Profiling - 2023.2 English

Vitis Libraries
Release Date
2023-12-20
Version
2023.2 English

The hardware resource utilizations are listed in the following table. Different tool versions may result slightly different resource.

Table 23 Table 1 Hardware resources for webp kernels
Kernel BRAM URAM DSP FF LUT Frequency(MHz)
kernel1 72 10 410 56498 48301 250
kernel2 11 0 5 23073 16375 250
  • One instance achieves about 6~14 times acceleration. Here are some examples:
Table 24 Table 2 Performance of Webp Encoder for FPGA
Kernel Width (pix) Height (pix) -q latency (ms) Throughput FPGA B (Mb/s) Throughput FPGA P (Mp/s) FPs (fps)
Kernel1 1920 1080 80 21.18 146.83 97.88 47.20
Kernel2 1920 1080 80 14.57 213.54 142.36 68.65
Kernel1 512 512 80 3.22 122.03 81.35 310.33
Kernel2 512 512 80 2.92 134.65 89.77 342.43
Kernel1 1920 1080 90 21.03 147.87 98.58 47.54
Kernel2 1920 1080 90 15.92 195.43 130.29 62.83
Kernel1 512 512 90 4.73 83.12 55.41 211.39
Kernel2 512 512 90 4.93 79.73 53.16 202.78
  • Platform: FPGA U200, CPU details are listd belowd (single thread)

Note

1. Kernels running on platform with Intel(R) Xeon(R) CPU E5-2603 v3 @ 1.60GHz, 48 Threads.
2. time unit: ms.
3. “-” Indicates that the result could not be obtained due to insufficient memory.
4. FPGA time is the kernel runtime by adding data transfer and executed with webp encoder.