CUDA 配置环境(二):Windows10+QT5.14+CUDA11.3+MSVC2017

发布时间 2023-08-10 16:18:14作者: 一杯清酒邀明月

准备
安装QT,参考教程:

安装CUDA11.3

配置环境

与在VS2017中的配置不同,VS2017可以在选项卡中对CUDA的编译调试环境进行配置,而在QT中,这都需要在工程文件.pro中,通过代码来实现。下面介绍如何新建一个可运行CUDA代码的QT工程。

(1)新建一个QT Console Application工程,kit selection选择 MSVC2017 64bit

 (2)右键工程文件夹->Add new->general->empty file,新建一个hello.cu文件,并新建对应的头文件hello.h

 (3)新建的hello.cu文件会被自动放入other files文件夹下,此时的.cu文件是不参与编译调试过程的,因此接下来需要在.pro文件中添加代码,我放出添加之前的.pro中的代码供大家参考:

 1 QT -= gui
 2 
 3 CONFIG += c++11 console
 4 CONFIG -= app_bundle
 5 
 6 # The following define makes your compiler emit warnings if you use
 7 # any Qt feature that has been marked deprecated (the exact warnings
 8 # depend on your compiler). Please consult the documentation of the
 9 # deprecated API in order to know how to port your code away from it.
10 DEFINES += QT_DEPRECATED_WARNINGS
11 
12 # You can also make your code fail to compile if it uses deprecated APIs.
13 # In order to do so, uncomment the following line.
14 # You can also select to disable deprecated APIs only up to a certain version of Qt.
15 #DEFINES += QT_DISABLE_DEPRECATED_BEFORE=0x060000    # disables all the APIs deprecated before Qt 6.0.0
16 
17 SOURCES += \
18         main.cpp
19 
20 # Default rules for deployment.
21 qnx: target.path = /tmp/$${TARGET}/bin
22 else: unix:!android: target.path = /opt/$${TARGET}/bin
23 !isEmpty(target.path): INSTALLS += target
24 
25 DISTFILES += \
26     hello.cu
27 
28 HEADERS += \
29     hello.h

我们在上述代码后加入以下代码:

 1 #-------------------------------------------------
 2 # CUDA settings
 3 CUDA_SOURCES += hello.cu
 4 CUDA_DIR = "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.3/"                # Path to cuda toolkit install
 5 SYSTEM_NAME = x64                 # Depending on your system either 'Win32', 'x64', or 'Win64'
 6 SYSTEM_TYPE = 64                    # '32' or '64', depending on your system
 7 CUDA_ARCH = compute_75                 # Type of CUDA architecture
 8 CUDA_CODE = sm_75
 9 NVCC_OPTIONS = --use_fast_math
10 # include paths
11 INCLUDEPATH += "$$CUDA_DIR/include" \
12 "C:\ProgramData\NVIDIA Corporation\CUDA Samples\v11.3\common\inc"
13 # library directories
14 QMAKE_LIBDIR += "$$CUDA_DIR/lib/x64"
15 # The following makes sure all path names (which often include spaces) are put between quotation marks
16 CUDA_INC = $$join(INCLUDEPATH,'" -I"','-I"','"')
17 # Add the necessary libraries
18 CUDA_LIB_NAMES += cublas \cuda \cudadevrt \cudart \cudart_static \cufft \cufftw \curand \
19 cusolver \cusparse \nppc \nppial \nppicc \nppidei \nppif \nppig \nppim \nppist \nppisu \nppitc \
20 npps \nvblas \nvml \nvrtc \OpenCL \kernel32 \user32 \gdi32 \winspool \comdlg32 \advapi32 \shell32 \ole32 \oleaut32 \uuid \odbc32 \odbccp32 \ucrt \MSVCRT
21 
22 for(lib, CUDA_LIB_NAMES) {
23     CUDA_LIBS += $$lib.lib
24 }
25 for(lib, CUDA_LIB_NAMES) {
26     NVCC_LIBS += -l$$lib
27 }
28 LIBS += $$NVCC_LIBS
29 # The following library conflicts with something in Cuda
30 QMAKE_LFLAGS_RELEASE = /NODEFAULTLIB:msvcrt.lib
31 QMAKE_LFLAGS_DEBUG   = /NODEFAULTLIB:msvcrtd.lib
32 # MSVCRT link option (static or dynamic, it must be the same with your Qt SDK link option)
33 MSVCRT_LINK_FLAG_DEBUG   = "/MDd"
34 MSVCRT_LINK_FLAG_RELEASE = "/MD"
35 # Configuration of the Cuda compiler
36 CONFIG(debug, debug|release) {
37     # Debug mode
38     DESTDIR = debug
39     OBJECTS_DIR = debug/obj
40     CUDA_OBJECTS_DIR = debug/cuda
41     cuda_d.input = CUDA_SOURCES
42     cuda_d.output = $$CUDA_OBJECTS_DIR/${QMAKE_FILE_BASE}_cuda.o
43     cuda_d.commands = $$CUDA_DIR/bin/nvcc.exe -D_DEBUG $$NVCC_OPTIONS $$CUDA_INC $$LIBS \
44                       --machine $$SYSTEM_TYPE -arch=$$CUDA_ARCH -code=$$CUDA_CODE \
45                       --compile -cudart static -g -DWIN32 -D_MBCS \
46                       -Xcompiler "/wd4819,/EHsc,/W3,/nologo,/Od,/Zi,/RTC1" \
47                       -Xcompiler $$MSVCRT_LINK_FLAG_DEBUG \
48                       -c -o ${QMAKE_FILE_OUT} ${QMAKE_FILE_NAME}
49     cuda_d.dependency_type = TYPE_C
50     QMAKE_EXTRA_COMPILERS += cuda_d
51 }
52 else {
53     # Release mode
54     DESTDIR = release
55     OBJECTS_DIR = release/obj
56     CUDA_OBJECTS_DIR = release/cuda
57     cuda.input = CUDA_SOURCES
58     cuda.output = $$CUDA_OBJECTS_DIR/${QMAKE_FILE_BASE}_cuda.o
59     cuda.commands = $$CUDA_DIR/bin/nvcc.exe $$NVCC_OPTIONS $$CUDA_INC $$LIBS \
60                     --machine $$SYSTEM_TYPE -arch=$$CUDA_ARCH -code=$$CUDA_CODE \
61                     --compile -cudart static -D_MBCS \
62                     -Xcompiler "/wd4819,/EHsc,/W3,/nologo,/O2,/Zi" \
63                     -Xcompiler $$MSVCRT_LINK_FLAG_RELEASE \
64                     -c -o ${QMAKE_FILE_OUT} ${QMAKE_FILE_NAME}
65     cuda.dependency_type = TYPE_C
66     QMAKE_EXTRA_COMPILERS += cuda
67 }

这样就完成了对.pro文件的配置。下面介绍可能出现的问题。

Debug出现问题

在一开始debug的时候我遇到了以下问题: Could not set up the environment for Microsoft Visual Studio using

nvcc fatal   : Could not set up the environment for Microsoft Visual Studio using 'C:/Program Files (x86)/Microsoft Visual Studio/2017/Professional/VC/Tools/MSVC/14.16.27023/bin/HostX64/x64/../../../../../../../VC/Auxiliary/Build/vcvars64.bat'

我在网上找了好久的解决办法,大致的方法包括检查系统的环境变量,CUDA的路径不能有空格,nvcc的输入太长等,但这些都不能解决这个问题。

在尝试了很长时间之后,我还是决定根据报错信息入手,我找到了出现问题的vcvars64.bat这个文件,默认路径在:
C:\Program Files (x86)\Microsoft Visual Studio\2017\Professional\VC\Auxiliary\Build
并用记事本打开,其中的内容如下,只有这一行:

@call "%~dp0vcvarsall.bat" x64 %*

之后我在QT的工具->选项->Kits->编译器,发现QT中的编译器使用都是与上述vcvars64.bat同一路径下的vcvarsall.bat文件。

 因此我首先想要改变我的.pro文件调用vcvars文件的代码,有关的代码如下:

SYSTEM_TYPE = 64 

改变等号右边的值,可以改变调试时使用的vavars文件的后缀,但是他只能接收数字输入,所以也无法使其调用vcvarsall.bat,因此只能另辟蹊径,考虑将vcvarsall.bat的内容复制到vcvars64.bat,达到调用vcvarsall.bat的效果。

(5)最终解决方案,用记事本打开vcvarsall.bat,复制其中的代码到vcvars64.bat,最终vcvars64.bat中的内容为:

  1 @if not "%VSCMD_DEBUG%" GEQ "3" echo off
  2 
  3 @REM -------------------------------------------------------------------------
  4 :parse_args
  5 set __VCVARSALL_STORE=
  6 set __VCVARSALL_WINSDK=
  7 set __VCVARSALL_PARSE_ERROR=
  8 set __VCVARSALL_TARGET_ARCH=
  9 set __VCVARSALL_HOST_ARCH=
 10 set __VCVARSALL_VER=
 11 
 12 @REM Parse command line arguments.  This implementation does not care about
 13 @REM argument order.
 14 
 15 if "%VSCMD_DEBUG%" GEQ "1" (
 16     @echo [DEBUG:%~nx0] init with arguments '%*'
 17 )
 18 
 19 set __VCVARSALL_PARSE_ERROR=0
 20 set "__VCVARSALL_ARGS_LIST=%*"
 21 call :parse_loop
 22 set __VCVARSALL_ARGS_LIST=
 23 
 24 if "%VSCMD_DEBUG%" GEQ "1" (
 25     @echo [DEBUG:%~nx0] Command line parse completed with values:
 26     @echo [DEBUG:%~nx0] __VCVARSALL_TARGET_ARCH='%__VCVARSALL_TARGET_ARCH%'
 27     @echo [DEBUG:%~nx0] __VCVARSALL_HOST_ARCH='%__VCVARSALL_HOST_ARCH%'
 28     @echo [DEBUG:%~nx0] __VCVARSALL_WINSDK='%__VCVARSALL_WINSDK%'
 29     @echo [DEBUG:%~nx0] __VCVARSALL_STORE='%__VCVARSALL_STORE%'
 30     @echo [DEBUG:%~nx0] __VCVARSALL_HELP='%__VCVARSALL_HELP%'
 31     @echo [DEBUG:%~nx0] __VCVARSALL_PARSE_ERROR='%__VCVARSALL_PARSE_ERROR%'
 32 )
 33 
 34 if "%__VCVARSALL_CLEAN_ENV%" NEQ "" goto :call_vsdevcmd
 35 if "%__VCVARSALL_PARSE_ERROR%" NEQ "0" goto :usage_error
 36 if "%__VCVARSALL_HELP%" NEQ "" goto :usage
 37 
 38 @REM -------------------------------------------------------------------------
 39 :check_platform
 40 @REM This script is installed to ...\VC\Auxiliary\Build.
 41 @REM vsdevcmd is installed to    ...\Common7\Tools.
 42 if not exist "%~dp0..\..\..\Common7\Tools\vsdevcmd.bat" goto missing
 43 
 44 @REM Assemble the arguments to pass to vsdevcmd.bat
 45 if "%__VCVARSALL_TARGET_ARCH%" == "" goto :usage_error
 46 if "%__VCVARSALL_HOST_ARCH%" == "" goto :usage_error
 47 
 48 set "__VCVARSALL_VSDEVCMD_ARGS=-arch=%__VCVARSALL_TARGET_ARCH% -host_arch=%__VCVARSALL_HOST_ARCH%"
 49 if "%__VCVARSALL_WINSDK%" NEQ "" (
 50     set "__VCVARSALL_VSDEVCMD_ARGS=%__VCVARSALL_VSDEVCMD_ARGS% -winsdk=%__VCVARSALL_WINSDK%"
 51 )
 52 if "%__VCVARSALL_STORE%" NEQ "" (
 53     set "__VCVARSALL_VSDEVCMD_ARGS=%__VCVARSALL_VSDEVCMD_ARGS% -app_platform=UWP"
 54 )
 55 if "%__VCVARSALL_VER%" NEQ "" (
 56     set "__VCVARSALL_VSDEVCMD_ARGS=%__VCVARSALL_VSDEVCMD_ARGS% -vcvars_ver=%__VCVARSALL_VER%"
 57 )
 58 if "%__VCVARSALL_SPECTRE%" NEQ "" (
 59     set "__VCVARSALL_VSDEVCMD_ARGS=%__VCVARSALL_VSDEVCMD_ARGS% -vcvars_spectre_libs=%__VCVARSALL_SPECTRE%"
 60 )
 61 
 62 goto :call_vsdevcmd
 63 
 64 @REM -------------------------------------------------------------------------
 65 @REM Call vsdevcmd.bat to setup the command prompt environment
 66 
 67 :call_vsdevcmd
 68 
 69 @REM This temporary environment variable is used to control setting of VC++ 
 70 @REM command prompt-specific environment variables that should not be set
 71 @REM by the VS Developer Command prompt (specifically vsdevcmd\ext\vcvars.bat). 
 72 @REM The known case this effects is the Platform environment variable, which 
 73 @REM will override platform target for .NET builds.
 74 set VSCMD_VCVARSALL_INIT=1
 75 
 76 @REM Special handling for the /clean_env argument
 77 if "%__VCVARSALL_CLEAN_ENV%" NEQ "" (
 78     call "%~dp0..\..\..\Common7\Tools\vsdevcmd.bat" /clean_env
 79     goto :end
 80 )
 81 
 82 call "%~dp0..\..\..\Common7\Tools\vsdevcmd.bat" %__VCVARSALL_VSDEVCMD_ARGS%
 83 
 84 if "%ERRORLEVEL%"=="0" (
 85     @REM Print the target command prompt architecture...
 86     if "%__VCVARSALL_HOST_ARCH%" NEQ "%__VCVARSALL_TARGET_ARCH%" (
 87         echo [%~nx0] Environment initialized for: '%__VCVARSALL_HOST_ARCH%_%__VCVARSALL_TARGET_ARCH%'
 88     ) else (
 89         echo [%~nx0] Environment initialized for: '%__VCVARSALL_TARGET_ARCH%'
 90     )
 91 )
 92 goto :end
 93 
 94 :parse_loop
 95 for /F "tokens=1,* delims= " %%a in ("%__VCVARSALL_ARGS_LIST%") do (
 96     if "%VSCMD_DEBUG%" GEQ "2" (
 97         @echo [DEBUG:%~nx0] inner argument {%%a}
 98     )
 99     call :parse_argument %%a
100     set "__VCVARSALL_ARGS_LIST=%%b"
101     goto :parse_loop
102 )
103 
104 exit /B 0
105 
106 :parse_argument
107 
108 @REM called by :parse_loop and expects the arguments to either be:
109 @REM 1. a single argument in %1
110 @REM 2. an argument pair from the command line specified as '%1=%2'
111 
112 set __local_ARG_FOUND=
113 @REM Architecture
114 if /I "%1"=="x86" (
115     set __VCVARSALL_TARGET_ARCH=x86
116     set __VCVARSALL_HOST_ARCH=x86
117     set __local_ARG_FOUND=1
118 )
119 if /I "%1"=="x86_amd64" (
120     set __VCVARSALL_TARGET_ARCH=x64
121     set __VCVARSALL_HOST_ARCH=x86
122     set __local_ARG_FOUND=1
123 )
124 if /I "%1"=="x86_x64" (
125     set __VCVARSALL_TARGET_ARCH=x64
126     set __VCVARSALL_HOST_ARCH=x86
127     set __local_ARG_FOUND=1
128 )
129 if /I "%1"=="x86_arm" (
130     set __VCVARSALL_TARGET_ARCH=arm
131     set __VCVARSALL_HOST_ARCH=x86
132     set __local_ARG_FOUND=1
133 )
134 if /I "%1"=="x86_arm64" (
135     set __VCVARSALL_TARGET_ARCH=arm64
136     set __VCVARSALL_HOST_ARCH=x86
137     set __local_ARG_FOUND=1
138 )
139 if /I "%1"=="amd64" (
140     set __VCVARSALL_TARGET_ARCH=x64
141     set __VCVARSALL_HOST_ARCH=x64
142     set __local_ARG_FOUND=1
143 )
144 if /I "%1"=="x64" (
145     set __VCVARSALL_TARGET_ARCH=x64
146     set __VCVARSALL_HOST_ARCH=x64
147     set __local_ARG_FOUND=1
148 )
149 if /I "%1"=="amd64_x86" (
150     set __VCVARSALL_TARGET_ARCH=x86
151     set __VCVARSALL_HOST_ARCH=x64
152     set __local_ARG_FOUND=1
153 )
154 if /I "%1"=="x64_x86" (
155     set __VCVARSALL_TARGET_ARCH=x86
156     set __VCVARSALL_HOST_ARCH=x64
157     set __local_ARG_FOUND=1
158 )
159 if /I "%1"=="amd64_arm" (
160     set __VCVARSALL_TARGET_ARCH=arm
161     set __VCVARSALL_HOST_ARCH=x64
162     set __local_ARG_FOUND=1
163 )
164 if /I "%1"=="x64_arm" (
165     set __VCVARSALL_TARGET_ARCH=arm
166     set __VCVARSALL_HOST_ARCH=x64
167     set __local_ARG_FOUND=1
168 )
169 if /I "%1"=="amd64_arm64" (
170     set __VCVARSALL_TARGET_ARCH=arm64
171     set __VCVARSALL_HOST_ARCH=x64
172     set __local_ARG_FOUND=1
173 )
174 if /I "%1"=="x64_arm64" (
175     set __VCVARSALL_TARGET_ARCH=arm64
176     set __VCVARSALL_HOST_ARCH=x64
177     set __local_ARG_FOUND=1
178 )
179 if /I "%1"=="-vcvars_ver" (
180     set "__VCVARSALL_VER=%2"
181     set __local_ARG_FOUND=1
182 )
183 if /I "%1"=="/vcvars_ver" (
184     set "__VCVARSALL_VER=%2"
185     set __local_ARG_FOUND=1
186 )
187 if /I "%1"=="--vcvars_ver" (
188     set "__vcvarsall_ver=%2"
189     set __local_ARG_FOUND=1
190 )
191 if /I "%1"=="-vcvars_spectre_libs" (
192     set "__VCVARSALL_SPECTRE=%2"
193     set __local_ARG_FOUND=1
194 )
195 if /I "%1"=="/vcvars_spectre_libs" (
196     set "__VCVARSALL_SPECTRE=%2"
197     set __local_ARG_FOUND=1
198 )
199 if /I "%1"=="--vcvars_spectre_libs" (
200     set "__vcvarsall_SPECTRE=%2"
201     set __local_ARG_FOUND=1
202 )
203 if /I "%1"=="help" (
204     set __VCVARSALL_HELP=1
205     set __local_ARG_FOUND=1
206 )
207 if /I "%1"=="/help" (
208     set __VCVARSALL_HELP=1
209     set __local_ARG_FOUND=1
210 )
211 if /I "%1"=="-help" (
212     set __VCVARSALL_HELP=1
213     set __local_ARG_FOUND=1
214 )
215 if /I "%1"=="/?" (
216     set __VCVARSALL_HELP=1
217     set __local_ARG_FOUND=1
218 )
219 if /I "%1"=="-?" (
220     set __VCVARSALL_HELP=1
221     set __local_ARG_FOUND=1
222 )
223 
224 @REM -- /clean_env --
225 @REM Mostly used for internal testing to restore the state of
226 @REM the command line environment to its state prior to vcvarsall.bat
227 @REM being executed.
228 if /I "%1"=="/clean_env" (
229     set __VCVARSALL_CLEAN_ENV=/clean_env
230     set __local_ARG_FOUND=1
231 )
232 if /I "%1"=="-clean_env" (
233     set __VCVARSALL_CLEAN_ENV=/clean_env
234     set __local_ARG_FOUND=1
235 )
236 
237 @REM Windows SDK Version
238 if /I "%1"=="8.1" (
239     set "__VCVARSALL_WINSDK=8.1"
240     set __local_ARG_FOUND=1
241 )
242 
243 set __temp1=%1
244 if /I "%__temp1:~0,3%"=="10." (
245     set "__VCVARSALL_WINSDK=%1"
246     set __local_ARG_FOUND=1
247 )
248 set __temp1=
249 
250 @REM Store/UWP
251 if /I "%1"=="store" (
252     set "__VCVARSALL_STORE=-app_platform=UWP"
253     set __local_ARG_FOUND=1
254 )
255 if /I "%1"=="uwp" (
256     set "__VCVARSALL_STORE=-app_platform=UWP"
257     set __local_ARG_FOUND=1
258 )
259 
260 if "%__local_ARG_FOUND%" NEQ "1" (
261     set /A __VCVARSALL_PARSE_ERROR=__VCVARSALL_PARSE_ERROR+1
262     if "%2"=="" (
263         @echo [ERROR:%~nx0] Invalid argument found : %1
264     ) else (
265         @echo [ERROR:%~nx0] Invalid argument found : %1=%2
266     )
267 )
268 set __local_ARG_FOUND=
269 exit /B 0
270 
271 :usage_error
272 echo [ERROR:%~nx0] Error in script usage. The correct usage is:
273 goto :usage
274 
275 :usage
276 echo Syntax:
277 echo     %~nx0 [arch] [platform_type] [winsdk_version] [-vcvars_ver=vc_version] [-vcvars_spectre_libs=spectre_mode]
278 echo where : 
279 echo     [arch]: x86 ^| amd64 ^| x86_amd64 ^| x86_arm ^| x86_arm64 ^| amd64_x86 ^| amd64_arm ^| amd64_arm64
280 echo     [platform_type]: {empty} ^| store ^| uwp 
281 echo     [winsdk_version] : full Windows 10 SDK number (e.g. 10.0.10240.0) or "8.1" to use the Windows 8.1 SDK.   
282 echo     [vc_version] : {none} for default VS 2017 VC++ compiler toolset ^|
283 echo                    "14.0" for VC++ 2015 Compiler Toolset ^|
284 echo                    "14.1x" for the latest 14.1x.yyyyy toolset installed (e.g. "14.11") ^|
285 echo                    "14.1x.yyyyy" for a specific full version number (e.g. 14.11.25503)
286 echo     [spectre_mode] : {none} for default VS 2017 libraries without spectre mitigations ^|
287 echo                      "spectre" for VS 2017 libraries with spectre mitigations
288 echo:
289 echo The store parameter sets environment variables to support Universal Windows Platform application 
290 echo development and is an alias for 'uwp'.
291 echo:
292 echo For example:
293 echo     %~nx0 x86_amd64
294 echo     %~nx0 x86_amd64 10.0.10240.0
295 echo     %~nx0 x86_arm uwp 10.0.10240.0
296 echo     %~nx0 x86_arm onecore 10.0.10240.0 -vcvars_ver=14.0
297 echo     %~nx0 x64 8.1
298 echo     %~nx0 x64 store 8.1
299 echo:
300 echo Please make sure either Visual Studio or C++ Build SKU is installed.
301 goto :end
302 
303 :missing
304 echo The specified configuration type is missing.  The tools for the
305 echo configuration might not be installed.
306 goto :end
307 
308 :end
309 set __VCVARSALL_TARGET_ARCH=
310 set __VCVARSALL_HOST_ARCH=
311 set __VCVARSALL_STORE=
312 set __VCVARSALL_WINSDK=
313 set __VCVARSALL_PARSE_ERROR=
314 set __VCVARSALL_CLEAN_ENV=
315 set VSCMD_VCVARSALL_INIT=
316 set __VCVARSALL_VSDEVCMD_ARGS=
317 set __VCVARSALL_HELP=
318 set __VCVARSALL_VER=
319 set __VCVARSALL_SPECTRE=

(6)再次debug程序,编译成功

(7)下面给出hello.cu 和hello.h、main.cpp中的代码

hello.cu

 1 #include "hello.h"
 2 
 3 extern "C"
 4 __global__ void hellofromGPU(void)
 5 {
 6     printf("GPU:hello world\n");
 7 }
 8 
 9 void showhello(void)
10 {
11     hellofromGPU <<<1,10>>>();
12     cudaDeviceSynchronize();
13 
14 }

hello.h

 1 #ifndef HELLO_H
 2 #define HELLO_H
 3 #include "cuda_runtime.h"
 4 #include "device_launch_parameters.h"
 5 #include <stdio.h>
 6 #include "malloc.h"
 7 #define  WIDTH 11
 8 
 9 #define  HEIGHT 10
10 #define  X_INTER 3
11 #define  Y_INTER 3
12 #define  BLOCK_SIZE 8
13 
14 void showhello(void);
15 #endif // HELLO_H

main.cpp

1 #include<stdio.h>
2 #include "hello.h"
3 
4 int main(void)
5 {
6    showhello();
7 }

(8)运行代码,输出结果如图所示: