cuda-ab: ctaid_probe (CTA-coordinate + A-row reader) + GEO geometry print in transformer_mt

2026-10-18T19:03:17+00:00

The vGPU GR engine's failing chains are correct for the by=0 band and replicate one tile for rows >= 16 with the QMD geometry and the input recs proven clean, so the remaining hypothesis is engine-side CTA-coordinate behaviour. ctaid_probe reads it directly in two phases on the exact (16,4,1)/(16,16,1) geometry: Phase 1 coord_map writes by*10000+bx*100+ty*10+tx per element (int32 64x256) and reports per-band row0 plus a bad count - a collapsed or wrong CTAID.Y shows instantly. Phase 2 arow_read writes A[(by*16+ty)*256+(bx*16+tx)] with A[i]=i, so a correct engine reads each band's own rows; a collapsed row register reads one band's rows everywhere. transformer_mt prints its per-chain geometry (GEO lines) for the host-side GRID-WIT comparison. Runtime <10 s. Built with nvcc -O2 --cudart=static -arch=sm_75 (deterministic static runtime).

cuda-ab: transformer + transformer_mt clients (single-head attention + 4-layer FFN, fp32, CPU-reference checked; multi-stream and 5-launch loop variants)

2026-10-18T16:05:31+00:00

open-iov.git/tools/cuda-ab/run/transformer_mt, branch bigpath

cuda-ab: ctaid_probe (CTA-coordinate + A-row reader) + GEO geometry print in transformer_mt

cuda-ab: transformer + transformer_mt clients (single-head attention + 4-layer FFN, fp32, CPU-reference checked; multi-stream and 5-launch loop variants)