Split compute into stage1 (compute → partial) and stage2 (partial → c_data). Halves CARRY4 chain depth from 16 to ~8 levels. Compute cycles: 16 → 32.
- Rewrite main.c to read from ExtRAM (0x1c400000), compute via matmul IP, write results back. Matches testbench expectation of 10 groups. - Add in matmul_axi_slave to trace error writes