From 83bb53f997754c2530718251680abaab33f28746 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Onur=20=C3=9Clgen?= <onur.ulgen@kcl.ac.uk>
Date: Fri, 2 Feb 2024 18:19:25 +0000
Subject: [PATCH] CUDA Test 6

---
 .github/workflows/release.yml | 15 ++++++++++-----
 niftyreg_build_version.txt    |  2 +-
 reg-lib/cuda/CMakeLists.txt   | 16 ++++++++--------
 3 files changed, 19 insertions(+), 14 deletions(-)

diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index a7382e6d..6aff37e4 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -10,13 +10,19 @@ jobs:
     strategy:
       matrix:
         os: [ubuntu-20.04, macos-latest, windows-2019]
+        platform: [cpu, cuda]
         include:
+          - platform: cpu
+            platform-name: ""
+            use-cuda: "OFF"
+          - platform: cuda
+            platform-name: "-CUDA"
+            use-cuda: "ON"
           - os: ubuntu-20.04    # For Ubuntu only
             os-name: "Ubuntu"
-            use_cuda: "ON"
           - os: macos-latest    # For macOS only
             os-name: "macOS"
-            use_cuda: "OFF"
+            use-cuda: "OFF"
           - sudo: "sudo"        # For Ubuntu and macOS
             c-compiler: "gcc"
             cxx-compiler: "g++"
@@ -25,7 +31,6 @@ jobs:
             sudo: ""
             c-compiler: "cl.exe"
             cxx-compiler: "cl.exe"
-            use_cuda: "ON"
           - build_type: "Release" # For all platforms
 
     steps:
@@ -54,7 +59,7 @@ jobs:
                 -DCMAKE_C_COMPILER=${{ matrix.c-compiler }} \
                 -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} \
                 -DBUILD_ALL_DEP=ON \
-                -DUSE_CUDA=${{ matrix.use_cuda }} \
+                -DUSE_CUDA=${{ matrix.use-cuda }} \
                 -DUSE_OPENCL=OFF \
                 -DUSE_SSE=ON \
                 -DUSE_OPENMP=ON \
@@ -94,4 +99,4 @@ jobs:
         with:
           repo_token: ${{ github.token }}
           file: build/NiftyReg.zip
-          asset_name: NiftyReg-${{ matrix.os-name }}-${{ steps.get_tag.outputs.tag }}.zip
\ No newline at end of file
+          asset_name: NiftyReg-${{ matrix.os-name }}${{ matrix.platform-name }}-${{ steps.get_tag.outputs.tag }}.zip
\ No newline at end of file
diff --git a/niftyreg_build_version.txt b/niftyreg_build_version.txt
index 45843d29..066cbfe9 100644
--- a/niftyreg_build_version.txt
+++ b/niftyreg_build_version.txt
@@ -1 +1 @@
-399
+401
diff --git a/reg-lib/cuda/CMakeLists.txt b/reg-lib/cuda/CMakeLists.txt
index b2272566..c4e43bc7 100755
--- a/reg-lib/cuda/CMakeLists.txt
+++ b/reg-lib/cuda/CMakeLists.txt
@@ -30,15 +30,15 @@
     #     return()
     # endif("${RUN_OUTPUT_VAR}" LESS "30")
     # string(REPLACE "." "" CAPABILITY_CODE ${RUN_OUTPUT_VAR})
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_52,code=sm_52")
+    # set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_52,code=sm_52")
     set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_60,code=sm_60")
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_61,code=sm_61")
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_70,code=sm_70")
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_75,code=sm_75")
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_80,code=sm_80")
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_86,code=sm_86")
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_89,code=sm_89")
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_89,code=compute_89")
+    # set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_61,code=sm_61")
+    # set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_70,code=sm_70")
+    # set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_75,code=sm_75")
+    # set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_80,code=sm_80")
+    # set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_86,code=sm_86")
+    # set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_89,code=sm_89")
+    # set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_89,code=compute_89")
     # If desired, add PIC flags
     if(CMAKE_POSITION_INDEPENDENT_CODE AND DEFINED CMAKE_C_COMPILE_OPTIONS_PIC)
         # Add (undocumented) CMake flag that should tell the host compiler to generate position independent code