS@"
 config.extra_include = "@CMAKE_CURRENT_SOURCE_DIR@/include"
+config.gpu_aot_target_opts = lit_config.params.get("gpu_aot_target_opts", "@GPU_AOT_TARGET_OPTS@")
 
 import lit.llvm
 lit.llvm.initialize(lit_config, config)