[sakuumomo]

Problem and result: One process could not use an NVIDIA GPU and an AMD GPU together. STRATA_HYBRID stays off. Turning it on keeps the CUDA host path and compiles the shared kernels again for HIP. Ordinals past the CUDA devices are HIP. The build is Linux x86-64 only. A CUDA-only build and a HIP-only build leave this path out.

A HIP device is named by its architecture, with compute capability left at 0 so it is not read as an NVIDIA architecture. Helpers are every other visible GPU. When hipBLASLt or rocBLAS is present, the HIP copies get the same switches as a pure HIP build, and the HIP prompt kernels use their own device table. When the current device is AMD, startup matches a pure HIP build: the probe accepts a non-gfx11 host, mapped expert fills go through one pinned buffer, the helper busy-waits for its stream, --vram-elastic stays off, and the tail skip, PCIe share, architecture defaults, gfx12 chunk, and pipeline windows follow that card. A HIP primary does not probe the canonical PCIe pack or say that the share will change.

Related discussion: no earlier issue agrees this scope.

Approach: The new files compile only with STRATA_HYBRID. Default builds keep their previous sources and flags. gfx906 is refused at configure. If CMake leaves a configuration's HIP flags empty, the init flags are written back. Two SHELL:-include generator expressions force-include the HIP headers, so CMake does not drop a repeated -include.

STRATA_HYBRID=OFF (default)
  CUDA build, or HIP build, or SYCL build, as before

STRATA_HYBRID=ON (Linux x86-64)
  CUDA host path
  shared kernels compiled again for HIP
  ordinals past the CUDA devices are HIP

Validation:
Passed: four compiles of 10fdcb31 on Linux x86-64. STRATA_BUILD_TESTS=OFF and STRATA_PORTABLE=ON on the Nix builds.
HIP, STRATA_HYBRID off: STRATA_ENABLE_HIP=ON, STRATA_ENABLE_CUDA=OFF, CMAKE_HIP_ARCHITECTURES=gfx1030, STRATA_PREFILL_MMQ=ON, HIP 7.2.3. CMake reported "Strata: HIP enabled, arch gfx1030" and did not report a hybrid build. Output /nix/store/087zqcrh6grkxlkqlmk5hfjm62flfaz5-strata-hip-0.2.1 (strata and strata-device). The compiler printed warnings while compiling for gfx1030. The link finished.
CUDA, STRATA_HYBRID off: STRATA_ENABLE_CUDA=ON, STRATA_ENABLE_HIP=OFF, CMAKE_CUDA_ARCHITECTURES=89-real, STRATA_PREFILL_MMQ=OFF, CUDA 12.9.86, gcc 14.4.0. CMake reported "Strata: CUDA enabled, arch 89-real" and did not report a hybrid build. Output /nix/store/3y0dc0nzihb4plvl8iaj087vw2b4clxm-strata-cuda-0.2.1 (strata and strata-device).
STRATA_HYBRID=ON: the same CUDA flags, plus HIP gfx1150. CMake reported CUDA arch 89-real and "Strata hybrid: gfx1150". The log shows CUDA objects and HIP objects (strata_hip_objs, compiled for gfx1150) in one build. Output /nix/store/h0v2fqhpi910zfx8c148gl828aw12y81-strata-hybrid-0.2.1 (strata and strata-device). The compiler printed warnings while compiling for gfx1150. The link finished.
SYCL: the strata-sycl-dev image from sycl/tools/Dockerfile (oneAPI 2026.1.0, no AOT device). sycl/tools/build.sh strata printed BUILD EXIT 0 and errors: 0. Binary /tmp/strata-wt-hybrid/build-sycl/strata.
The author ran it on two Linux machines and reports that both function as intended. The laptop has an RTX 4060 and a Radeon 890M and serves with STRATA_HYBRID. The desktop has an RX 6800 and an RX 6900 XT and serves with the HIP build, so STRATA_HYBRID stays off there. No command log or timing was attached to that report.
Failed: none observed.
Not run: a SYCL process on a device. A saved log of the laptop startup.

Limits: Linux x86-64 only. A HIP primary refuses --vram-elastic. prime_hw_queues stays on the STRATA_USE_HIP path. Data-type enums are not remapped between the two runtimes. The first host call on an AMD-only process can still enter the CUDA copy before a trampoline is installed. Compute capability 0 sends host code that branches on compute capability down the portable path while a HIP device is current.

Changes since the last update (2026-10-27)

• Rebased onto upstream main (0.2.6, 4b9e05a2); content re-landed verbatim. New tip b32dee9c — seven commits: the opt-in hybrid build, its 0.2.2/0.2.3 HIP-rule extensions, the hip_compat fixes main's drafter required, and the dsv41 glob guard. The commit hashes in the Validation section belong to the pre-rebase lineage.
• Equivalent content (the same rebased delta on the owner's deploy branch nixos, pin 3032afd8) has served daily on both machines since the rebase: the laptop as the CUDA+HIP hybrid build (RTX 4060 + Radeon 890M), the desktop as a HIP-only build where STRATA_HYBRID stays off (RX 6800 + RX 6900 XT).
• Serve-side follow-up PRs are now open in the same stack: [Other] serve: session ladder — conversations survive an engine unload #4306 (session ladder) and [Other] serve+webui: unload that works — pre-staged checkpoints, polite waits, honest answers #4307 (unload that works). Neither touches this diff.

Changes since the last update (2026-10-30)

• Rebased onto upstream v0.2.8 (870aedd9); content re-landed as before. The single conflict - the expert-cache auto message against upstream [Engine] v0.2.7.1 / engine 0.1.53: Excessive memory usage causes VRAM exhaustion and frequent request overflows #4474's late-priced note - resolved in the rebase. New tip 5fa28548, eight commits: the seven above plus the moe_ws glob exclusion (upstream [Engine] GP100 prompt path: a fully resident layer's experts weight-stationary, one launch per product (STRATA_MMQ_WS=1, opt-in) #3752's GP100 weight-stationary expert kernels are pure PTX: the HIP lane can neither compile nor dispatch them, and the hybrid glob previously force-compiled them for gfx1150).
• The owner's rebuilt deploy lineage (branch nixos: every open PR branch of this stack re-merged in order onto 870aedd9) passes every owner gate: HIP-only (gfx1030) and CUDA+HIP hybrid (gfx1150) flavor builds green with installCheck, and the session-file CPU and ThreadSanitizer suites green on the rebuilt lineage.

Branch commits were re-signed (GPG ed25519) on 2026-10-30; trees are unchanged from the validated work.

Dependencies: none — #3553 stacks on this branch.

[Niko1221]

Thank you for the careful work on a combined CUDA and HIP build. It is planned for a later release, with no date yet, and this will stay open until then.