gpuSimulateN2kCorr¶

class gpuSimulateN2kCorr : public kotekan::Stage¶

Perform on CPU the equivalent of the CudaCorrelator stage: N2 correlation.

An example of this stage being used can be found in config/tests/verify_cuda_n2k.yaml.

voltage_in_buf The input voltages.
- Format: 4+4-bit complex int, encoded with CHIME convention: offset encoded by +8, imaginary in low bits, real in high. In kotekan this is a int4x2_swapped_withoffset.
- Shape: [samples_per_data_set, num_local_freq, num_elements] or equivalently [samples_per_data_set, num_local_freq, num_polarizations, num_dishes]
- Metadata: chordMetadata
rfimask_in_buf The input fast cadence RFI mask.
- Format: bitmask in (eqivalently) uint8_t or uint1x8_t.
- Shape: [samples_per_data_set / 1024, num_local_freq, 128]. The time axis is split into a 1024 sample (128 byte) fast index and a coarse samples_per_data_set / 1024 index. The frame size is samples_per_data_set * num_local_freq / 8 bytes.
- Metadata: chordMetadata
corr_out_buf The output correlation matrix, tiled into 16x16 element blocks, storing the lower triangular blocks (e.g. B00, B10, B11, B20, …). There are num_blocks = num_elements / 16 blocks, and num_integrations = samples_per_data_set / sub_integration_ntime times represented in the output.
- Format: int32+32 complex. Each 2 int entry has the real part in the first int and the imaginary part in the second int.
- Shape: [num_integrations, num_local_freq, num_blocks, 16, 16, 2]. The final axis is over real/imaginary components. The size is num_integrations * num_local_freq * num_blocks * 16^2 * 2 int32s.

Param num_elements:: Int. Number of feeds or (antennas x polarizations).
Param num_local_freq:: Int. Number of frequencies.
Param samples_per_data_set:: Int. Number of samples per frame.
Param sub_integration_ntime:: Int. Number of samples to sum over for each N^2 correlation matrix.

Public Functions

gpuSimulateN2kCorr(kotekan::Config &config, const std::string &unique_name, kotekan::bufferContainer &buffer_container)¶