Profiling and Benchmarks - 2023.2 English

The General Similarity Kernel is validated on Alveo U50 board at 295MHz frequency. The hardware resource utilization and benchmark results are shown in the two table below.

Table 80 Table 1 Hardware resources
Name	LUT	Register	BRAM	URAM	DSP
generalSimilarityKernel	168516	230995	230.5	80	645

Table 81 Table 2 Performance comparison of sparse graph between TigerGraph on CPU and FPGA
Datasets	Vertex	Edges	Similarity Type	FPGA Time / ms	TigerGraph (32 core 512 GB)
Datasets	Vertex	Edges	Similarity Type	FPGA Time / ms	Time / ms	Speed up
as-Skitter	1694616	11094209	Cosine	21.3	278	13.5
as-Skitter	1694616	11094209	Jaccard	21.2	271	12.7
coPaperDBLP	540486	15245729	Cosine	24.4	289	11.8
coPaperDBLP	540486	15245729	Jaccard	24.5	271	11.0
coPaperCiteseer	434102	16036720	Cosine	30.6	282	9.2
coPaperCiteseer	434102	16036720	Jaccard	30.7	283	9.2
cit-Patents	3774768	16518948	Cosine	35.3	268	7.6
cit-Patents	3774768	16518948	Jaccard	35.1	262	7.5
europe_osm	50912018	54054660	Cosine	188.4	309	1.6
europe_osm	50912018	54054660	Jaccard	188.3	315	1.7
hollywood	1139905	57515616	Cosine	110.3	280	2.6
hollywood	1139905	57515616	Jaccard	110.1	329	3.0
soc-LiveJournal1	4847571	68993773	Cosine	105.2	293	2.8
soc-LiveJournal1	4847571	68993773	Jaccard	105.3	288	2.7
ljournal-2008	5363260	79023142	Cosine	120.6	281	2.4
ljournal-2008	5363260	79023142	Jaccard	120.5	384	3.2
GEOMEAN				79.4	292.7	3.70

Table 82 Table 2 Performance comparison of dense graph between TigerGraph on CPU and FPGA
Datasets	Vertex	Edges	Similarity Type	FPGA Time / ms	TigerGraph (32 core 512 GB)
					Time / ms	Speed up
Patients(1 GB)	1250000	200	Cosine	22.1	585.7	26.5

Note

1. Tigergraph running on platform with Intel(R) Xeon(R) CPU E5-2640 v3 @2.600GHz, 32 Threads (16 Core(s)).

2. Time unit: ms.