feat(amd): first-class R9700 (gfx1201/RDNA4) support for Qwen3.6-27B

[DeanoC]

Summary

Makes the AMD Radeon AI PRO R9700 (gfx1201, RDNA4 / Navi 48) a first-class AMD target for the DFlash Qwen3.6-27B stack, and fixes a CMake-version build break that affects the cuda12 / rocm Docker images.

1. R9700 / gfx1201 support

The HIP build pipeline compiled RDNA4 as gfx1200 (Navi 44, RX 9060-class) and mislabeled it "RX 9070". The R9700 is gfx1201, which is not code-object compatible with gfx1200 — so the published :rocm image and the documented build produced no native R9700 kernels (ggml-hip fails to load / falls back badly on the R9700).

• Add gfx1201 to the default fat-binary HIP arch list in docker-bake.hcl, Dockerfile.rocm, and the CI main/release matrix (.github/workflows/docker.yml). PR builds stay gfx1151-only for fast CI. Corrected the RDNA4 labels (gfx1200 = RX 9060, gfx1201 = RX 9070 / R9700).
• Build the rocWMMA flashprefill numerics test (test_flashprefill_kernels) under HIP too — it was CUDA-only, so there was previously no way to validate the rocWMMA path on AMD. The CUDA-spelled test compiles via the existing hip_compat/ shim.
• Document the R9700 in the README "Tested Machines" table and the server/README.md AMD HIP section (gfx1201 build, --ddtree-budget=22, multi-GPU / Fedora-PIE build notes).

No kernel changes: ROCm 7.1's rocWMMA handles the gfx12 WMMA operand-format change internally.

2. Build fix: json FetchContent on CMake < 3.24

#433 added DOWNLOAD_EXTRACT_TIMESTAMP TRUE to the json FetchContent_Declare. That keyword (and policy CMP0135) only exist on CMake ≥ 3.24, but the cuda12 / rocm Docker base images ship CMake 3.22, where the unknown token is parsed as extra URL list entries and configure fails with At least one entry of URL is a path (invalid in a list). The option is now applied only on CMake ≥ 3.24 (where it still silences the dev warning); older CMake parses the declare correctly again. Reproduced and verified on CMake 3.22.6 and 4.3.

Validation — real R9700 (gfx1201, ROCm 7.1.1)

Native gfx1201 build (Phase 1 + Phase 2 rocWMMA); the built libggml-hip.so contains native hipv4-amdgcn-amd-amdhsa--gfx1201 code objects.

Check	Result
test_server_unit	1984 assertions, 0 failures
test_flashprefill_kernels (rocWMMA)	PASS — max diff 5e-4; e2e flash_prefill_forward_bf16 @ S=8192 = 10.7 ms/iter
Qwen3.6-27B Q4_K_M + DFlash, --ddtree-budget=22	54.65 tok/s mean decode (bench_he.py --n-gen 256, AL 7.14, range 36.9–93.0 tok/s)
16K-context generation	coherent

Notes

• Fedora's system ROCm lives under /usr and links PIE by default; the local build used -DCMAKE_HIP_COMPILER_ROCM_ROOT=/usr -DROCM_PATH=/usr -DCMAKE_EXE_LINKER_FLAGS=-no-pie (documented in server/README.md). The Ubuntu-based Dockerfile.rocm is unaffected.

[cubic-dev-ai]

No issues found across 6 files

_{Re-trigger cubic}