Add and update samples for cuda 11.0 support

2026-10-23 13:18:16 +08:00 · 2020-05-18 22:22:06 +05:30
parent 6be514679b
commit 4f6e02970b
410 changed files with 21920 additions and 3864 deletions
--- a/Samples/cuSolverSp_LinearSolver/Makefile
+++ b/Samples/cuSolverSp_LinearSolver/Makefile
@@ -72,9 +72,9 @@ endif
 # architecture
 HOST_ARCH   := $(shell uname -m)
 TARGET_ARCH ?= $(HOST_ARCH)
-ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 ppc64le armv7l))
+ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 sbsa ppc64le armv7l))
    ifneq ($(TARGET_ARCH),$(HOST_ARCH))
-        ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 ppc64le))
+        ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 sbsa ppc64le))
            TARGET_SIZE := 64
        else ifneq (,$(filter $(TARGET_ARCH),armv7l))
            TARGET_SIZE := 32
@@ -85,8 +85,17 @@ ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 ppc64le armv7l))
 else
    $(error ERROR - unsupported value $(TARGET_ARCH) for TARGET_ARCH!)
 endif
+
+# sbsa and aarch64 systems look similar. Need to differentiate them at host level for now.
+ifeq ($(HOST_ARCH),aarch64)
+    ifeq ($(CUDA_PATH)/targets/sbsa-linux,$(shell ls -1d $(CUDA_PATH)/targets/sbsa-linux))
+        HOST_ARCH := sbsa
+        TARGET_ARCH := sbsa
+    endif
+endif
+
 ifneq ($(TARGET_ARCH),$(HOST_ARCH))
-    ifeq (,$(filter $(HOST_ARCH)-$(TARGET_ARCH),aarch64-armv7l x86_64-armv7l x86_64-aarch64 x86_64-ppc64le))
+    ifeq (,$(filter $(HOST_ARCH)-$(TARGET_ARCH),aarch64-armv7l x86_64-armv7l x86_64-aarch64 x86_64-sbsa x86_64-ppc64le))
        $(error ERROR - cross compiling from $(HOST_ARCH) to $(TARGET_ARCH) is not supported!)
    endif
 endif
@@ -141,6 +150,8 @@ else ifneq ($(TARGET_ARCH),$(HOST_ARCH))
        else ifeq ($(TARGET_OS), android)
            HOST_COMPILER ?= aarch64-linux-android-clang++
        endif
+    else ifeq ($(TARGET_ARCH),sbsa)
+        HOST_COMPILER ?= aarch64-linux-gnu-g++
    else ifeq ($(TARGET_ARCH),ppc64le)
        HOST_COMPILER ?= powerpc64le-linux-gnu-g++
    endif
@@ -185,19 +196,27 @@ ifneq ($(TARGET_ARCH),$(HOST_ARCH))
                CCFLAGS += --sysroot=$(TARGET_FS)
            endif
            LDFLAGS += --sysroot=$(TARGET_FS)
-            LDFLAGS += -rpath-link=$(TARGET_FS)/lib -L $(TARGET_FS)/lib
-            LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib -L $(TARGET_FS)/usr/lib
-            LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib/aarch64-linux-gnu -L $(TARGET_FS)/usr/lib/aarch64-linux-gnu
+            LDFLAGS += -rpath-link=$(TARGET_FS)/lib -L$(TARGET_FS)/lib
+            LDFLAGS += -rpath-link=$(TARGET_FS)/lib/aarch64-linux-gnu -L$(TARGET_FS)/lib/aarch64-linux-gnu
+            LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib -L$(TARGET_FS)/usr/lib
+            LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib/aarch64-linux-gnu -L$(TARGET_FS)/usr/lib/aarch64-linux-gnu
            LDFLAGS += --unresolved-symbols=ignore-in-shared-libs
-            CCFLAGS += -isystem=$(TARGET_FS)/usr/include
-            CCFLAGS += -isystem=$(TARGET_FS)/usr/include/aarch64-linux-gnu
+            CCFLAGS += -isystem=$(TARGET_FS)/usr/include  -I$(TARGET_FS)/usr/include
+            CCFLAGS += -isystem=$(TARGET_FS)/usr/include/aarch64-linux-gnu -I$(TARGET_FS)/usr/include/aarch64-linux-gnu
+        endif
+    endif
+    ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-qnx)
+        CCFLAGS += -DWIN_INTERFACE_CUSTOM -I/usr/include/aarch64-qnx-gnu
+        LDFLAGS += -lsocket
+        LDFLAGS += -rpath=/usr/lib/aarch64-qnx-gnu -L/usr/lib/aarch64-qnx-gnu
+        ifneq ($(TARGET_FS),)
+            LDFLAGS += -rpath=$(TARGET_FS)/usr/lib -L $(TARGET_FS)/usr/lib
+            LDFLAGS += -rpath=$(TARGET_FS)/usr/libnvidia -L $(TARGET_FS)/usr/libnvidia
+        endif
+        ifdef TARGET_OVERRIDE # cuda toolkit targets override
+            NVCCFLAGS += -target-dir $(TARGET_OVERRIDE)
        endif
    endif
-endif
-
-ifeq ($(TARGET_OS),qnx)
-    CCFLAGS += -DWIN_INTERFACE_CUSTOM
-    LDFLAGS += -lsocket
 endif

 # Install directory of different arch
@@ -206,6 +225,8 @@ ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-linux)
    CUDA_INSTALL_TARGET_DIR = targets/armv7-linux-gnueabihf/
 else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-linux)
    CUDA_INSTALL_TARGET_DIR = targets/aarch64-linux/
+else ifeq ($(TARGET_ARCH)-$(TARGET_OS),sbsa-linux)
+    CUDA_INSTALL_TARGET_DIR = targets/sbsa-linux/
 else ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-android)
    CUDA_INSTALL_TARGET_DIR = targets/armv7-linux-androideabi/
 else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-android)
@@ -251,9 +272,9 @@ LIBRARIES :=

 # Gencode arguments
 ifeq ($(TARGET_ARCH),$(filter $(TARGET_ARCH),armv7l aarch64))
-SMS ?= 30 35 37 50 52 60 61 70 72 75
+SMS ?= 35 37 50 52 60 61 70 72 75 80
 else
-SMS ?= 30 35 37 50 52 60 61 70 75
+SMS ?= 35 37 50 52 60 61 70 75 80
 endif

 ifeq ($(SMS),)
--- a/Samples/cuSolverSp_LinearSolver/NsightEclipse.xml
+++ b/Samples/cuSolverSp_LinearSolver/NsightEclipse.xml
@@ -45,7 +45,6 @@
    <scope>1:CUDA Basic Topics</scope>
    <scope>3:Linear Algebra</scope>
  </scopes>
-  <sm-arch>sm30</sm-arch>
  <sm-arch>sm35</sm-arch>
  <sm-arch>sm37</sm-arch>
  <sm-arch>sm50</sm-arch>
@@ -55,6 +54,7 @@
  <sm-arch>sm70</sm-arch>
  <sm-arch>sm72</sm-arch>
  <sm-arch>sm75</sm-arch>
+  <sm-arch>sm80</sm-arch>
  <supported_envs>
    <env>
      <arch>x86_64</arch>
--- a/Samples/cuSolverSp_LinearSolver/README.md
+++ b/Samples/cuSolverSp_LinearSolver/README.md
@@ -10,11 +10,11 @@ Linear Algebra, CUSOLVER Library

 ## Supported SM Architectures

-[SM 3.0 ](https://developer.nvidia.com/cuda-gpus)  [SM 3.5 ](https://developer.nvidia.com/cuda-gpus)  [SM 3.7 ](https://developer.nvidia.com/cuda-gpus)  [SM 5.0 ](https://developer.nvidia.com/cuda-gpus)  [SM 5.2 ](https://developer.nvidia.com/cuda-gpus)  [SM 6.0 ](https://developer.nvidia.com/cuda-gpus)  [SM 6.1 ](https://developer.nvidia.com/cuda-gpus)  [SM 7.0 ](https://developer.nvidia.com/cuda-gpus)  [SM 7.2 ](https://developer.nvidia.com/cuda-gpus)  [SM 7.5 ](https://developer.nvidia.com/cuda-gpus)
+[SM 3.5 ](https://developer.nvidia.com/cuda-gpus)  [SM 3.7 ](https://developer.nvidia.com/cuda-gpus)  [SM 5.0 ](https://developer.nvidia.com/cuda-gpus)  [SM 5.2 ](https://developer.nvidia.com/cuda-gpus)  [SM 6.0 ](https://developer.nvidia.com/cuda-gpus)  [SM 6.1 ](https://developer.nvidia.com/cuda-gpus)  [SM 7.0 ](https://developer.nvidia.com/cuda-gpus)  [SM 7.2 ](https://developer.nvidia.com/cuda-gpus)  [SM 7.5 ](https://developer.nvidia.com/cuda-gpus)  [SM 8.0 ](https://developer.nvidia.com/cuda-gpus)

 ## Supported OSes

-Linux, Windows, MacOSX
+Linux, Windows

 ## Supported CPU Architecture

@@ -27,7 +27,7 @@ x86_64, ppc64le, armv7l

 ## Prerequisites

-Download and install the [CUDA Toolkit 10.2](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
+Download and install the [CUDA Toolkit 11.0](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
 Make sure the dependencies mentioned in [Dependencies]() section above are installed.

 ## Build and Run
@@ -67,29 +67,5 @@ The samples makefiles can take advantage of certain options:
    $ make HOST_COMPILER=g++
 ```

-### Mac
-The Mac samples are built using makefiles. To use the makefiles, change directory into the sample directory you wish to build, and run make:
-```
-$ cd <sample_dir>
-$ make
-```
-
-The samples makefiles can take advantage of certain options:
-
-*  **dbg=1** - build with debug symbols
-    ```
-    $ make dbg=1
-    ```
-
-*  **SMS="A B ..."** - override the SM architectures for which the sample will be built, where "A B ..." is a space-delimited list of SM architectures. For example, to generate SASS for SM 50 and SM 60, use SMS="50 60".
-    ```
-    $ make SMS="A B ..."
-    ```
-
-*  **HOST_COMPILER=<host_compiler>** - override the default clang host compiler. See the [Mac Installation Guide](http://docs.nvidia.com/cuda/cuda-installation-guide-mac-os-x/index.html#system-requirements) for a list of supported host compilers.
-    ```
-    $ make HOST_COMPILER=clang
-    ```
-
 ## References (for more details)

--- a/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver.cpp
+++ b/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver.cpp
@@ -24,53 +24,63 @@
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */
+
 /*
 *  Test three linear solvers, including Cholesky, LU and QR.
- *  The user has to prepare a sparse matrix of "matrix market format" (with extension .mtx).
- *  For example, the user can download matrices in Florida Sparse Matrix Collection.
+ *  The user has to prepare a sparse matrix of "matrix market format" (with
+ extension .mtx).
+ *  For example, the user can download matrices in Florida Sparse Matrix
+ Collection.
 *  (http://www.cise.ufl.edu/research/sparse/matrices/)
 *
 *  The user needs to choose a solver by the switch -R<solver> and
 *  to provide the path of the matrix by the switch -F<file>, then
 *  the program solves
- *          A*x = b  
+ *          A*x = b
 *  and reports relative error
 *          |b-A*x|/(|A|*|x|+|b|)
 *
 *  How does it work?
 *     The example solves A*x = b by the following steps
- *  step 1: B = A(Q,Q) 
+ *  step 1: B = A(Q,Q)
 *     Q is the ordering to minimize zero fill-in.
 *     The user can choose symrcm or symamd.
 *  step 2: solve B*z = Q*b
 *  step 3: x = inv(Q)*z
- * 
+ *
 *  Above three steps can be combined by the formula
 *        (Q*A*Q')*(Q*x) = (Q*b)
 *
- *  The elapsed time is also reported so the user can compare efficiency of different solvers.
+ *  The elapsed time is also reported so the user can compare efficiency of
+ different solvers.
 *
 *  How to use
-        /cuSolverSp_LinearSolver            // Default: Cholesky, symrcm & file=lap2D_5pt_n100.mtx
- *     ./cuSolverSp_LinearSolver -R=chol  -file=<file>   // cholesky factorization
- *     ./cuSolverSp_LinearSolver -R=lu -P=symrcm -file=<file>     // symrcm + LU with partial pivoting
- *     ./cuSolverSp_LinearSolver -R=qr -P=symamd -file=<file>     // symamd + QR factorization
+        /cuSolverSp_LinearSolver            // Default: Cholesky, symrcm &
+ file=lap2D_5pt_n100.mtx
+ *     ./cuSolverSp_LinearSolver -R=chol  -file=<file>   // cholesky
+ factorization
+ *     ./cuSolverSp_LinearSolver -R=lu -P=symrcm -file=<file>     // symrcm + LU
+ with partial pivoting
+ *     ./cuSolverSp_LinearSolver -R=qr -P=symamd -file=<file>     // symamd + QR
+ factorization
 *
 *
- *  Remark: the absolute error on solution x is meaningless without knowing condition number of A.
- *     The relative error on residual should be close to machine zero, i.e. 1.e-15.
+ *  Remark: the absolute error on solution x is meaningless without knowing
+ condition number of A.
+ *     The relative error on residual should be close to machine zero,
+ i.e. 1.e-15.
 */

+#include <assert.h>
+#include <ctype.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
-#include <ctype.h>
-#include <assert.h>

 #include <cuda_runtime.h>

-#include "cusparse.h"
 #include "cusolverSp.h"
+#include "cusparse.h"

 #include "helper_cuda.h"
 #include "helper_cusolver.h"
@@ -461,10 +471,38 @@ int main(int argc, char *argv[]) {
                                  cudaMemcpyDeviceToDevice, stream));
  checkCudaErrors(cudaMemcpyAsync(d_x, h_x, sizeof(double) * colsA,
                                  cudaMemcpyHostToDevice, stream));
-  checkCudaErrors(cusparseDcsrmv(cusparseHandle,
-                                 CUSPARSE_OPERATION_NON_TRANSPOSE, rowsA, colsA,
-                                 nnzA, &minus_one, descrA, d_csrValA,
-                                 d_csrRowPtrA, d_csrColIndA, d_x, &one, d_r));
+
+  /* Wrap raw data into cuSPARSE generic API objects */
+  cusparseSpMatDescr_t matA = NULL;
+  if (baseA) {
+    checkCudaErrors(cusparseCreateCsr(&matA, rowsA, colsA, nnzA, d_csrRowPtrA,
+                                      d_csrColIndA, d_csrValA,
+                                      CUSPARSE_INDEX_32I, CUSPARSE_INDEX_32I,
+                                      CUSPARSE_INDEX_BASE_ONE, CUDA_R_64F));
+  } else {
+    checkCudaErrors(cusparseCreateCsr(&matA, rowsA, colsA, nnzA, d_csrRowPtrA,
+                                      d_csrColIndA, d_csrValA,
+                                      CUSPARSE_INDEX_32I, CUSPARSE_INDEX_32I,
+                                      CUSPARSE_INDEX_BASE_ZERO, CUDA_R_64F));
+  }
+
+  cusparseDnVecDescr_t vecx = NULL;
+  checkCudaErrors(cusparseCreateDnVec(&vecx, colsA, d_x, CUDA_R_64F));
+  cusparseDnVecDescr_t vecAx = NULL;
+  checkCudaErrors(cusparseCreateDnVec(&vecAx, rowsA, d_r, CUDA_R_64F));
+
+  /* Allocate workspace for cuSPARSE */
+  size_t bufferSize = 0;
+  checkCudaErrors(cusparseSpMV_bufferSize(
+      cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE, &minus_one, matA, vecx,
+      &one, vecAx, CUDA_R_64F, CUSPARSE_MV_ALG_DEFAULT, &bufferSize));
+  void *buffer = NULL;
+  checkCudaErrors(cudaMalloc(&buffer, bufferSize));
+
+  checkCudaErrors(cusparseSpMV(cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
+                               &minus_one, matA, vecx, &one, vecAx, CUDA_R_64F,
+                               CUSPARSE_MV_ALG_DEFAULT, &buffer));
+
  checkCudaErrors(cudaMemcpyAsync(h_r, d_r, sizeof(double) * rowsA,
                                  cudaMemcpyDeviceToHost, stream));
  /* wait until h_r is ready */
@@ -518,10 +556,11 @@ int main(int argc, char *argv[]) {
  printf("step 8: evaluate residual r = b - A*x (result on GPU)\n");
  checkCudaErrors(cudaMemcpyAsync(d_r, d_b, sizeof(double) * rowsA,
                                  cudaMemcpyDeviceToDevice, stream));
-  checkCudaErrors(cusparseDcsrmv(cusparseHandle,
-                                 CUSPARSE_OPERATION_NON_TRANSPOSE, rowsA, colsA,
-                                 nnzA, &minus_one, descrA, d_csrValA,
-                                 d_csrRowPtrA, d_csrColIndA, d_x, &one, d_r));
+
+  checkCudaErrors(cusparseSpMV(cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
+                               &minus_one, matA, vecx, &one, vecAx, CUDA_R_64F,
+                               CUSPARSE_MV_ALG_DEFAULT, &buffer));
+
  checkCudaErrors(cudaMemcpyAsync(h_x, d_x, sizeof(double) * colsA,
                                  cudaMemcpyDeviceToHost, stream));
  checkCudaErrors(cudaMemcpyAsync(h_r, d_r, sizeof(double) * rowsA,
@@ -566,6 +605,15 @@ int main(int argc, char *argv[]) {
  if (descrA) {
    checkCudaErrors(cusparseDestroyMatDescr(descrA));
  }
+  if (matA) {
+    checkCudaErrors(cusparseDestroySpMat(matA));
+  }
+  if (vecx) {
+    checkCudaErrors(cusparseDestroyDnVec(vecx));
+  }
+  if (vecAx) {
+    checkCudaErrors(cusparseDestroyDnVec(vecAx));
+  }

  if (h_csrValA) {
    free(h_csrValA);
--- a/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2012.vcxproj
+++ b/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2012.vcxproj
@@ -62,7 +62,7 @@
      <OutputFile>$(OutDir)/cuSolverSp_LinearSolver.exe</OutputFile>
    </Link>
    <CudaCompile>
-      <CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;</CodeGeneration>
+      <CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;</CodeGeneration>
      <AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
      <Include>./;../../Common</Include>
      <Defines>WIN32</Defines>
--- a/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2013.vcxproj
+++ b/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2013.vcxproj
@@ -62,7 +62,7 @@
      <OutputFile>$(OutDir)/cuSolverSp_LinearSolver.exe</OutputFile>
    </Link>
    <CudaCompile>
-      <CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;</CodeGeneration>
+      <CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;</CodeGeneration>
      <AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
      <Include>./;../../Common</Include>
      <Defines>WIN32</Defines>
--- a/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2015.vcxproj
+++ b/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2015.vcxproj
@@ -62,7 +62,7 @@
      <OutputFile>$(OutDir)/cuSolverSp_LinearSolver.exe</OutputFile>
    </Link>
    <CudaCompile>
-      <CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;</CodeGeneration>
+      <CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;</CodeGeneration>
      <AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
      <Include>./;../../Common</Include>
      <Defines>WIN32</Defines>
--- a/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2017.vcxproj
+++ b/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2017.vcxproj
@@ -67,7 +67,7 @@
      <OutputFile>$(OutDir)/cuSolverSp_LinearSolver.exe</OutputFile>
    </Link>
    <CudaCompile>
-      <CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;</CodeGeneration>
+      <CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;</CodeGeneration>
      <AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
      <Include>./;../../Common</Include>
      <Defines>WIN32</Defines>
--- a/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2019.vcxproj
+++ b/Samples/cuSolverSp_LinearSolver/cuSolverSp_LinearSolver_vs2019.vcxproj
@@ -63,7 +63,7 @@
      <OutputFile>$(OutDir)/cuSolverSp_LinearSolver.exe</OutputFile>
    </Link>
    <CudaCompile>
-      <CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;</CodeGeneration>
+      <CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;</CodeGeneration>
      <AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
      <Include>./;../../Common</Include>
      <Defines>WIN32</Defines>