Hash Join V2

Hash Join V2 resides in L1/benchmarks/hash_join_v2 directory. This benchmark tests the performance of hashJoinMPU from hash_join_v2.hpp with the following query.

SELECT
       SUM(l_extendedprice * (1 - l_discount)) as revenue
FROM
       Orders,
       Lineitem
WHERE
       l_orderkey = o_orderkey
;
Here, Orders is a self-made table filled with random data,which contains a column named o_orderkey.
Lineitem is also a table, which contains 3 columns named l_orderkey, l_extendedprice and l_discount.

Dataset

This project uses 32-bit data for numeric fields. To benchmark 64-bit performance, edit host/table_dt.h and make TPCH_INT an int64_t.

Executable Usage

  • Work Directory(Step 1)

The steps for library download and environment setup can be found in Vitis Database Library. For getting the design,

cd L1/benchmarks/hash_join_v2
  • Build kernel(Step 2)

Run the following make command to build your XCLBIN and host binary targeting a specific device. Please be noticed that this process will take a long time, maybe couple of hours.

make run TARGET=hw DEVICE=xilinx_u280_xdma_201920_3 HOST_ARCH=x86
  • Run kernel(Step 3)

To get the benchmark results, please run the following command.

./build_dir.hw.xilinx_u280_xdma_201920_3/test_join.exe -xclbin build_dir.hw.xilinx_u280_xdma_201920_3/hash_join.xclbin

Hash Join V2 Input Arguments:

Usage: test_join.exe -xclbin
       -xclbin:      the kernel name

Note: Default arguments are set in Makefile, you can use other platforms to build and run.

  • Example output(Step 4)
------------- Hash-Join Test ----------------
Data integer width is 32.
Host map buffer has been allocated.
Lineitem 6001215 rows
Orders 227597rows
Lineitem table has been read from disk
Orders table has been read from disk
INFO: CPU ref matched 13659906 rows, sum = 6446182945969752
Found Platform
Platform Name: Xilinx
Selected Device xilinx_u280_xdma_201920_3
INFO: Importing build_dir.hw.xilinx_u280_xdma_201920_3/hash_join_v2.xclbin
Loading: 'build_dir.hw.xilinx_u280_xdma_201920_3/hash_join_v2.xclbin'
Kernel has been created
DDR buffers have been mapped/copy-and-mapped
FPGA result 0: 644618294596.9752
Golden result 0: 644618294596.9752
FPGA execution time of 1 runs: 63985 usec
Average execution per run: 63985 usec
INFO: kernel 0: execution time 55945 usec
---------------------------------------------

Profiling

The hash join v2 design is validated on Alveo U280 board at 282 MHz frequency. The hardware resource utilizations are listed in the following table.

Table 1 Hardware resources for hash join v2
Name LUT BRAM URAM DSP
Platform 122083 202 0 4
join_kernel 63697 98 64 3
User Budget 1180637 1814 960 9020
Percentage 5.40% 5.40% 6.67% 0.03%
The performance is shown below.
In above test, table Lineitem has 3 columns and 6001215 rows and Orders does 1 column and 227597 rows. This means that the design takes 55.5945ms to process 69.55MB data, so it achieves 1.21GB/s throughput.