🚚 move reduce and transpose to root dir

Tongkaio · Jun 4, 2024 · 84d9f2b · 84d9f2b
1 parent 23bc3ef
commit 84d9f2b
Show file tree

Hide file tree

Showing 11 changed files with 20 additions and 11 deletions.
diff --git a/example/reduce/README.md b/example/reduce/README.md
diff --git a/example/reduce/CMakeLists.txt → reduce/CMakeLists.txt b/example/reduce/CMakeLists.txt → reduce/CMakeLists.txt
diff --git a/reduce/README.md b/reduce/README.md
@@ -0,0 +1,20 @@
+# 归约计算
+内容：求给定数组求和
+
+1. device_reduce_v0：仅使用**全局内存**，且 N 必须是 BLOCK_SIZE 的整数倍
+2. device_reduce_v1：使用（静态）**共享内存**，不再要求 N 是 BLOCK_SIZE 的整数倍，归约的过程中不会改变全局内存的数据
+3. device_reduce_v2：在v1基础上修改，使用（动态）**共享内存**，性能不变
+4. device_reduce_v3：在v2基础上修改，通过原子函数，不再需要到CPU上再归约一次
+
+## 结果
+N=100000000，BLOCK_SIZE = 128 的测试结果：
+```
+[reduce_host]: sum = -1209.635986, total_time_h = 388.534760 ms
+[reduce_v0]: sum = -22739588.000000, total_time_0 = 31.805029 ms
+[reduce_v1]: sum = -1208.930542, total_time_1 = 19.669153 ms
+[reduce_v2]: sum = -1208.930542, total_time_2 = 19.637846 ms
+[reduce_v3]: sum = -1208.927124, total_time_3 = 15.914701 ms
+```
+
+## 参考：
+1. cuda编程基础与实践 (樊哲勇)
diff --git a/example/reduce/include/utils.cuh → reduce/include/utils.cuh b/example/reduce/include/utils.cuh → reduce/include/utils.cuh
diff --git a/example/reduce/reduce.cu → reduce/reduce.cu b/example/reduce/reduce.cu → reduce/reduce.cu
diff --git a/example/reduce/src/utils.cu → reduce/src/utils.cu b/example/reduce/src/utils.cu → reduce/src/utils.cu
diff --git a/example/transpose/CMakeLists.txt → transpose/CMakeLists.txt b/example/transpose/CMakeLists.txt → transpose/CMakeLists.txt
diff --git a/example/transpose/README.md → transpose/README.md b/example/transpose/README.md → transpose/README.md
@@ -22,7 +22,6 @@ shared_memory 优化：
 [device_transpose_v3] Average time: (10.317618) ms
 ```
 
-
 ## 参考 
 1. cuda编程基础与实践 (樊哲勇)
 2. [CUDA笔记-内存合并访问](https://zhuanlan.zhihu.com/p/641639133)

diff --git a/example/transpose/include/utils.cuh → transpose/include/utils.cuh b/example/transpose/include/utils.cuh → transpose/include/utils.cuh
diff --git a/example/transpose/src/utils.cu → transpose/src/utils.cu b/example/transpose/src/utils.cu → transpose/src/utils.cu
diff --git a/example/transpose/transpose.cu → transpose/transpose.cu b/example/transpose/transpose.cu → transpose/transpose.cu