Tools/ApplicationDebugger update (#2347)

* Tools/ApplicationDebugger: add -g flag explicitly for debug builds.

* Tools/ApplicationDebugger: remove "$" before bash commands in README.

* Tools/ApplicationDebugger: use ONEAPI_DEVICE_SELECTOR to choose the device.

* Tools/ApplicationDebugger: update guided_matrix_mult_SLMSize.

Author: nsaiapova

Tools/ApplicationDebugger/array-transform/README.md

@@ -82,7 +82,7 @@ To learn more about the extensions, see the
 > Linux*:
 > - For system wide installations: `. /opt/intel/oneapi/setvars.sh`
 > - For private installations: `. ~/intel/oneapi/setvars.sh`
-> - For non-POSIX shells, like csh, use the following command: `$ bash -c 'source <install-dir>/setvars.sh ; exec csh'`
+> - For non-POSIX shells, like csh, use the following command: `bash -c 'source <install-dir>/setvars.sh ; exec csh'`
 >
 > Windows*:
 > - `C:\"Program Files (x86)"\Intel\oneAPI\setvars.bat`
@@ -112,7 +112,7 @@ batch or interactive mode.  For the array transform sample, a node
 with GPU and an interactive shell is recommended.
 
 ```
-$ qsub -I -l nodes=1:gpu:ppn=2
+qsub -I -l nodes=1:gpu:ppn=2
 ```
 
 For more information, see the Intel® oneAPI
@@ -127,12 +127,12 @@ offloaded to the GPU can be debugged conveniently.  Auto-attach is
 by default enabled.  To turn this feature off, if desired (e.g., if
 interested in debugging CPU or FPGA-emu only), do:
 ```
-$ export INTELGT_AUTO_ATTACH_DISABLE=1
+export INTELGT_AUTO_ATTACH_DISABLE=1
 ```
 
 To turn the feature back on:
 ```
-$ unset INTELGT_AUTO_ATTACH_DISABLE
+unset INTELGT_AUTO_ATTACH_DISABLE
 ```
 
 ### On a Linux* System
@@ -141,33 +141,33 @@ Perform the following steps:
 
 1.  Build the program using the following `cmake` commands.
     ```
-    $ cd array-transform
-    $ mkdir build
-    $ cd build
-    $ cmake ..
-    $ make
+    cd array-transform
+    mkdir build
+    cd build
+    cmake ..
+    make
     ```
     > Note: The cmake configuration enforces the `Debug` build type.
 
 2.  Run the program:
     ```
-    $ ./array-transform
+    ./array-transform
     ```
     > Note: to specify a device type to offload the kernel, use
     > the `ONEAPI_DEVICE_SELECTOR` environment variable.
     > E.g.  to restrict the offload only to CPU devices use:
     ```
-    $ ONEAPI_DEVICE_SELECTOR=*:cpu ./array-transform
+    ONEAPI_DEVICE_SELECTOR=*:cpu ./array-transform
     ```
 
 3.  Start a debugging session:
     ```
-    $ gdb-oneapi array-transform
+    gdb-oneapi array-transform
     ```
 
 4.  Clean the program using:
     ```
-    $ make clean
+    make clean
     ```
 
 By default, CMake configures the build for Just-in-Time (JIT)
@@ -182,7 +182,7 @@ compiler.
 For example, to do AoT compilation for a specific GPU device ID:
 
 ```
-$ cmake .. -DSYCL_COMPILE_TARGET=<device id>
+cmake .. -DSYCL_COMPILE_TARGET=<device id>
 ```
 where the `<device id>` specifies the target device, e.g., "xe".
 
@@ -233,7 +233,7 @@ please see [Get Started with Intel® Distribution for GDB* on Windows* OS Host](
 ### Example Outputs
 
 ```
-$ ONEAPI_DEVICE_SELECTOR=*:cpu gdb-oneapi -q ./array-transform
+ONEAPI_DEVICE_SELECTOR=*:cpu gdb-oneapi -q ./array-transform
 Reading symbols from ./array-transform...
 (gdb) break 54
 Breakpoint 1 at 0x4057b7: file array-transform.cpp, line 54.
@@ -248,7 +248,7 @@ Thread 4 "array-transform" hit Breakpoint 1, main::{lambda(auto:1&)#1}::operator
 ```
 
 ```
-$ ONEAPI_DEVICE_SELECTOR=*:fpga gdb-oneapi -q ./array-transform
+ONEAPI_DEVICE_SELECTOR=*:fpga gdb-oneapi -q ./array-transform
 Reading symbols from ./array-transform...
 (gdb) break 54
 Breakpoint 1 at 0x4057b7: file array-transform.cpp, line 54.
@@ -263,7 +263,7 @@ Thread 6 "array-transform" hit Breakpoint 1, main::{lambda(auto:1&)#1}::operator
 ```
 
 ```
-$ ONEAPI_DEVICE_SELECTOR=level_zero:gpu gdb-oneapi -q ./array-transform
+ONEAPI_DEVICE_SELECTOR=level_zero:gpu gdb-oneapi -q ./array-transform
 Reading symbols from ./array-transform...
 (gdb) break 54
 Breakpoint 1 at 0x4057b7: file array-transform.cpp, line 54.

Tools/ApplicationDebugger/guided_matrix_mult_BadBuffers/CMakeLists.txt

@@ -12,7 +12,7 @@ if (NOT CMAKE_BUILD_TYPE MATCHES Debug)
   message (STATUS "Build Type is changed to 'Debug'.")
 endif ()
 
-set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -O0")
+set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -g -O0")
 message (STATUS "Optimization level is set to -O0.")
 
 # Check that debugger executables exist in PATH

Tools/ApplicationDebugger/guided_matrix_mult_BadBuffers/README.md

@@ -178,7 +178,7 @@ In `a1_matrix_mul_zero_buff`, a zero-element buffer is passed to a SYCL submit `
 
 2. Start the debugger to watch the application failure and find out where it failed.
    ```
-   $ gdb-oneapi ./a1_matrix_mul_zero_buff
+   gdb-oneapi ./a1_matrix_mul_zero_buff
    ```
 
 3. You should get the prompt `(gdb)`.
@@ -297,7 +297,7 @@ You will run the application and review what happens when the program is run usi
 
 1. Run the program on the GPU using Level Zero.
    ```
-   SYCL_DEVICE_FILTER=level_zero:gpu ./b1_matrix_mul_null_usm
+   ONEAPI_DEVICE_SELECTOR=level_zero:gpu ./b1_matrix_mul_null_usm
    ```
    This run produces troublesome output.
    ```
@@ -310,14 +310,14 @@ You will run the application and review what happens when the program is run usi
 
 2. Check the output if we run on the GPU again but using OpenCL.
    ```
-   SYCL_DEVICE_FILTER=opencl:gpu ./b1_matrix_mul_null_usm
+   ONEAPI_DEVICE_SELECTOR=opencl:gpu ./b1_matrix_mul_null_usm
    ```
 
    The results should be the same as the Level Zero output.
 
 3. Check the output we get by bypassing the GPU entirely and using the OpenCL driver for CPU.
    ```
-   SYCL_DEVICE_FILTER=opencl:cpu ./b1_matrix_mul_null_usm
+   ONEAPI_DEVICE_SELECTOR=opencl:cpu ./b1_matrix_mul_null_usm
    ```
    ```
    Problem size: c(150,600) = a(150,300) * b(300,600)
@@ -333,7 +333,7 @@ Let's see what caused the problem by running in the debugger using the OpenCL dr
 
 1. Start the debugger using OpenCL™ on the CPU.
    ```
-   SYCL_DEVICE_FILTER=opencl:cpu gdb-oneapi ./b1_matrix_mul_null_usm
+   ONEAPI_DEVICE_SELECTOR=opencl:cpu gdb-oneapi ./b1_matrix_mul_null_usm
    ```
 2. You should get the prompt `(gdb)`.
 

Tools/ApplicationDebugger/guided_matrix_mult_Exceptions/CMakeLists.txt

@@ -12,7 +12,7 @@ if (NOT CMAKE_BUILD_TYPE MATCHES Debug)
   message (STATUS "Build Type is changed to 'Debug'.")
 endif ()
 
-set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -O0")
+set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -g -O0")
 message (STATUS "Optimization level is set to -O0.")
 
 # Check that debugger executables exist in PATH

Tools/ApplicationDebugger/guided_matrix_mult_InvalidContexts/CMakeLists.txt

@@ -12,7 +12,7 @@ if (NOT CMAKE_BUILD_TYPE MATCHES Debug)
   message (STATUS "Build Type is changed to 'Debug'.")
 endif ()
 
-set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -O0")
+set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -g -O0")
 message (STATUS "Optimization level is set to -O0.")
 
 # Check that debugger executables exist in PATH

Tools/ApplicationDebugger/guided_matrix_mult_InvalidContexts/README.md

@@ -159,7 +159,7 @@ You must download the [Tracing and Profiling Tool](https://github.com/intel/pti-
 
 2. Check the results on a **GPU** with OpenCL.
    ```
-   SYCL_DEVICE_FILTER=opencl:gpu ./1_matrix_mul_invalid_contexts
+   ONEAPI_DEVICE_SELECTOR=opencl:gpu ./1_matrix_mul_invalid_contexts
    ```
    The output might look similar to the following:
    ```
@@ -176,7 +176,7 @@ You must download the [Tracing and Profiling Tool](https://github.com/intel/pti-
    ```
 3. Check the results on the **CPU** using OpenCL. You should see similar problems.
    ```
-   SYCL_DEVICE_FILTER=opencl:cpu ./1_matrix_mul_invalid_contexts
+   ONEAPI_DEVICE_SELECTOR=opencl:cpu ./1_matrix_mul_invalid_contexts
    ```
    The output might look like the following:
    ```
@@ -199,7 +199,7 @@ In this section, you will use the Intel® Distribution for GDB* to determine wha
 
 1. Start the debugger using OpenCL™ on the **GPU**.
    ```
-   SYCL_DEVICE_FILTER=opencl:gpu  gdb-oneapi ./1_matrix_mul_invalid_contexts
+   ONEAPI_DEVICE_SELECTOR=opencl:gpu  gdb-oneapi ./1_matrix_mul_invalid_contexts
    ```
 2. You should get the prompt `(gdb)`.
 
@@ -319,7 +319,7 @@ One of the things that the Tracing and Profiling utility can help us identify is
    >**Note**: You must modify the command shown below to include the path to where you installed the `onetrace` utility.
 
    ```
-   SYCL_DEVICE_FILTER=opencl:gpu [path]/onetrace -c ./1_matrix_mul_invalid_contexts
+   ONEAPI_DEVICE_SELECTOR=opencl:gpu [path]/onetrace -c ./1_matrix_mul_invalid_contexts
    ```
 
    The `onetrace` utility outputs extensive results. A few key excerpts with areas of interest are shown below.
@@ -343,7 +343,7 @@ One of the things that the Tracing and Profiling utility can help us identify is
 
 2. Look at the output from Level Zero, and see if we could have detected the issue.
    ```
-   SYCL_DEVICE_FILTER=level_zero:gpu [path]onetrace -c ./1_matrix_mul_invalid_contexts
+   ONEAPI_DEVICE_SELECTOR=level_zero:gpu [path]onetrace -c ./1_matrix_mul_invalid_contexts
    ```
    Your output might be similar to the following:
    ```

Tools/ApplicationDebugger/guided_matrix_mult_RaceCondition/CMakeLists.txt

@@ -12,7 +12,7 @@ if (NOT CMAKE_BUILD_TYPE MATCHES Debug)
   message (STATUS "Build Type is changed to 'Debug'.")
 endif ()
 
-set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -O0")
+set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -g -O0")
 message (STATUS "Optimization level is set to -O0.")
 
 # Check that debugger executables exist in PATH

Tools/ApplicationDebugger/guided_matrix_mult_RaceCondition/README.md

@@ -290,14 +290,14 @@ But what if the developer didn't delete `c_back`, and let program termination cl
 Is the behavior different if you run it on OpenCL or Level 0?  The default is to use the Level Zero run time, but we can explicitly force the use of either Level Zero or OpenCL, which can be helpful when troubleshooting.
 
 ```
-$ SYCL_DEVICE_FILTER=level_zero:gpu ./1_matrix_mul_race_condition
-$ SYCL_DEVICE_FILTER=opencl:gpu ./1_matrix_mul_race_condition
+ONEAPI_DEVICE_SELECTOR=level_zero:gpu ./1_matrix_mul_race_condition
+ONEAPI_DEVICE_SELECTOR=opencl:gpu ./1_matrix_mul_race_condition
 ```
 Unfortunately not; pretty much the same thing happens.
 
 Similarly, we specify targeting the CPU, which sometimes can avoid problems in your code that are specific to offloading to the GPU.
 ```
-$ SYCL_DEVICE_FILTER=CPU ./1_matrix_mul_race_condition
+ONEAPI_DEVICE_SELECTOR=*:cpu ./1_matrix_mul_race_condition
 ```
 This also has problems, but if you run this in the debugger you will see lots of threads all running in the third `q.submit` kernel, but no thread running `main`.   This is because these threads have been abandoned when `main` deleted `c_back` and exited!
 

Tools/ApplicationDebugger/guided_matrix_mult_SLMSize/CMakeLists.txt

@@ -12,7 +12,7 @@ if (NOT CMAKE_BUILD_TYPE MATCHES Debug)
   message (STATUS "Build Type is changed to 'Debug'.")
 endif ()
 
-set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -O0")
+set (CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -fsycl -g -O0")
 message (STATUS "Optimization level is set to -O0.")
 
 # Check that debugger executables exist in PATH