SPRUIL1D May 2019 – December 2024 DRA829J , DRA829J-Q1 , DRA829V , DRA829V-Q1 , TDA4VM , TDA4VM-Q1
VPAC_MSC consists of 10 programmable resizers performing multi-thread/multi-scaling operations (1-input to N-outputs and 1-input to M-outputs where N+M is 10 or less). Each processing thread of MSC HWA reads its input plane data from a Shared Memory buffers (SL2) circular line buffers, performs multi-scaling operations (ratios between X and 0.25X) on the selected thread channel input, and writes out results to SL2 circular line buffers. In case of OTF operation, the source data is generated from another HWA. In case of M2M operation, the source data is read from the DDR memory. Data transfers from/to SL2 to/from external memory (or another HWA) are handled by VPAC level DMA controller with transfer request events coming from VPAC top level HWA Thread Scheduler (HTS), see HWA Threads Scheduler (HTS).
A typical configuration of the MSC module uses one input plane data to generate 7 intra-octave scales and the next octave image. Multi-pass processing is then utilized to generate up to next 6 or 7 sets of scales/octaves. Since the resolution after each octave is essentially reduced by 4x, the performance requirement for generating an infinite number of scales is bounded by 1.33x of the base input image pixel rate. The second input can be used to enable a pyramid generation processing for another input data or a set of scales/pyramid generation for chroma data as long as the total number of scales needed by both threads is 10 or less. Each resizer in a processing thread can access any coefficient set.
For details on MSC integration in VPAC, see Section 6.9.3,VPAC Subsystem Level.