From d6068c8f61210f5c7833560186ceb04686df1730 Mon Sep 17 00:00:00 2001
From: Neil Gershenfeld <gersh@cba.mit.edu>
Date: Sun, 24 Oct 2021 16:12:52 -0400
Subject: [PATCH] add CUDA errors
---
CUDA/cudapi.cu | 22 +++++++++++-----------
1 file changed, 11 insertions(+), 11 deletions(-)
mode change 100755 => 100644 CUDA/cudapi.cu
diff --git a/CUDA/cudapi.cu b/CUDA/cudapi.cu
old mode 100755
new mode 100644
index 2586924..b755b84
--- a/CUDA/cudapi.cu
+++ b/CUDA/cudapi.cu
@@ -6,8 +6,8 @@
//
#include <iostream>
#include <chrono>
-#include <cstdint>
#include <string>
+using namespace std;
uint64_t blocks = 1024;
uint64_t threads = 1024;
uint64_t nloop = 1000000;
@@ -18,16 +18,6 @@ void cudaCheck(string msg) {
if (cudaSuccess != err)
cerr << msg << ": " << cudaGetErrorString(err) << endl;
}
-void reduce(double *arr) {
- uint64_t len = npts >> 1;
- while (1) {
- reduce_sum<<<blocks,threads>>>(arr,len);
- cudaCheck("reduce");
- len = len >> 1;
- if (len == 0)
- return;
- }
- }
__global__ void init(double *arr,uint64_t nloop) {
uint64_t i = blockIdx.x*blockDim.x+threadIdx.x;
uint64_t start = nloop*i+1;
@@ -41,6 +31,16 @@ __global__ void reduce_sum(double *arr,uint64_t len) {
if (i < len)
arr[i] += arr[i+len];
}
+void reduce(double *arr) {
+ uint64_t len = npts >> 1;
+ while (1) {
+ reduce_sum<<<blocks,threads>>>(arr,len);
+ cudaCheck("reduce");
+ len = len >> 1;
+ if (len == 0)
+ return;
+ }
+ }
int main(void) {
double harr[1],*darr;
cudaMalloc(&darr,npts*sizeof(double));
--
GitLab