Understanding estimate_layer_config_alternatives.json

[kalahel]

Hello,

Can anyone explain the estimate_layer_config_alternatives.json, how it is generated and how to use it fully ?

I only found this on the documentation :

report/estimate_layer_config_alternatives.json – resources per layer estimation from analytical model, including what other config alternatives would have yielded

But it is not clear what are the other config nor how can we input different config.

For example, for the MobileNetV1, I have this kind of alternative configs :

  "ConvolutionInputGenerator_rtl_0": [
    {
      "BRAM_18K": 1,
      "BRAM_efficiency": 1,
      "LUT": 300,
      "URAM": 0,
      "URAM_efficiency": 1,
      "DSP": 0
    },
    {
      "BRAM_18K": 0,
      "BRAM_efficiency": 1,
      "LUT": 724,
      "URAM": 0,
      "URAM_efficiency": 1,
      "DSP": 0
    },
    {
      "BRAM_18K": 0,
      "BRAM_efficiency": 1,
      "LUT": 300,
      "URAM": 1,
      "URAM_efficiency": 1,
      "DSP": 0
    }
  ],
 [...]
"MVAU_rtl_2": [
    {
      "BRAM_18K": 15,
      "BRAM_efficiency": 0.11851851851851852,
      "LUT": 0,
      "URAM": 0,
      "URAM_efficiency": 1,
      "DSP": 32
    },
    {
      "BRAM_18K": 15,
      "BRAM_efficiency": 0.11851851851851852,
      "LUT": 0,
      "URAM": 0,
      "URAM_efficiency": 1,
      "DSP": 32
    }
  ],