TPG Test Application - Binary File Format

Overview

This document specifies the binary file format used by the TPG test application.

1. Binary File Format (.bin/.val)

Binary file with header and int16 data.

1.1 File Structure

┌─────────────────┐
│ File Header     │ (12 bytes)
├─────────────────┤
│ Data Block 0    │ (N samples × 2 bytes each)
├─────────────────┤
│ Data Block 1    │ (N samples × 2 bytes each)
├─────────────────┤
│ Data Block 2    │ (N samples × 2 bytes each)
├─────────────────┤
│ ...             │
└─────────────────┘

This binary file format can be used for:

• Input data files (.bin): Time step data for processor testing
• Validation files (.bin): Expected results for validation steps
• Pipeline tests: Multiple time step payloads for different channels

1.2 Header Specification

Offset	Size (bytes)	Type	Name	Description
0	4	uint32	magic_number	Magic number: 0x54504754 ("TPGT")
4	4	uint32	version	Format version: 1.0.4 (matches tpglibs)
8	4	uint32	reserved	Reserved for future use

1.3 Data Format

• Data Type: int16 (2 bytes per sample)
• Byte Order: Little-endian
• Layout: One data block per read operation
• Header: 16 bytes with magic number and version

1.4 Usage with BinarySignalReader

samples_per_block is a logical abstraction for the size of data to be fed into a processing entity. If it is a single processor, then it is 16 integer type data for the 16 channels. If it is a TPGenerator data block, then it would be 4 pipelines * 16 channels.

BinarySignalReader<int16_t> reader("data.bin");
 
// Read one data block at a time
while (!reader.eof()) {
    auto data_block = reader.next(samples_per_block);
    if (data_block.empty()) break;
    
    // Process this data block
    auto result = processor->process(data_block);
}

1.5 Header Validation

struct BinaryFileHeader {
    uint32_t magic_number;
    uint32_t version;
    uint32_t reserved;
};
 
// Validate header when opening file
BinaryFileHeader header;
file.read(reinterpret_cast<char*>(&header), sizeof(BinaryFileHeader));
 
if (header.magic_number != 0x54504754) {
    throw std::runtime_error("Invalid binary file format");
}
if (header.version != 0x010004) {  // 1.0.4 in hex
    throw std::runtime_error("Unsupported file version");
}

2. Configuration File Format (.json)

Simple JSON configuration file.

2.1 Example Configuration

{
  "processor_config": {
    "accum_limit": 10,
    "metric_collect_toggle_state": false
  },
  "test_config": {
    "processor_name": "AVXFrugalPedestalSubtractProcessor",
    "max_steps": 250000,
    "samples_per_time_step": 16,
    "validation_steps": [1, 200000]
  }
}

2.2 Configuration Fields

• processor_config: Processor-specific parameters (varies by processor type)
• test_config.processor_name: String specifying the processor name to use for the test. This must correspond to a known processor class name (e.g., "AVXFrugalPedestalSubtractProcessor")
• test_config.max_steps: Integer specifying the maximum number of processing steps/time steps to run processing (even if no validation steps are being performed). If not provided, the maximum step will be inferred from the largest value in validation_steps.
• test_config.samples_per_time_step: Number of samples per time step (should in most cases be 16 for a single processor)
• test_config.validation_steps: Array of processing step numbers to perform validation (exact integer comparison)

3. Test File Examples

3.1 AVXFrugalPedestalSubtractProcessor Sanity Test

Files:

• tpg_processor_avx_fps_sanity_test.bin - Input file with constant 0 values
• tpg_processor_avx_fps_sanity_val.bin - Validation file with expected outputs
• tpg_processor_avx_fps_sanity_config.json - Configuration file

Test Logic:

• Input: Constant 0 values for all channels across all time steps
• Step 1: Baseline validation - output should be -16384 (0 - initial_pedestal)
• Step 200000: Convergence validation - output should be 0 (0 - converged_pedestal)

Configuration:

{
  "processor_config": {
    "accum_limit": 10,
    "metric_collect_toggle_state": false
  },
  "test_config": {
    "samples_per_time_step": 16,
    "validation_steps": [1, 200000]
  }
}

4. General Test File Naming Convention

Pattern: tpg_processor_[processor_name]_[test_name]_[test/val/config]

Examples:

• tpg_processor_avx_fps_sanity_test.bin - AVX Frugal Pedestal Subtract sanity test input
• tpg_processor_avx_fps_sanity_val.val - AVX Frugal Pedestal Subtract sanity test validation
• tpg_processor_avx_fps_sanity_config.json - AVX Frugal Pedestal Subtract sanity test config