531     float *h_a, *h_b, *h_c;  in runStream()  local
534     if ((h_a = (float *)calloc(N, sizeof(float))) == (float *)NULL) {  in runStream()
559     PetscCallCUDA(cudaMemcpy(h_a, d_a, sizeof(float) * N, cudaMemcpyDeviceToHost));  in runStream()
561     errorSTREAMkernel = STREAM_Copy_verify(h_a, h_c, N);  in runStream()
573     PetscCallCUDA(cudaMemcpy(h_a, d_a, sizeof(float) * N, cudaMemcpyDeviceToHost));  in runStream()
575     errorSTREAMkernel = STREAM_Copy_verify(h_a, h_c, N);  in runStream()
601     PetscCallCUDA(cudaMemcpy(h_a, d_a, sizeof(float) * N, cudaMemcpyDeviceToHost));  in runStream()
604     errorSTREAMkernel = STREAM_Add_verify(h_a, h_b, h_c, N);  in runStream()
616     PetscCallCUDA(cudaMemcpy(h_a, d_a, sizeof(float) * N, cudaMemcpyDeviceToHost));  in runStream()
619     errorSTREAMkernel = STREAM_Triad_verify(h_b, h_c, h_a, scalar, N);  in runStream()
625     free(h_a);  in runStream()
745     double *h_a, *h_b, *h_c;  in runStreamDouble()  local
748     if ((h_a = (double *)calloc(N, sizeof(double))) == (double *)NULL) {  in runStreamDouble()
773     PetscCallCUDA(cudaMemcpy(h_a, d_a, sizeof(double) * N, cudaMemcpyDeviceToHost));  in runStreamDouble()
775     errorSTREAMkernel = STREAM_Copy_verify_double(h_a, h_c, N);  in runStreamDouble()
787     PetscCallCUDA(cudaMemcpy(h_a, d_a, sizeof(double) * N, cudaMemcpyDeviceToHost));  in runStreamDouble()
789     errorSTREAMkernel = STREAM_Copy_verify_double(h_a, h_c, N);  in runStreamDouble()
814     PetscCallCUDA(cudaMemcpy(h_a, d_a, sizeof(double) * N, cudaMemcpyDeviceToHost));  in runStreamDouble()
817     errorSTREAMkernel = STREAM_Add_verify_double(h_a, h_b, h_c, N);  in runStreamDouble()
829     PetscCallCUDA(cudaMemcpy(h_a, d_a, sizeof(double) * N, cudaMemcpyDeviceToHost));  in runStreamDouble()
832     errorSTREAMkernel = STREAM_Triad_verify_double(h_b, h_c, h_a, scalar, N);  in runStreamDouble()
838     free(h_a);  in runStreamDouble()