You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
GPUPerfAPI no longer uses git submodules to reference dependent repositories. Instead, you need to follow these instructions in
11
-
order to clone/update any dependent repositories.
11
+
order to clone and update any dependent repositories.
12
12
13
13
#### Prerequisites
14
14
* Python 3.x, which can be installed from https://www.python.org/.
15
-
* CMake 3.19 or newer
16
-
* For Windows, this can be downloaded from https://cmake.org/download/
17
-
* For Linux, this can be installed using: sudo apt-get install cmake
15
+
* CMake 3.19 and newer.
16
+
* For Windows, this can be downloaded from https://cmake.org/download/.
17
+
* For Linux, this can be installed using:
18
+
*`sudo apt-get install cmake`
18
19
* To build the documentation:
19
20
* Install Sphinx:
20
-
*`pip install Sphinx`
21
+
*`pip install sphinx`
21
22
* Install Sphinx Read The Docs theme:
22
23
*`pip install sphinx-rtd-theme` (read the docs theme is not installed by default)
23
24
* Install spelling checker (optional)
24
25
*`pip install pyenchant`
25
26
*`pip install sphinxcontrib-spelling`
27
+
***Note:** Be sure to add Python scripts to PATH in order to build documentation from prebuild scripts.
26
28
27
29
#### Instructions
28
-
* Simply execute the [pre_build.py](build/pre_build.py)python script located in the GPA directory:
30
+
* Simply execute the [pre_build.py](build/pre_build.py)Python script located in the GPA directory:
29
31
*`python build/pre_build.py`
30
32
* This script will clone any dependent repositories that are not present on the system. If any of the dependent repositories are already
31
33
present on the system, this script will instead do a "git pull" on those repositories to ensure that they are up to date. Please re-run
32
34
this script everytime you pull new changes from GPA repository.
33
-
*NOTE: For GPA 3.3 or newer, if you are updating an existing clone of the GPA repo from a GPA release prior than 3.3, you will first need to delete the Common/Lib/Ext/GoogleTest directory. Starting with GPA 3.3, GPA is now using a fork of the official GoogleTest repo. Failure to remove this directory will lead to git errors when running [pre_build.py](build/pre_build.py) or [fetch_dependencies.py](scripts/fetch_dependencies.py).
34
-
*NOTE: For GPA 3.11 and newer, the Common/ directory has been removed, and the external/Lib/Ext/GoogleTest will first need to be deleted instead.
35
-
* This script will also download and execute the Vulkan� SDK installer.
36
-
* On Windows, running the installer may require elevation. If you've previously installed the required Vulkan version, fetch_dependencies.py will simply copy the files from the default installation location into the correct place into the GPUPerfAPI directory tree.
35
+
***Note:** For GPA 3.3 and newer, if you are updating an existing clone of the GPA repo from a GPA release prior to 3.3, you will first need to delete the Common/Lib/Ext/GoogleTest directory. Starting with GPA 3.3, GPA is now using a fork of the official GoogleTest repo. Failure to remove this directory will lead to Git errors when running [pre_build.py](build/pre_build.py) or [fetch_dependencies.py](scripts/fetch_dependencies.py).
36
+
***Note:** For GPA 3.11 and newer, the Common/ directory has been removed, and the external/Lib/Ext/GoogleTest will first need to be deleted instead.
37
+
* This script will also download and execute the Vulkan SDK installer.
38
+
* On Windows, running the installer may require elevation. If you've previously installed the required Vulkan version, fetch_dependencies.py will simply copy the files from the default installation location into the correct place into the GPUPerfAPI directory tree.
37
39
* fetch_dependencies.py is set up to install the version of the Vulkan SDK which was used during development. If you want to use a newer version of the SDK, the following file will need to be updated:
* By default the build will expect the Vulkan SDK to be found in a directory pointed to by the `VULKAN_SDK` environment variable. This environment variable is automatically set by the Windows SDK installer, but you may need to set it manually after running the Linux SDK installer. The Linux SDK includes a script called `setup-env.sh` to aid in setting this environment variable:
40
-
*`source ~/VulkanSDK/1.0.68.0/setup-env.sh` (adjust path as necessary)
41
-
* This script also executes cmake to generate all required files to build GPA.
42
-
* If you want to generate all cmake build files without trying to clone/pull dependent repos, you can add "--nofetch" to the [pre_build.py](build/pre_build.py) command line.
42
+
*`source ~/VulkanSDK/1.0.68.0/setup-env.sh` (adjust PATH as necessary)
43
+
* This script also executes CMake to generate all required files to build GPA.
44
+
* If you want to generate all CMake build files without trying to clone/pull dependent repos, you can add "--nofetch" to the [pre_build.py](build/pre_build.py) command line.
43
45
* Additional switches that can be used with the [pre_build.py](build/pre_build.py) script:
44
-
*`--vs=[2015,2017,2019,2022]`: Specify the Visual Studio version for which to generate projects. Default is 2019.
46
+
*`--vs=[2015,2017,2019,2022]`: Specify the Visual Studio version for which to generate projects. Default is 2022.
45
47
*`--config=[debug,release]`: Specify the config for which to generate makefiles. Default is both. A specific config can only be specified on Linux. On Windows, both configs are always supported by the generated VS solution and project files.
46
48
*`--platform=[x86,x64]`: Specify the platform for which to generate build files. Default is both.
47
49
*`--clean`: Deletes CMakeBuild directory and regenerates all build files from scratch
@@ -57,60 +59,67 @@ this script everytime you pull new changes from GPA repository.
57
59
## Windows Build Information
58
60
59
61
##### Prerequisites
60
-
* Microsoft Visual Studio 2019
61
-
* Windows 10 SDK Version 10.0.10586.0 from https://developer.microsoft.com/en-US/windows/downloads/windows-10-sdk
62
-
* You can override the version of the Windows 10 SDK used by modifying external/Lib/Ext/Windows-Kits/Global-WindowsSDK.cmake
63
-
* Microsoft .NET 4.6.2 SDK from https://dotnet.microsoft.com/en-us/download/dotnet-framework/thank-you/net462-developer-pack-offline-installer
62
+
* Microsoft Visual Studio 2022
63
+
* Within the Visual Studio Installer, the following workloads:
64
+
* Desktop development with C++
65
+
* Within the Visual Studio Installer, the following individual components:
66
+
* Windows 11 SDK (10.0.22621.0)
67
+
* MSVC v143 - VS 2022 C++ x74/x86 build tools (latest)
68
+
* C++ CMake tools for Windows
69
+
* C++ Clang Compiler for Windows (15.0.1)
70
+
* .NET Framework 4.6.2-4.8.1 SDKs and targeting packs
71
+
* C# and Visual Basic
72
+
* C# and Visual Basic Roslyn compilers
64
73
65
74
##### Build Instructions
66
-
* Load cmake_bld\x64\GPUPerfAPI.sln into Visual Studio to build the 64-bit version of GPA
67
-
* Load cmake_bld\x86\GPUPerfAPI.sln into Visual Studio to build the 32-bit version of GPA
68
-
* After a successful build, the GPUPerfAPI binaries can be found in `gpu_performance_api\output\$(Configuration)` (for example gpu_performance_api\output\release)
75
+
* Load cmake_bld\x64\GPUPerfAPI.sln into Visual Studio to build the 64-bit version of GPA.
76
+
* Load cmake_bld\x86\GPUPerfAPI.sln into Visual Studio to build the 32-bit version of GPA.
77
+
* After a successful build, the GPUPerfAPI binaries can be found in `gpu_performance_api\output\$(Configuration)` (e.g. gpu_performance_api\output\release).
69
78
70
79
#### Additional Information
71
-
* The Windows projects each include a .rc file that embeds the VERSIONINFO resource into the final binary. Internally within AMD, a Jenkins build system will dynamically update
80
+
* The Windows projects each include a .rc file that embeds the VERSIONINFO resource into the final binary. Internally within AMD, a Jenkins build system will dynamically update.
72
81
the build number. The version and build numbers can be manually updated by modifying the [gpa_version.h](source/gpu_perf_api_common/gpa_version.h) file.
73
82
74
83
## Linux Build Information
75
84
76
85
##### Prerequisites
77
-
* Install the Mesa common development package: sudo apt-get install mesa-common-dev
86
+
* Install the Mesa common development package:
87
+
*`sudo apt-get install mesa-common-dev`
78
88
79
89
##### Build Instructions
80
-
* Execute "make" in the cmake_bld/x64/debug to build the 64-bit debug version of GPA
81
-
* Execute "make" in the cmake_bld/x64/release to build the 64-bit release version of GPA
82
-
* After a successful build, the GPUPerfAPI binaries can be found in `gpu_performance_api/output/$(Configuration)` (for example gpu_performance_api/output/release)
90
+
* Execute "make" in the cmake_bld/x64/debug to build the 64-bit debug version of GPA.
91
+
* Execute "make" in the cmake_bld/x64/release to build the 64-bit release version of GPA.
92
+
* After a successful build, the GPUPerfAPI binaries can be found in `gpu_performance_api/output/$(Configuration)` (e.g. gpu_performance_api/output/release).
83
93
84
94
## PublicCounterCompiler Tool
85
-
86
-
The PublicCounterCompiler Tool is a utility, written in C#, that will generate C++ code to define the public (or derived) counters.
95
+
The PublicCounterCompiler Tool is a C# utility that will generate C++ code to define the public (or derived) counters.
87
96
It takes as input text files contained in the [public_counter_compiler_input_files](source/public_counter_compiler_input_files) directory and
88
97
outputs files in the [gpu_perf_api_counter_generator](source/auto_generated/gpu_perf_api_counter_generator), [gpu_perf_api_unit_tests](source/auto_generated/gpu_perf_api_unit_tests)
89
98
and [docs](docs) directories.
90
99
91
100
There are three ways to execute the tool:
92
-
* With no parameters - it opens the user interface with no fields prepopulated
93
-
* With two parameters - it opens the user interface with the two main fields prepopulated. When you press the "Compile Public Counters" button it will load the correct input files and generate the output files in the correct location.
94
-
* Param 1: API -- the API to compile counters for (ex: GL, CL, DX11, DX12, VK, etc).
95
-
* Param 2: HW generation: the generation to compile counters for (ex: Gfx6, Gfx7, Gfx8, etc.)
96
-
* With six or seven parameters - the user interface does not open. It simply generates the c++ files using the specified input and output file locations
97
-
* Param 1: Counter names file - text file containing hardware counter names and type (CounterNames[API][GEN].txt)
98
-
* Param 2: Public counter definition file - text file defining how the public counters are calculated (PublicCounterDefinitions\*.txt)
99
-
* Param 3: Output Dir - the directory to generate the output in (Ex: the path to the GPUPerfAPICounterGenerator directory)
100
-
* Param 4: Test output Dir - the directory to generate the test output in (Ex: the path to the GPUPerfAPIUnitTests/counters directory)
101
-
* Param 5: API - the API to take the counter names from (ex: DX12)
102
-
* Param 6: GPU - the GPU to take the counter names from (ex: Gfx9)
103
-
* Param 7: GPU ASIC - (optional) the subversion of GPU to take the counter names from
101
+
* With no parameters - the user interface opens with no fields prepopulated.
102
+
* With two parameters - the user interface opens with the two main fields prepopulated. Note that pressing the "Compile Public Counters" button will load the correct input files and generate the output files in the correct location.
103
+
* Param 1: **API** - the API to compile counters for (e.g. GL, CL, DX11, DX12, VK, etc.)
104
+
* Param 2: **HW generation** - the generation to compile counters for (ex: Gfx9, Gfx10, Gfx11 etc.)
105
+
* With six or seven parameters - the user interface does not open and will generate the C++ files using the specified input and output file locations.
106
+
* Param 1: **Counter names file** - text file containing hardware counter names and type (CounterNames[API][GEN].txt)
107
+
* Param 2: **Public counter definition file** - text file defining how the public counters are calculated (PublicCounterDefinitions\*.txt)
108
+
* Param 3: **Output Dir** - the directory to generate the output in (e.g. the path to the GPUPerfAPICounterGenerator directory)
109
+
* Param 4: **Test output Dir** - the directory to generate the test output in (e.g. the path to the GPUPerfAPIUnitTests/counters directory)
110
+
* Param 5: **API** - the API to take the counter names from (e.g. DX12)
111
+
* Param 6: **GPU** - the GPU to take the counter names from (e.g. Gfx11)
112
+
* Param 7: **GPU ASIC** - the subversion of GPU to take the counter names from (optional)
104
113
105
114
See the various `public_counter_definitions_*.txt` files in the [public_counter_compiler_input_files](source/public_counter_compiler_input_files) directory. These contain all the counter definitions.
106
115
Each counter is given a name, a description, a type, an optional usage type, a list of hardware counters required and a formula applied to the values of the hardware counters to calculate the value of the counter.
107
116
108
117
Counter formulas are expressed in a Reverse Polish Notation and are made up the following elements:
109
-
* numbers: these are zero-based counter indexes referring to individual counters within the list of hardware counters
110
-
* hardware counters may also be referred to by name (e.g.: GPUTime_Bottom_To_Bottom) or templated name (e.g.: SPI*_SPI_PERF_CSG_BUSY) which will automatically refer to the correct number of instances
111
-
* math operators: The supported operators are +, -, /, *
118
+
* numbers: Zero-based counter indexes referring to individual counters within the list of hardware counters
119
+
* math operators: Supported operators are +, -, /, *
112
120
* numeric literals: Numbers contained within parentheses are numeric literals (as opposed to counter indexes)
113
-
* functions: The supported functions are: min, max, sum, ifnotzero, and vcomparemax4. "max and "sum" have variants that work on multiple items at once (i.e. sum16, sum64, etc.)
114
-
* hardware params: The supported hardware params are "num_shader_engines". "num_simds", "su_clock_prim", "num_prim_pipes", and "TS_FREQ"
121
+
* functions: Supported functions are: min, max, sum, ifnotzero, and vcomparemax4. "max and "sum" have variants that work on multiple items at once (i.e. sum16, sum64, etc.)
122
+
* hardware params: Supported hardware params are "num_shader_engines", "num_simds", "su_clock_prim", "num_prim_pipes", and "TS_FREQ"
123
+
***Note:** Hardware counters may also be referred to by name (e.g.: GPUTime_Bottom_To_Bottom) or templated name (e.g.: SPI*_SPI_PERF_CSG_BUSY) which will automatically refer to the correct number of instances.
115
124
116
125
For more details, see the "EvaluateExpression" function in the [gpa_derived_counter.cc](source/gpu_perf_api_counter_generator/gpa_derived_counter.cc) file.
0 commit comments