diff --git a/parallel/para_gibbs.cu b/parallel/para_gibbs.cu
index ac50c70..d8d04d5 100644
--- a/parallel/para_gibbs.cu
+++ b/parallel/para_gibbs.cu
@@ -87,7 +87,9 @@ __global__ void seqMetroProcess(int K, int nBlocks, int *y, float *n, curandStat
 __device__ void sample_theta_seq(float *theta, float *log_theta, int *y, float *n,
                              float a, float b, int K, curandState *state);
 
-__global__ void mergePosterior(int trials, float *dev_a_out,float *dev_b_out,int *tDot);
+__global__ void mergePosterior(int trials, float *dev_a_out,float *dev_b_out,int *tDot,curandState *state, int nBlocks);
+
+__global__ void sampleTdot(int trials, int *tDot,curandState *state);
 
 int main(int argc, char **argv){
 
@@ -100,10 +102,11 @@ int main(int argc, char **argv){
     nBlocks = atoi(argv[3]);
     nThreads = atoi(argv[4]);
   }
-  else if(argc > 2)
+  else if(argc > 2){
     trials = atoi(argv[2]);
     nBlocks = 64;
     nThreads = 1;
+  }
 
   load_data(argc, argv, &K, &y, &n);
 
@@ -143,7 +146,15 @@ int main(int argc, char **argv){
   seqMetroProcess<<<nBlocks,nThreads>>>(K,nBlocks,dev_y,dev_n,devStates,dev_theta,dev_log_theta,a,b,dev_a_out,dev_b_out,trials);
   
   int *tDot;
-  mergePosterior<<<1,1>>>(trials,dev_a_out,dev_b_out,tDot);
+  CUDA_CALL(cudaMalloc((void **)&tDot,trials*sizeof(int)));
+
+  float *h;
+  CUDA_CALL(cudaMalloc((void **)&h,trials*sizeof(float)));
+
+  sampleTdot<<<nBlocks,nThreads>>>(trials, tDot,devStates);
+
+
+  mergePosterior<<<1,1>>>(trials,dev_a_out,dev_b_out,tDot,devStates,nBlocks);
  /*------ Free Memory -------------------------------------------*/
 
   free(y);
@@ -158,7 +169,16 @@ int main(int argc, char **argv){
   return EXIT_SUCCESS;
 }
 
-__global__ void mergePosterior(int trials, float *dev_a_out,float *dev_b_out,int *tDot){
+__global__ void sampleTdot(int trials, int *tDot,curandState *state){
+   int id = threadIdx.x + blockIdx.x * blockDim.x;
+   int u = (curand(&state[id]) % (trials-1)) + 1;
+   //printf("thread %d sample: %d",id,u);
+   tDot[id] = u; 
+}
+
+
+__global__ void mergePosterior(int trials, float *dev_a_out,float *dev_b_out,int *tDot, curandState *state,int nBlocks){
+   
 /*   printf("\n all blocks finished\n");
    for(int j = 0; j < trials ; j++) {
         printf(" %f ", *dev_a_out);
@@ -168,6 +188,15 @@ __global__ void mergePosterior(int trials, float *dev_a_out,float *dev_b_out,int
    }
 */
 
+ printf("\n all blocks finished\n");
+   for(int j = 0; j < nBlocks*2 ; j++) {
+       // printf(" %f ", *dev_a_out);
+        printf(" %d \n",*tDot);
+        tDot++;
+        //dev_b_out++;
+   }
+
+
 }