Merge pull request #143 from vais-ral/cudareset

adding cudaDeviceSynchronize() before exit

10 files changed, 11 insertions, 6 deletions

diff --git a/src/Core/regularisers_GPU/Diffus_4thO_GPU_core.cu b/src/Core/regularisers_GPU/Diffus_4thO_GPU_core.cu
index afd2026..3cdf0a3 100644
--- a/src/Core/regularisers_GPU/Diffus_4thO_GPU_core.cu
+++ b/src/Core/regularisers_GPU/Diffus_4thO_GPU_core.cu
@@ -363,5 +363,6 @@ extern "C" int Diffus4th_GPU_main(float *Input, float *Output, float *infovector
         /*adding info into info_vector */
         infovector[0] = (float)(n);  /*iterations number (if stopped earlier based on tolerance)*/
         infovector[1] = re;  /* reached tolerance */
+        cudaDeviceSynchronize();
         return 0;
 }
diff --git a/src/Core/regularisers_GPU/LLT_ROF_GPU_core.cu b/src/Core/regularisers_GPU/LLT_ROF_GPU_core.cu
index 3c578f3..3109cfa 100644
--- a/src/Core/regularisers_GPU/LLT_ROF_GPU_core.cu
+++ b/src/Core/regularisers_GPU/LLT_ROF_GPU_core.cu
@@ -599,6 +599,6 @@ extern "C" int LLT_ROF_GPU_main(float *Input, float *Output,  float *infovector,
 
         infovector[0] = (float)(n);  /*iterations number (if stopped earlier based on tolerance)*/
         infovector[1] = re;  /* reached tolerance */
-
+        cudaDeviceSynchronize();
         return 0;
 }
diff --git a/src/Core/regularisers_GPU/NonlDiff_GPU_core.cu b/src/Core/regularisers_GPU/NonlDiff_GPU_core.cu
index d7a6bac..1d592d6 100644
--- a/src/Core/regularisers_GPU/NonlDiff_GPU_core.cu
+++ b/src/Core/regularisers_GPU/NonlDiff_GPU_core.cu
@@ -545,5 +545,6 @@ extern "C" int NonlDiff_GPU_main(float *Input, float *Output, float *infovector,
 
         infovector[0] = (float)(n);  /*iterations number (if stopped earlier based on tolerance)*/
         infovector[1] = re;  /* reached tolerance */
+        cudaDeviceSynchronize();
         return 0;
 }
diff --git a/src/Core/regularisers_GPU/PatchSelect_GPU_core.cu b/src/Core/regularisers_GPU/PatchSelect_GPU_core.cu
index fb6fa95..cde252e 100644
--- a/src/Core/regularisers_GPU/PatchSelect_GPU_core.cu
+++ b/src/Core/regularisers_GPU/PatchSelect_GPU_core.cu
@@ -505,6 +505,6 @@ extern "C" int PatchSelect_GPU_main(float *A, unsigned short *H_i, unsigned shor
     cudaFree(H_j_d);
     cudaFree(Weights_d);
     cudaFree(Eucl_Vec_d);
-    cudaDeviceReset();
+    cudaDeviceSynchronize();
     return 0;
 }
diff --git a/src/Core/regularisers_GPU/TGV_GPU_core.cu b/src/Core/regularisers_GPU/TGV_GPU_core.cu
index fc462fe..4e5e830 100644
--- a/src/Core/regularisers_GPU/TGV_GPU_core.cu
+++ b/src/Core/regularisers_GPU/TGV_GPU_core.cu
@@ -816,6 +816,6 @@ extern "C" int TGV_GPU_main(float *U0, float *U, float *infovector, float lambda
         /*adding info into info_vector */
         infovector[0] = (float)(n);  /*iterations number (if stopped earlier based on tolerance)*/
         infovector[1] = re;  /* reached tolerance */
-
+        cudaDeviceSynchronize();
         return 0;
 }
diff --git a/src/Core/regularisers_GPU/TV_FGP_GPU_core.cu b/src/Core/regularisers_GPU/TV_FGP_GPU_core.cu
index ce2548f..049dae1 100755
--- a/src/Core/regularisers_GPU/TV_FGP_GPU_core.cu
+++ b/src/Core/regularisers_GPU/TV_FGP_GPU_core.cu
@@ -593,9 +593,9 @@ extern "C" int TV_FGP_GPU_main(float *Input, float *Output, float *infovector, f
             cudaFree(R2);
             cudaFree(R3);
     }
-    //cudaDeviceReset();
     /*adding info into info_vector */
     infovector[0] = (float)(i);  /*iterations number (if stopped earlier based on tolerance)*/
     infovector[1] = re;  /* reached tolerance */
+    cudaDeviceSynchronize();
     return 0;
 }
diff --git a/src/Core/regularisers_GPU/TV_PD_GPU_core.cu b/src/Core/regularisers_GPU/TV_PD_GPU_core.cu
index 01e0ab5..3fcd1e8 100644
--- a/src/Core/regularisers_GPU/TV_PD_GPU_core.cu
+++ b/src/Core/regularisers_GPU/TV_PD_GPU_core.cu
@@ -515,5 +515,6 @@ extern "C" int TV_PD_GPU_main(float *Input, float *Output, float *infovector, fl
    /*adding info into info_vector */
    infovector[0] = (float)(i);  /*iterations number (if stopped earlier based on tolerance)*/
    infovector[1] = re;  /* reached tolerance */
+   cudaDeviceSynchronize();
    return 0;
 }
diff --git a/src/Core/regularisers_GPU/TV_ROF_GPU_core.cu b/src/Core/regularisers_GPU/TV_ROF_GPU_core.cu
index c6c6e88..e9e6d9b 100755
--- a/src/Core/regularisers_GPU/TV_ROF_GPU_core.cu
+++ b/src/Core/regularisers_GPU/TV_ROF_GPU_core.cu
@@ -480,7 +480,8 @@ extern "C" int TV_ROF_GPU_main(float* Input, float* Output, float *infovector, f
         CHECK(cudaFree(d_D2));
         CHECK(cudaFree(lambdaPar_d));
 
-	      infovector[0] = (float)(n);  /*iterations number (if stopped earlier based on tolerance)*/
+	infovector[0] = (float)(n);  /*iterations number (if stopped earlier based on tolerance)*/
         infovector[1] = re;  /* reached tolerance */
+        cudaDeviceSynchronize();
         return 0;
 }
diff --git a/src/Core/regularisers_GPU/TV_SB_GPU_core.cu b/src/Core/regularisers_GPU/TV_SB_GPU_core.cu
index 0353868..58b1157 100755
--- a/src/Core/regularisers_GPU/TV_SB_GPU_core.cu
+++ b/src/Core/regularisers_GPU/TV_SB_GPU_core.cu
@@ -528,6 +528,6 @@ extern "C" int TV_SB_GPU_main(float *Input, float *Output, float *infovector, fl
     /*adding info into info_vector */
     infovector[0] = (float)(ll);  /*iterations number (if stopped earlier based on tolerance)*/
     infovector[1] = re;  /* reached tolerance */
-
+    cudaDeviceSynchronize();
     return 0;
 }
diff --git a/src/Core/regularisers_GPU/dTV_FGP_GPU_core.cu b/src/Core/regularisers_GPU/dTV_FGP_GPU_core.cu
index 89fca06..12473c0 100644
--- a/src/Core/regularisers_GPU/dTV_FGP_GPU_core.cu
+++ b/src/Core/regularisers_GPU/dTV_FGP_GPU_core.cu
@@ -753,5 +753,6 @@ extern "C" int dTV_FGP_GPU_main(float *Input, float *InputRef, float *Output, fl
     /*adding info into info_vector */
     infovector[0] = (float)(i);  /*iterations number (if stopped earlier based on tolerance)*/
     infovector[1] = re;  /* reached tolerance */
+    cudaDeviceSynchronize();
     return 0;
 }