Parallel Calculation of Local Scores in Bayesian Network Structure Learning using FPGA
n30-m9-P0
: a single-core software execution (SW)n30-m9-P1024
: a single-core software execution using an FPGA accelerator with parallelism 1024 (HW(P=1024))n30-m9-P2048
: a single-core software execution using an FPGA accelerator with parallelism 2048 (HW(P=2048))
Vitis 2020.2 + Alveo U50
- make new Application Project with template "Vector Addition"
File
\> New
\> Application Project
\> Next
\> Select a platform from repository (xilinx_u50_gen3x16_xdma_201920_3)
\> Application Project name (enter any names you want)
\> SW acceleration templates (Vector Addition)
\> Finish
-
add files
- replace
{Project_name}_system/{Project_name}_kernels/src/krnl_vadd.cpp
->{Project_name}_system/{Project_name}_kernels/src/localscore_kernel.cpp
- replace
{Project_name}_system/{Project_name}/src/vadd.cpp
- replace
{Project_name}_system/{Project_name}/src/vadd.h
- add
{Project_name}_system/{Project_name}/src/dataset/asia30.idt
- replace
-
Build HW (several hours)
-
Run
- Ryota Miyagi
- Graduate School of Informatics, Kyoto University
- Email: [email protected]