HBM Large Buffers¶
This is a simple example of vector addition to describe how HBM pseudo-channels can be grouped to handle buffers larger than 256 MB.
KEY CONCEPTS: High Bandwidth Memory, Multiple HBM Pseudo-channel Groups
This is a simple example of vector addition to describe how HBM pseudo-channels(PC) can be grouped to handle buffers larger than 256 MB.
Compute unit interfaces are associated to fixed HBM PC groups using
[connectivity] sp=krnl_vadd_1.in1:HBM[0:31] sp=krnl_vadd_1.in2:HBM[0:31] sp=krnl_vadd_1.out_r:HBM[0:31]
Here, All three buffers will be created into different bank groups and the application will run and performance will be reported. Following is the real log reported while running on U50 platform :
Found Platform Platform Name: Xilinx INFO: Reading ./build_dir.hw.xilinx_u50_gen3x16_xdma_201920_3/krnl_vadd.xclbin Loading: './build_dir.hw.xilinx_u50_gen3x16_xdma_201920_3/krnl_vadd.xclbin' Trying to program device: xilinx_u50_gen3x16_xdma_201920_3 Device: program successful! Running Three Separate Groups for Three Buffers Each buffer is allocated with different HBM Group. input 1 -> PC 0 : 31 input 2 -> PC 0 : 31 output -> PC 0 : 31 THROUGHPUT = 39.2318 GB/s TEST PASSED
Platforms containing following strings in their names are not supported for this example :
_u25_ u30 u200 zc vck u250 u280 aws-vu9p-f1 _u2_ nodma
Application code is located in the src directory. Accelerator binary files will be compiled to the xclbin directory. The xclbin directory is required by the Makefile and its contents will be filled during compilation. A listing of all the files in this example is shown below
Access these files in the github repo by clicking here.
COMMAND LINE ARGUMENTS¶
Once the environment has been configured, the application can be executed by
./hbm_large_buffers <krnl_vadd XCLBIN>