ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-11 17:00:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

update flops calculation

Browse Source
This commit is contained in:
Chao Liu
2019-04-10 15:35:00 -05:00
parent 5696c81ffd
commit edc89778c3
4 changed files with 12 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
driver/device_implicit_gemm_convolution_1_chwn_cyxk_khwn.hpp
View File

@@ -423,7 +423,7 @@ void device_implicit_gemm_convolution_1_chwn_cyxk_khwn(InDesc,
printf("Elapsed time : %f ms, %f TFlop/s\n",
time,
(float)calculate_convolution_flops(InDesc{}, WeiDesc{}, OutDesc{}) /
(std::size_t(1024) * 1024 * 1024 * 1024) / (time / 1000));
(std::size_t(1000) * 1000 * 1000) / time);
usleep(std::min(time * 1000, float(10000)));
}

2
driver/device_implicit_gemm_convolution_2_chwn_cyxk_khwn.hpp
View File

@@ -314,7 +314,7 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
printf("Elapsed time : %f ms, %f TFlop/s\n",
time,
(float)calculate_convolution_flops(InDesc{}, WeiDesc{}, OutDesc{}) /
(std::size_t(1024) * 1024 * 1024 * 1024) / (time / 1000));
(std::size_t(1000) * 1000 * 1000) / time);
usleep(std::min(time * 1000, float(10000)));
}

1
src/include/blockwise_batched_gemm.hip.hpp
View File

@@ -259,7 +259,6 @@ struct BlockwiseBatchGemmBlockABlockBThreadCTransANormalBNormalC_V2
c_thread_mtx,
False,
p_c_thread + ib * ThreadMatrixStrideC);
}
}
}

20
src/include/gridwise_convolution_implicit_gemm_v2_chwn_cyxk_khwn_lds_double_buffer.hip.hpp
View File

@@ -269,9 +269,9 @@ struct GridwiseConvolutionImplicitGemm_v2_chwn_cyxk_khwn_lds_double_buffer
#elif 0
blockwise_gemm.Run_asm
#endif
(p_wei_block_now + wei_cyxk_block_desc.Get1dIndex(0, y, x, 0),
p_in_block_now + y * Wi + x,
p_out_thread);
(p_wei_block_now + wei_cyxk_block_desc.Get1dIndex(0, y, x, 0),
p_in_block_now + y * Wi + x,
p_out_thread);
}
}
@@ -310,9 +310,9 @@ struct GridwiseConvolutionImplicitGemm_v2_chwn_cyxk_khwn_lds_double_buffer
#elif 0
blockwise_gemm.Run_asm
#endif
(p_wei_block_double + wei_cyxk_block_desc.Get1dIndex(0, y, x, 0),
p_in_block_double + y * Wi + x,
p_out_thread);
(p_wei_block_double + wei_cyxk_block_desc.Get1dIndex(0, y, x, 0),
p_in_block_double + y * Wi + x,
p_out_thread);
}
}
@@ -335,10 +335,10 @@ struct GridwiseConvolutionImplicitGemm_v2_chwn_cyxk_khwn_lds_double_buffer
#elif 0
blockwise_gemm.Run_asm
#endif
(p_wei_block_double + wei_block_space +
wei_cyxk_block_desc.Get1dIndex(0, y, x, 0),
p_in_block_double + in_block_space + y * Wi + x,
p_out_thread);
(p_wei_block_double + wei_block_space +
wei_cyxk_block_desc.Get1dIndex(0, y, x, 0),
p_in_block_double + in_block_space + y * Wi + x,
p_out_thread);
}
}
}