ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-13 09:45:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

add lds doble buffer to nchw padded v4r1 and v4r4

Browse Source
This commit is contained in:
Chao Liu
2019-09-15 16:58:16 -05:00
parent 2c93b3057d
commit bf97542846
7 changed files with 153 additions and 135 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
driver/src/driver.cpp
View File

@@ -92,8 +92,8 @@ int main(int argc, char* argv[])
// 3x3, 34x34
constexpr index_t N = 64;
constexpr index_t C = 256;
constexpr index_t HI = 34;
constexpr index_t WI = 34;
constexpr index_t HI = 32;
constexpr index_t WI = 32;
constexpr index_t K = 128;
constexpr index_t Y = 3;
constexpr index_t X = 3;
@@ -101,8 +101,8 @@ int main(int argc, char* argv[])
using ConvStrides = Sequence<1, 1>;
using ConvDilations = Sequence<1, 1>;
using LeftPads = Sequence<0, 0>;
using RightPads = Sequence<0, 0>;
using LeftPads = Sequence<1, 1>;
using RightPads = Sequence<1, 1>;
#elif 0
// 1x1 filter, 8x8 image
// cudnn@V100 68%, ck@V100 72%, ck@P100 52%, ck@VII 42%