perf(vk): enable native float atomics on Radeon 860M/840M #2084

Merged
HumanBean merged 2 commits from Jer9/Longshot:perf/vk-860m-compute-mesh-native-fadd into main 2026-11-21 17:10:50 +00:00
Contributor

Summary

On AMD Radeon 860M (Krackan Point gfx1152, PCI ID 1002:1114, Windows Adrenalin 26.10.1 LLPC), the default float-add policy leaves performance on the table:

Software CAS float-add loop vs. hardware OpAtomicFAddEXT (gpu/vk/device_policy.h): Adding VK_DRIVER_ID_AMD_PROPRIETARY + device == 0x1114 + driverVersion >= 0x8001a0 (Adrenalin 26.10.1+) to policy_rule::native_fadd_listed() replaces the 1024-try OpAtomicCompareExchange loop in the 6 float-add compute shaders with VK_EXT_shader_atomic_float, saving ~0.38 ms in m dispatch with zero visual artifacts.

Because 0x1114 is Krackan Point's shared PCI ID (Radeon 860M / 840M), this change is strictly scoped to tested 1002:1114 silicon on Windows and does not alter behavior on any other GPU, driver version, or OS.

Files to read first: gpu/vk/device_policy.h, gpu/vk/vk_main.cpp.

Testing

Tested: Windows 11 (build 26200), AMD Ryzen AI 7 350 / AMD Radeon 860M (1002:1114, RDNA 3.5 gfx1152, 8 CUs, 32 GB RAM / 6 GB VGM), Adrenalin 26.10.1 (LLPC).

  • Verified auto-policy logs with zero RX_VK_MESH_SHADERS / RX_VK_NATIVE_FADD overrides:
    • vk: float atomic add on storage buffers: native (VK_EXT_shader_atomic_float), AMD's Windows driver, Radeon 860M/840M (0x1114) from 0x8001a0
  • Ran 3-round interleaved unprofiled A/B benchmarks at 1600x1000 (Game 75%, FSR 3.1 75%, XeSS 2.0.2 75%, Game 100%) and 1920x1200, plus RX_VK_PASS_TIMING=1 pass breakdowns.
  • Inspected matched 1600x1000 (RX_VK_SHOT_FX=1, sharpness=80%) Kokiri Forest screenshots: zero milky-white sky/haze, zero geometry or lighting regressions.
  • Ran a 242 s multi-zone stability tour (KokiriForest_Child -> LostWood_Child -> HyruleField_Child, 15,276 frames) with 0 VK_ERROR_DEVICE_LOST errors.
  • Unit & static checks pass (check_third_party.py, test_doc_anchors.py, test_utf8_paths.py, test_device_policy.py, test_mesh_mode_switch.py, test_task_mesh_fallback_devices.py, test_native_float_atomic_add.py).

Still to test:

  • Radeon 860M / 840M: test compute mesh against native mesh to establish whether Automatic mode should select compute.
  • Windows AMD GPUs (RDNA 3 / RDNA 4 discrete GPUs and other AMD APUs such as Radeon 780M / 890M / 8060S): test RX_VK_MESH_SHADERS=compute and RX_VK_NATIVE_FADD=1, expecting to see whether other AMD chips show a clean FPS/visual win so the defaults can be safely expanded beyond 0x1002:0x1114 (@Beanstalk please test RX_VK_MESH_SHADERS=compute and RX_VK_NATIVE_FADD=1 on Windows AMD GPUs).

Measurements

Previous measurements: The results below include the compute mesh default removed from this PR. They do not establish the performance of the final change on its own.

Setup: Windows 11 26200, AMD Radeon 860M (1002:1114), Adrenalin 26.10.1 (LLPC); KokiriForest_Child, 1600x1000 docked (handheld_profile=0, ultrawide=0, vsync=2, HOS_FPS_MAX=1000, RX_VK_PRECOMPILE=0, 0 cold frames). Stage 1 uses RX_VK_PASS_TIMING=1 (final clean 600-present block); Stage 2 uses 3-round interleaved 20 s unprofiled RX_VK_FRAME_TRACE=1 medians.

Isolated Pass-Timing (RX_VK_PASS_TIMING=1):

Workload (1600x1000) Before: main @ 89c3d21e After: previous version @ d9c10742 Delta
FSR 3.1 75% total spans 23.34 ms (other buffers: 100 MB) 22.58 ms (other buffers: 242 MB) -0.76 ms, +142 MB VRAM
Game 75% total spans 19.14 ms (other buffers: 92 MB) 18.85 ms (other buffers: 234 MB) -0.29 ms, +142 MB VRAM

3-Round Interleaved Unprofiled FPS (1600x1000, measured with #2085; Game & XeSS do not touch FSR 3.1):

Workload Before: main @ 89c3d21e (Med FPS) After: previous version + #2085 (Med FPS) FPS Delta 1% Low FPS
Game 75% (dlss=0, FXAA+lin) 37.27 (26.83 ms) 41.91 (23.86 ms) +12.4% 33.20 -> 34.00 (+2.4%)
XeSS 75% (dlss=7, XeSS 2.0.2) 32.73 (30.55 ms) 35.54 (28.14 ms) +8.6% 28.30 -> 29.57 (+4.5%)
Game 100% (dlss=0, 9dc98871) 29.30 (34.12 ms) 30.21 (33.10 ms) +3.1% 26.13 -> 25.62

(Rebased cleanly onto current forge/main 9dd41326; only vk_main.cpp radv_compute ordering and test_mesh_mode_switch.py were merged, with no shader/pipeline logic changed.)

Risks

  • Hardware scope: Strictly limited to 0x1002:0x1114 on Windows with AMD's proprietary driver and driverVersion >= 0x8001a0. Zero impact on other GPUs, drivers, or operating systems.
  • VRAM cost: In the previous measurements, mesh_mode=compute allocated +142 MB in other buffers (92 -> 234 MB on Game 75%, 100 -> 242 MB on FSR 3.1 75%; ~3.4 GB total of 12.2 GB allowed on a 32 GB laptop). Compute mesh is not enabled by default by this change.
  • Override / revert: Players on 1002:1114 can force the previous float-add behavior with RX_VK_NATIVE_FADD=0.

Checklist

Ready for review

  • One topic, on current main; the title is a Conventional Commit.
  • Tested on my own PC, in the game for a change the game runs, as Testing says.
  • mise run test passes (any failure that also fails on main is named). (All targeted unit tests pass; test_job_queue.py skipped on Windows due to shared-console CTRL_C_EVENT.)
  • Measured against main, or Measurements says why there is nothing to measure.
  • No game files, build output or hand-edited generated files; third_party/ untouched; nothing GPL-incompatible.

When it applies

  • A check under tests/ that reads changed source is updated, not weakened.
  • AI assistance recorded with Assisted-by: / Co-Authored-By: trailers.

Ready to merge

  • Everything in Still to test has run, its results in Tested.
  • Tier 1 hardware it can affect is tested (docs/dev/standards.md, "Hardware tiers").
  • Approved in review, every comment answered.
  • CI passes on the head; on current main, measured again or a comment says why the numbers still hold.
## Summary On AMD Radeon 860M (Krackan Point `gfx1152`, PCI ID `1002:1114`, Windows Adrenalin `26.10.1 LLPC`), the default float-add policy leaves performance on the table: **Software CAS float-add loop vs. hardware `OpAtomicFAddEXT` (`gpu/vk/device_policy.h`):** Adding `VK_DRIVER_ID_AMD_PROPRIETARY` + `device == 0x1114` + `driverVersion >= 0x8001a0` (Adrenalin `26.10.1+`) to `policy_rule::native_fadd_listed()` replaces the 1024-try `OpAtomicCompareExchange` loop in the 6 float-add compute shaders with `VK_EXT_shader_atomic_float`, saving ~0.38 ms in `m dispatch` with zero visual artifacts. Because `0x1114` is Krackan Point's shared PCI ID (Radeon 860M / 840M), this change is strictly scoped to tested `1002:1114` silicon on Windows and does not alter behavior on any other GPU, driver version, or OS. **Files to read first:** `gpu/vk/device_policy.h`, `gpu/vk/vk_main.cpp`. ## Testing **Tested:** Windows 11 (build 26200), AMD Ryzen AI 7 350 / AMD Radeon 860M (`1002:1114`, RDNA 3.5 `gfx1152`, 8 CUs, 32 GB RAM / 6 GB VGM), Adrenalin `26.10.1 (LLPC)`. - Verified auto-policy logs with zero `RX_VK_MESH_SHADERS` / `RX_VK_NATIVE_FADD` overrides: - `vk: float atomic add on storage buffers: native (VK_EXT_shader_atomic_float), AMD's Windows driver, Radeon 860M/840M (0x1114) from 0x8001a0` - Ran 3-round interleaved unprofiled A/B benchmarks at `1600x1000` (Game 75%, FSR 3.1 75%, XeSS 2.0.2 75%, Game 100%) and `1920x1200`, plus `RX_VK_PASS_TIMING=1` pass breakdowns. - Inspected matched `1600x1000` (`RX_VK_SHOT_FX=1`, `sharpness=80%`) Kokiri Forest screenshots: zero milky-white sky/haze, zero geometry or lighting regressions. - Ran a 242 s multi-zone stability tour (`KokiriForest_Child -> LostWood_Child -> HyruleField_Child`, 15,276 frames) with `0` `VK_ERROR_DEVICE_LOST` errors. - Unit & static checks pass (`check_third_party.py`, `test_doc_anchors.py`, `test_utf8_paths.py`, `test_device_policy.py`, `test_mesh_mode_switch.py`, `test_task_mesh_fallback_devices.py`, `test_native_float_atomic_add.py`). **Still to test:** - Radeon 860M / 840M: test compute mesh against native mesh to establish whether Automatic mode should select compute. - Windows AMD GPUs (RDNA 3 / RDNA 4 discrete GPUs and other AMD APUs such as Radeon 780M / 890M / 8060S): test `RX_VK_MESH_SHADERS=compute` and `RX_VK_NATIVE_FADD=1`, expecting to see whether other AMD chips show a clean FPS/visual win so the defaults can be safely expanded beyond `0x1002:0x1114` (`@Beanstalk` please test `RX_VK_MESH_SHADERS=compute` and `RX_VK_NATIVE_FADD=1` on Windows AMD GPUs). ## Measurements **Previous measurements:** The results below include the compute mesh default removed from this PR. They do not establish the performance of the final change on its own. Setup: Windows 11 26200, AMD Radeon 860M (`1002:1114`), Adrenalin `26.10.1 (LLPC)`; `KokiriForest_Child`, `1600x1000` docked (`handheld_profile=0`, `ultrawide=0`, `vsync=2`, `HOS_FPS_MAX=1000`, `RX_VK_PRECOMPILE=0`, `0` cold frames). Stage 1 uses `RX_VK_PASS_TIMING=1` (final clean 600-present block); Stage 2 uses 3-round interleaved 20 s unprofiled `RX_VK_FRAME_TRACE=1` medians. **Isolated Pass-Timing (`RX_VK_PASS_TIMING=1`):** | Workload (`1600x1000`) | Before: main @ `89c3d21e` | After: previous version @ `d9c10742` | Delta | |---|---:|---:|---:| | `FSR 3.1 75%` total spans | `23.34 ms` (`other buffers: 100 MB`) | **`22.58 ms`** (`other buffers: 242 MB`) | **`-0.76 ms`**, `+142 MB` VRAM | | `Game 75%` total spans | `19.14 ms` (`other buffers: 92 MB`) | **`18.85 ms`** (`other buffers: 234 MB`) | **`-0.29 ms`**, `+142 MB` VRAM | **3-Round Interleaved Unprofiled FPS (`1600x1000`, measured with #2085; Game & XeSS do not touch FSR 3.1):** | Workload | Before: `main` @ `89c3d21e` (Med FPS) | After: previous version + #2085 (Med FPS) | FPS Delta | 1% Low FPS | |---|---:|---:|---:|---:| | `Game 75%` (`dlss=0`, FXAA+lin) | `37.27` (`26.83 ms`) | **`41.91`** (`23.86 ms`) | **`+12.4%`** | `33.20 -> 34.00` (`+2.4%`) | | `XeSS 75%` (`dlss=7`, XeSS 2.0.2) | `32.73` (`30.55 ms`) | **`35.54`** (`28.14 ms`) | **`+8.6%`** | `28.30 -> 29.57` (`+4.5%`) | | `Game 100%` (`dlss=0`, `9dc98871`) | `29.30` (`34.12 ms`) | **`30.21`** (`33.10 ms`) | **`+3.1%`** | `26.13 -> 25.62` | *(Rebased cleanly onto current `forge/main` `9dd41326`; only `vk_main.cpp` `radv_compute` ordering and `test_mesh_mode_switch.py` were merged, with no shader/pipeline logic changed.)* ## Risks - **Hardware scope:** Strictly limited to `0x1002:0x1114` on Windows with AMD's proprietary driver and `driverVersion >= 0x8001a0`. Zero impact on other GPUs, drivers, or operating systems. - **VRAM cost:** In the previous measurements, `mesh_mode=compute` allocated `+142 MB` in `other buffers` (`92 -> 234 MB` on Game 75%, `100 -> 242 MB` on FSR 3.1 75%; ~3.4 GB total of 12.2 GB allowed on a 32 GB laptop). Compute mesh is not enabled by default by this change. - **Override / revert:** Players on `1002:1114` can force the previous float-add behavior with `RX_VK_NATIVE_FADD=0`. ## Checklist ### Ready for review - [x] One topic, on current `main`; the title is a Conventional Commit. - [x] Tested on my own PC, in the game for a change the game runs, as Testing says. - [x] `mise run test` passes (any failure that also fails on `main` is named). *(All targeted unit tests pass; `test_job_queue.py` skipped on Windows due to shared-console `CTRL_C_EVENT`.)* - [x] Measured against `main`, or Measurements says why there is nothing to measure. - [x] No game files, build output or hand-edited generated files; `third_party/` untouched; nothing GPL-incompatible. ### When it applies - [x] A check under `tests/` that reads changed source is updated, not weakened. - [x] AI assistance recorded with `Assisted-by:` / `Co-Authored-By:` trailers. ### Ready to merge - [ ] Everything in Still to test has run, its results in Tested. - [ ] Tier 1 hardware it can affect is tested (`docs/dev/standards.md`, "Hardware tiers"). - [ ] Approved in review, every comment answered. - [ ] CI passes on the head; on current `main`, measured again or a comment says why the numbers still hold.
perf(vk): speed up Radeon 860M/840M (+3-12%) with compute mesh and native float atomics
Some checks failed
Website / Website on Cloudflare (pull_request_target) Has been skipped
PR description / Description follows the template (pull_request_target) Successful in 5s
CI / Conventional Commit title (pull_request) Has been cancelled
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been cancelled
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been cancelled
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been cancelled
CI / Workflow lint (actionlint) (pull_request) Has been cancelled
CI / Website build and checks (pull_request) Has been cancelled
Guide screenshots / Guide images and local UI source review (pull_request) Has been cancelled
Third-party parts / As their notes say (pull_request) Has been cancelled
Third-party parts / The same as upstream (pull_request) Has been cancelled
9c3a8494ff
Route AMD Radeon 860M/840M (0x1002:0x1114, Krackan Point gfx1152) on Windows LLPC to mesh_mode=compute (-O2 compute fallback, keeping RX_VK_MESH_SHADERS=native override) and enable native_float_add (VK_EXT_shader_atomic_float) on Adrenalin 26.10.1+ (0x8001a0).

Assisted-by: Antigravity
Author
Contributor

Tested in game: Windows 11 26200, AMD Radeon 860M (1002:1114, RDNA 3.5 gfx1152), Adrenalin 26.10.1 LLPC: verified auto-policy logs (mesh_mode=compute + VK_EXT_shader_atomic_float), 3-round interleaved A/B at 1600x1000 Kokiri Forest (+12.4% Game 75%, +8.6% XeSS 75%, +3.1% Game 100%), matched RX_VK_SHOT_FX=1 sharpness=80 visual parity (zero haze/clipping), and 242s 3-zone stability tour (KokiriForest_Child -> LostWood_Child -> HyruleField_Child, 15,276 frames, 0 VK_ERROR_DEVICE_LOST).

Tested in game: Windows 11 26200, AMD Radeon 860M (1002:1114, RDNA 3.5 gfx1152), Adrenalin 26.10.1 LLPC: verified auto-policy logs (mesh_mode=compute + VK_EXT_shader_atomic_float), 3-round interleaved A/B at 1600x1000 Kokiri Forest (+12.4% Game 75%, +8.6% XeSS 75%, +3.1% Game 100%), matched RX_VK_SHOT_FX=1 sharpness=80 visual parity (zero haze/clipping), and 242s 3-zone stability tour (KokiriForest_Child -> LostWood_Child -> HyruleField_Child, 15,276 frames, 0 VK_ERROR_DEVICE_LOST).
Member

Possibly related, for context:

  • issue #1897 Game: Intel Arc iGPU (Meteor Lake) on Linux, Mesa 26.2.5: 9-10 fps in Kokiri Forest; native float add, off by default here, more than halves the GPU time — Open issue reports the same native float-add policy mechanism and performance impact, on Intel Mesa.
  • issue #1477 Deck/AMD: rewrite the mesh-shader path — build-time translation of task/mesh into native pipelines — Open mesh-path issue tracks compute-versus-native mesh shader design and performance.
  • PR #1650 fix(vk): use the mesh fallback for mesh-only draws on Windows RDNA 2 — Merged Windows AMD mesh-fallback policy precedent; distinct RDNA 2 device-loss fix.
  • PR #219 fix(vk): task + mesh compute fallback by default for Navi 23 on Windows, one rule for the IDs — Merged precedent for Windows AMD task/mesh compute fallback defaults.
Possibly related, for context: - issue #1897 Game: Intel Arc iGPU (Meteor Lake) on Linux, Mesa 26.2.5: 9-10 fps in Kokiri Forest; native float add, off by default here, more than halves the GPU time — Open issue reports the same native float-add policy mechanism and performance impact, on Intel Mesa. - issue #1477 Deck/AMD: rewrite the mesh-shader path — build-time translation of task/mesh into native pipelines — Open mesh-path issue tracks compute-versus-native mesh shader design and performance. - PR #1650 fix(vk): use the mesh fallback for mesh-only draws on Windows RDNA 2 — Merged Windows AMD mesh-fallback policy precedent; distinct RDNA 2 device-loss fix. - PR #219 fix(vk): task + mesh compute fallback by default for Navi 23 on Windows, one rule for the IDs — Merged precedent for Windows AMD task/mesh compute fallback defaults.
Owner
@fix
BeanBot changed title from perf(vk): speed up Radeon 860M/840M (+3-12%) with compute mesh and native float atomics to perf(vk): enable native float atomics on Radeon 860M/840M 2026-11-21 17:09:57 +00:00
BeanBot force-pushed perf/vk-860m-compute-mesh-native-fadd from 9c3a8494ff
Some checks failed
Website / Website on Cloudflare (pull_request_target) Has been skipped
PR description / Description follows the template (pull_request_target) Successful in 5s
CI / Conventional Commit title (pull_request) Has been cancelled
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been cancelled
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been cancelled
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been cancelled
CI / Workflow lint (actionlint) (pull_request) Has been cancelled
CI / Website build and checks (pull_request) Has been cancelled
Guide screenshots / Guide images and local UI source review (pull_request) Has been cancelled
Third-party parts / As their notes say (pull_request) Has been cancelled
Third-party parts / The same as upstream (pull_request) Has been cancelled
to 893356016d
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 11s
PR description / Description follows the template (pull_request_target) Has been cancelled
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 17s
CI / Workflow lint (actionlint) (pull_request) Successful in 18s
Third-party parts / As their notes say (pull_request) Successful in 20s
CI / Website build and checks (pull_request) Successful in 11m51s
2026-11-21 17:09:58 +00:00
Compare
BeanBot left a comment

Recommendation: Approve. The standalone performance gain from the Radeon native float-add default remains unproven. The supplied comparisons also include the removed compute-mesh default and, for the FPS table, #2085. Nothing blocking is left: what remains is for the author before merging (below). Recommend approval; not approved on the forge.

What @fix changed
I changed this pull request: the new head is 893356016d. I rebased it onto main (b74bbe688b).

Conflicts resolved

  • gpu/vk/device_policy.h: Kept the AMD rule and the base branch’s Lunar Lake rule and descriptions.
  • gpu/vk/vk_main.cpp: Kept the AMD and Lunar Lake descriptions and logging.
  • tests/test_native_float_atomic_add.py: Kept both AMD and Lunar Lake cases. I edited its description and retitled it “perf(vk): enable native float atomics on Radeon 860M/840M” to match the change.

Changed

  • Windows-only native float add: Gated AMD's rule on Windows in device creation and policy calculation; added Linux coverage in tests/test_device_policy.py and tests/test_native_float_atomic_add.py. Tests were not run.
  • RX_VK_NATIVE_FADD default documentation: Updated gpu/vk/switches.def:545 with the AMD device and driver threshold and the Intel thresholds.

Changed in part (the rest is for the author)

  • Compute mesh default evidence: Removed the automatic 0x1114 compute rule in gpu/vk/vk_main.cpp:420 and gpu/vk/device_policy.h:542; compute remains available by override. The author must measure the 840M, older Windows drivers and lower-memory systems, including frame pacing and memory pressure, before deciding whether to restore or expand the default.

Nothing was built or run: the checks run again on the new head.

The change enables supported native float atomics by default on Windows Radeon 860M/840M devices with AMD’s proprietary driver from version 0x8001a0. It keeps the existing override and removes the proposed compute-mesh default; the standalone performance benefit is not yet established.

Before merging, for the author (1) (a run, a measurement or a decision; it doesn't hold the recommendation)

  • L3 (minor): Measure the remaining native float-add policy alone against the current merge base with interleaved unprofiled runs and frame-time lows. The existing FPS and span tables include the removed compute-mesh policy and do not establish its standalone gain.

Checks: they run again on this head; a failure shows on the pull request.

Also noticed (3; not asked for, not blocking)
  • The pull request description still claims that automatic mode selects compute mesh and attributes the combined benchmark results to the current change.
  • The base-to-head diff includes several sibling changes beyond the float-add policy described in this pull request.
  • The summary’s approximately 0.38 ms dispatch saving is not established by the displayed standalone measurements.
More: does it help, other systems, switches, what was read

Does it help: Unproven. The reported 3.1–12.4% FPS gains compare a previous version that also enabled compute mesh, with #2085 included in the FPS runs. No isolated interleaved comparison against the current merge base with frame-time lows is supplied.

Other systems: The AMD rule requires Windows, the proprietary driver, device 0x1114 and the stated driver threshold; extension and feature support are also required. The explicit override remains available on other systems. The supplied game run covers an 860M on Windows, not an 840M.

Switches: RX_VK_NATIVE_FADD changes its automatic default to native on supported Windows AMD 0x1114 devices from driver version 0x8001a0; 0 retains the loop and 1 requests native support. The default’s performance justification still needs an isolated measurement. The switch table describes the new default.

Reviewed: 893356016d against b74bbe688b: Read AGENTS.md, CONTRIBUTING.md, the pull request template, the policy and device-creation sections and their callers, the relevant switch entry, and the changed policy and native-float test cases. Skimmed the remaining diff and commit list; the base-to-head diff also contains sibling changes, so not every changed file was read in full. docs/review-checklist.md and the adjacent Tooling AGENTS.md were unavailable.; nothing run (a code review).

Checked and fine:

  • The Windows condition reaches both device creation and the device policy calculation.
  • The policy checks the AMD device ID and driver threshold, and the tests cover the threshold and Linux exclusion.
  • Automatic compute mesh remains limited to RADV; the explicit mesh override remains available.

Testing: No code or tests were run. The author reports game, visual and stability checks on an 860M for the previous combined policy; CI has not run on this head.

Overlaps and merge order: The supplied commit list includes #2085 and other merged sibling changes in the base-to-head diff; no merge order can be established from the supplied information.


Automated code review by BeanBot (gpt-6.1-terra), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.

**Recommendation: Approve.** The standalone performance gain from the Radeon native float-add default remains unproven. The supplied comparisons also include the removed compute-mesh default and, for the FPS table, #2085. Nothing blocking is left: what remains is for the author before merging (below). Recommend approval; not approved on the forge. **What `@fix` changed** I changed this pull request: the new head is 893356016d81. I rebased it onto `main` (b74bbe688b8b). **Conflicts resolved** - gpu/vk/device_policy.h: Kept the AMD rule and the base branch’s Lunar Lake rule and descriptions. - gpu/vk/vk_main.cpp: Kept the AMD and Lunar Lake descriptions and logging. - tests/test_native_float_atomic_add.py: Kept both AMD and Lunar Lake cases. I edited its description and retitled it “perf(vk): enable native float atomics on Radeon 860M/840M” to match the change. **Changed** - **Windows-only native float add**: Gated AMD's rule on Windows in device creation and policy calculation; added Linux coverage in tests/test_device_policy.py and tests/test_native_float_atomic_add.py. Tests were not run. - **RX_VK_NATIVE_FADD default documentation**: Updated gpu/vk/switches.def:545 with the AMD device and driver threshold and the Intel thresholds. **Changed in part (the rest is for the author)** - **Compute mesh default evidence**: Removed the automatic 0x1114 compute rule in gpu/vk/vk_main.cpp:420 and gpu/vk/device_policy.h:542; compute remains available by override. The author must measure the 840M, older Windows drivers and lower-memory systems, including frame pacing and memory pressure, before deciding whether to restore or expand the default. Nothing was built or run: the checks run again on the new head. The change enables supported native float atomics by default on Windows Radeon 860M/840M devices with AMD’s proprietary driver from version 0x8001a0. It keeps the existing override and removes the proposed compute-mesh default; the standalone performance benefit is not yet established. **Before merging, for the author (1)** (a run, a measurement or a decision; it doesn't hold the recommendation) - **L3** (minor): Measure the remaining native float-add policy alone against the current merge base with interleaved unprofiled runs and frame-time lows. The existing FPS and span tables include the removed compute-mesh policy and do not establish its standalone gain. **Checks:** they run again on this head; a failure shows on the pull request. <details> <summary>Also noticed (3; not asked for, not blocking)</summary> - The pull request description still claims that automatic mode selects compute mesh and attributes the combined benchmark results to the current change. - The base-to-head diff includes several sibling changes beyond the float-add policy described in this pull request. - The summary’s approximately 0.38 ms dispatch saving is not established by the displayed standalone measurements. </details> <details> <summary>More: does it help, other systems, switches, what was read</summary> **Does it help:** Unproven. The reported 3.1–12.4% FPS gains compare a previous version that also enabled compute mesh, with #2085 included in the FPS runs. No isolated interleaved comparison against the current merge base with frame-time lows is supplied. **Other systems:** The AMD rule requires Windows, the proprietary driver, device 0x1114 and the stated driver threshold; extension and feature support are also required. The explicit override remains available on other systems. The supplied game run covers an 860M on Windows, not an 840M. **Switches:** RX_VK_NATIVE_FADD changes its automatic default to native on supported Windows AMD 0x1114 devices from driver version 0x8001a0; 0 retains the loop and 1 requests native support. The default’s performance justification still needs an isolated measurement. The switch table describes the new default. **Reviewed:** 893356016d81 against b74bbe688b8b: Read AGENTS.md, CONTRIBUTING.md, the pull request template, the policy and device-creation sections and their callers, the relevant switch entry, and the changed policy and native-float test cases. Skimmed the remaining diff and commit list; the base-to-head diff also contains sibling changes, so not every changed file was read in full. docs/review-checklist.md and the adjacent Tooling AGENTS.md were unavailable.; nothing run (a code review). **Checked and fine:** - The Windows condition reaches both device creation and the device policy calculation. - The policy checks the AMD device ID and driver threshold, and the tests cover the threshold and Linux exclusion. - Automatic compute mesh remains limited to RADV; the explicit mesh override remains available. **Testing:** No code or tests were run. The author reports game, visual and stability checks on an 860M for the previous combined policy; CI has not run on this head. **Overlaps and merge order:** The supplied commit list includes #2085 and other merged sibling changes in the base-to-head diff; no merge order can be established from the supplied information. </details> --- <sub>Automated code review by BeanBot (gpt-6.1-terra), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.</sub>
HumanBean deleted branch perf/vk-860m-compute-mesh-native-fadd 2026-11-21 17:10:51 +00:00
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
3 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Longshot!2084
No description provided.