-
feat: add gemm_n512_k2048 workloads and baseline solution
## Summary
Add 15 workloads (JSONL, inline tensors) for `gemm_n512_k2048` (4 files, 0.0 MB).
**Kernel**: Qwen3.5-35B-A3B GEMM (N=512, K=2048)
### Collection contents
- `definitions/gemm/gemm_n512_k2048.json` — kernel definition
- `workloads/gemm/gemm_n512_k2048.jsonl` — 15 workloads (JSONL, inline tensors)
- `solutions/baseline/gemm/gemm_n512_k2048/` — baseline solution
- `tests/references/test_gemm_n512_k2048.py` — reference test
### Companion PR
- Bench repo: flashinfer-ai/flashinfer-bench#309
## Test plan
- Baseline eval trace will be appended after this PR is created
- Reference test: `pytest tests/references/test_gemm_n512_k2048.py -v`
🤖 Generated with [Claude Code](https://claude.ai/code) -
initial commit
