Add fused_dynamic_mxfp4_quant_moe_sort_hip by junhaha666 · Pull Request #2620 · ROCm/aiter

junhaha666 · 2026-04-05T14:24:41Z

python3 op_tests/test_moe_sorting_mxfp4.py -ek 256,8 -dim 7168
test on MI355:

Motivation

Technical Details

Test Plan

Test Result

Submission Checklist

Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

github-actions · 2026-04-05T14:25:12Z

🏷️ CI Guide

Runs automatically on every PR:

✅ Pre-checks (submodule verification, code formatting)
✅ Aiter op tests (gfx942 + gfx950)
✅ Triton tests (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
`ci:triton-355`	Run Triton tests on MI355 in addition to MI325
`ci:sglang`	SGLang integration tests
`ci:atom`	ATOM benchmark (DeepSeek-R1 + GPT-OSS)
`ci:vllm`	vLLM benchmark
`ci:all`	All of the above

Add labels via the sidebar or gh pr edit 2620 --add-label <label>

Copilot

Pull request overview

This PR introduces a new HIP fused kernel path to perform dynamic MXFP4 (fp4x2) quantization while also producing MoE-sorted (swizzled) e8m0 scale bytes, and wires it into the Python API and MoE flow with an accompanying benchmark/test.

Changes:

Add fused_dynamic_mxfp4_quant_moe_sort_hip HIP kernel + pybind export.
Add Python wrapper fused_dynamic_mxfp4_quant_moe_sort(...) and switch fused_moe.py to use it.
Extend op_tests/test_moe_sorting_mxfp4.py to benchmark/validate HIP vs Triton scale sorting.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
`csrc/kernels/quant_kernels.cu`	Adds the fused quant+MoE-scale-sort HIP kernel and C++ entrypoint.
`csrc/include/quant.h`	Declares the new HIP entrypoint in the quant header.
`csrc/include/rocm_ops.hpp`	Exposes the new entrypoint to Python via pybind.
`aiter/ops/quant.py`	Adds Python bindings + a convenience wrapper that allocates output/scale buffers.
`aiter/fused_moe.py`	Switches MoE quant+sort path to use the new top-level fused helper.
`op_tests/test_moe_sorting_mxfp4.py`	Adds HIP-vs-reference checks and updates CLI options/bench coverage.

Comments suppressed due to low confidence (1)

op_tests/test_moe_sorting_mxfp4.py:25

run_torch uses sorted_ids[num_valid_ids:] slicing, but in both call sites num_valid_ids is a CUDA tensor (num_valid_ids = num_valid_ids[0]). CUDA tensors can’t be used as Python slice indices, so this will raise at runtime. Convert to a Python int before calling (e.g., num_valid_ids = int(num_valid_ids[0].item())) or change run_torch to do the .item() internally for the slice boundary.

def run_torch(scale, sorted_ids, num_valid_ids, token_num):
    topk = 1
    if len(scale.shape) == 3:
        topk = scale.shape[1]
        scale = scale.view(-1, scale.shape[-1])
    sorted_ids[num_valid_ids:] = token_num
    topk_ids = sorted_ids >> 24

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

…_mxfp4_quant_moe_sort

…fp4_sort in fused_moe

zejunchen-zejun · 2026-04-16T06:37:12Z

+                scale[addr]     = bs_e8m0;
+            }
+
+            if(topk_id < topk)


Hi @junhaha666
We found EP MXFP4 acc issue may be from here.
For stage1 and topk=1, the code is equal to if(topk_id == 0). When TP mode, there is no issue because all tokens' topk_id is in this rank, while for EP mode, the token's top-1 expert maybe on other rank and here if condition exclude this token computation even when this token picks the expert in this rank.
CC: @ZhangLirong-amd

* add fused_dynamic_mxfp4_quant_moe_sort_hip * use hip fused_dynamic_mxfp4_quant_moe_sort in fuse_moe * update * add mxfp4_moe_sort_hip * add dispatch to choose the fused kernel or not in aiter.fused_dynamic_mxfp4_quant_moe_sort * rm topk in api and use mxfp4_moe_sort_fwd instead of fp4_utils.moe_mxfp4_sort in fused_moe * format * update --------- Co-authored-by: Lingpeng Jin <103567126+valarLip@users.noreply.github.com>

aiter/fused_moe.py: - Restore to origin/main. Per sunway513's own comment, #2457 and #2547 were excluded from this bulk merge; per valarLip, #2687 was rejected. No source PR should land changes in this file. The previous state (+110/-119 vs main) was collateral damage from auto-resolved conflicts taking older sides, which silently reverted #2262 (xbf16 asm fmoe path), #2726 (FlyDSL a8w4 MoE wrapper params + fuse_quant), #2658 (CK fp8 blockscale splitk tuner support), and #2620 (mxfp4_moe_sort_hip, flagged by valarLip). op_tests/test_gemm_a8w8_blockscale.py: - Replace with a clean 3-way merge of origin/main + #2541. Now +55/-0 vs main, matching #2541's actual contribution exactly. The previous state was silently reverting #2645 (CK GEMM multi-arch + test infra: TEST_NUM_ITERS, --csv/--output args, kernel_name= param). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

@xaguilar-amd

…3-Next, pa_mqa OOB) (#3005) * fix: remap QuantType.No to per_1x32 for fp4x2 MoE weights (W4A6 support) * Fixing two cascading bugs when running the MoE tuner * Enable split-K for block-scale A8W8 CK and CKTile GEMMs Propagate the splitK parameter (as KBatch = 2^splitK) through the block-scale GEMM kernel infrastructure so that the tuning scripts can sweep split-K values to improve occupancy on small-M shapes. CK path: add KBatch parameter to gemm_a8w8_blockscale_impl and call SetKBatch on the device argument. The CK invoker handles output zeroing and atomic accumulation internally. CKTile path: add k_batch parameter to gemm_a8w8_blockscale_cktile_impl, remove the "split-k is not supported yet" runtime guard, and add hipMemsetAsync to zero the output buffer before atomic accumulation. Non-tune entry points pass KBatch=1 (no split-K) to preserve existing behavior. Code generation scripts (gen_instances.py, gen_instances_cktile.py) updated to include the new parameter in generated wrappers and manifests. Made-with: Cursor * Wire splitK from tuning CSV through production blockscale GEMM dispatch The tuning infrastructure already sweeps splitK and writes it to the CSV, but the production dispatch ignored it and hardcoded KBatch=1. Add splitK as a runtime parameter to the non-tune entry points so tuned split-K values are used without compiling the full _tune instance set. Made-with: Cursor * fix: ck_moe_stage1 split-K output buffer overflow from padding scatter The CK kernel scatters output via sorted_token_ids using: token_offset = (fused_token & 0xffffff) * topk + (fused_token >> 24) Padding entries use the sentinel value (topk << 24 | token_num), which decodes to scatter position (token_num * topk + topk) -- beyond the valid output range [0, token_num * topk). The original buffer (token_num, topk, w1.shape[1]) only has token_num * topk rows, so the padding scatter writes out of bounds, causing "HIP runtime error: invalid argument" during CUDA graph capture (e.g. DeepSeek-R1 decode with token_num=1, topk=8, block_m=16). Fix: allocate (token_num * topk + topk + 1) rows -- the exact minimum needed to absorb all padding scatter writes. After the kernel, slice only the valid [0, token_num * topk) rows for the activation. Related: #2508 Made-with: Cursor * Address PR review feedback: validate splitK, fix hipMemset stride issue, add correctness test Agent-Logs-Url: https://github.com/ROCm/aiter/sessions/e3b37b0f-e151-4935-ad89-fd72436d41e2 Co-authored-by: samremes <181322991+samremes@users.noreply.github.com> * black format * fix splitk test dimensions * Add gdn fusions * style: fix ruff F841 and black-format Triton PR files Remove unused variable in rmsnorm FP8 test ref. Apply Black to kernels, launchers, tests, and gated_delta_rule decode __init__. Made-with: Cursor * Update fused_rearrange_sigmoid_gdr.py * Update op_tests * Fix BLACK format problem * Fix black check failure * Update test_fused_rearrange_sigmoid_gdr.py * Allow callers to pass pre-allocated moe_buf to avoid output copy Add an optional `moe_buf` parameter through the moe_sorting and fused_moe call chain. When provided, the sorting kernel writes directly into the caller's buffer instead of allocating a new one, eliminating a redundant copy on the output path. Made-with: Cursor * Add moe_buf pass-through test to existing test_moe_sorting Made-with: Cursor * Replace _fast with _single_token for causal conv1d update kernels for single token decoding * Fix blck format error * Add tuned a8w8 blockscale GEMM config for Qwen3-Next-80B-A3B on MI355X Tuned 1482 shapes (TP1/TP2/TP4) for Qwen/Qwen3-Next-80B-A3B-Instruct-FP8 on MI355X using CK + CK-TILE backends with splitK support. Depends on: - PR #2862 (CK bump for stride fix in CK-TILE blockscale) - PR #2541 (splitK support for CK/CK-TILE blockscale GEMMs) - PR #2487 (AQLayout tunable for CK-TILE blockscale 8-warp kernels) * refactor(triton): rename gated RMSNorm+FP8 op to fused_rms_gated_fp8_group_quant Colocate the gated RMSNorm + FP8 group quant path with the other fused FP8 ops. The Triton kernel is now _fused_rms_gated_fp8_group_quant_kernel in _triton_kernels/quant/fused_fp8_quant.py; the Python entry point is fused_rms_gated_fp8_group_quant in quant/fused_fp8_quant.py, with a docstring that contrasts it with fused_rms_fp8_group_quant. Remove the old rmsnorm_input_quant_fp8 module and rms_norm_input_quant_fp8 kernel file. Re-export the new symbol and helpers (get_fp8_min_max_bounds, calc_rows_per_block) from aiter.ops.triton.quant. Rename the test file to test_fused_rms_gated_fp8_group_quant.py and update test.sh. BREAKING CHANGE: rmsnorm_input_quant_fp8 is removed; use fused_rms_gated_fp8_group_quant instead. Made-with: Cursor * Retune blockscale GEMM configs to fix invalid kernelId+splitK combinations Full retune of all 1482 shapes on MI355X (gfx950, cu_num=256). Key changes: - SplitK usage dropped from 613 to 88 CK shapes (splitK > 0) - All shapes validated via --run_config (1482/1482 OK) - E2e perf: 2-8% output throughput improvement vs untuned heuristic * [Bug] pa_mqa_logits: mask OOB stores on OutLogits_buffer The gluon `_gluon_deepgemm_fp8_paged_mqa_logits_preshuffle` and `_gluon_deepgemm_fp8_paged_mqa_logits_preshuffle_varctx` kernels have 10 `buffer_store(ptr=OutLogits_buffer, ...)` call sites that are missing the upper-bound mask present on their sibling stores. When `context_length == max_model_len` (the last-token position in a long- context decode step), `split_context_length` is rounded UP to a `KVBlockSize` multiple at line 427 and the final prefix/suffix store then writes up to `ChunkKPerStage` float32 elements past the logical row end. With `stride_out_batch == max_model_len`, those writes cross into the next row / the next allocation, causing intermittent HIP memory-access faults on gfx950 during DeepSeek V3.2 MTP decoding. This change adds `mask=<offset> < max_model_len` to every unmasked `buffer_store` on `OutLogits_buffer` in both preshuffle kernels, matching the pattern of their already-masked neighbours. The existing `tl.where(..., -inf)` masking of the *values* is preserved; the only behavioural change is that out-of-row lanes no longer emit buffer stores. Hardware overhead is negligible: `buffer_store` with a predicate is the same SMEM descriptor path as the unmasked variant, just with a VCC mask setup. Repro + end-to-end fix evidence: see PR description. Signed-off-by: Markus Hartikainen <markus.hartikainen@amd.com> * style: fix Black formatting * style: fix Black formatting (Python 3.12 compatible) * ci: replace deprecated zmq package with pyzmq The `zmq` meta-package fails to install on some CI runners because it cannot resolve the `pyzmq` dependency. Use `pyzmq` directly, which is the actual package providing ZeroMQ bindings for Python. Fixes Triton Test Shard 7 setup failures. * ci: increase pip retries and timeout for CI reliability Set pip global retries=15 and timeout=120s in build_aiter_triton.sh to handle transient PyPI network failures on self-hosted runners. Shard 5/7 failures were caused by RemoteDisconnected during pip install. * ci: make pyzmq install non-blocking in triton test setup pyzmq is only used by aiter.dist.shm_broadcast, not by any triton test. When PyPI is unreachable on self-hosted runners, the pyzmq install failure should not block the entire CI shard. Split pyzmq into a separate pip install with || fallback so triton tests can proceed even when PyPI connectivity is degraded. * ci: retry pip install individually on batch failure When batch pip install fails (e.g., PyPI connectivity issues on self-hosted runners), retry each package individually. Only pyzmq is allowed to fail silently since it's only used by aiter.dist.shm_broadcast and not required by any CI test suite. Critical packages (pandas, einops, numpy) must still succeed. * [MLA] Fix nhead=32 non-persistent decode crash on gfx950 Commit c849fd5 ("Add bf16 MLA decode kernel for gqa_ratio=64, qseqlen=1 (non-persistent)") zeroed ptr_RP and out_16_nosplit for all non-persistent dispatch. The legacy QH16 ASM kernel used for nhead=32 (MLA_A16W16_1TG_4W_32mx1_16nx1_Coex0_Msk1_QH16.co) still writes directly to the output buffer via ptr_RP when kv_split==1. Dereferencing nullptr causes a GPU memory access fault during CUDA graph capture on MI355X (gfx950) with DeepSeek-V3.2 at TP4. Fix: - Conditionally restore ptr_RP and out_16_nosplit in the non-persistent path for legacy kernels (gqa_ratio * max_seqlen_q <= 64) while keeping nullptr for newer kernels (e.g. gqa_ratio=64). - Restore the bf16 nhead in [32,64] early-return after stage1 when num_kv_splits==1 to prevent stage2 from overwriting the kernel's direct output. Tested on MI355X TP4 with deepseek-ai/DeepSeek-V3.2 (nhead=32): - No crash during CUDA graph capture - Correct GSM8K accuracy Made-with: Cursor * revert: remove #2983 (MLA nhead=32 fix) — causes test_mla CI failures Reverting cherry-pick of #2983 from this bulk merge. The MLA nhead=32 non-persistent decode fix causes deterministic test_mla k_cache and mla_decode-absorb precision failures on CI MI35X runners (Shard 1 & 2). #2983 should go through its own PR with proper CI validation by the original author (frida-andersson). * fix: restore tuple unpack for FlyDSL fused-quant stage1 return flydsl_moe_stage1 returns (out, out_scale_sorted) when the kernel uses fused fp4/fp8 quantization. The tuple unpack logic was removed during earlier refactoring but the kernel behavior was not changed, causing fused_moe_2stages to crash with: AttributeError: 'tuple' object has no attribute 'view' Restore the unpack: detect tuple return, extract tensor and scale, handle fp4 byte-packing trim, and skip redundant Python-side requant when the kernel already produced sorted scales. * Revert leaked changes from excluded PRs #2457/#2547/#2687 in fused_moe.py - Restore import to match main: use `from aiter import fused_dynamic_mxfp4_quant_moe_sort, mxfp4_moe_sort_fwd` instead of importing from internal triton path and fp4_utils - Replace all fp4_utils.moe_mxfp4_sort() calls with mxfp4_moe_sort_fwd() using correct parameter names (cols= instead of block_size=) - Remove all moe_buf preallocated buffer additions (PR #2687 rejected): parameter defaults, if-guards, and pass-throughs in _moe_sorting_impl, moe_sorting, fused_moe, fused_moe_fake, and fused_moe_ - Fix moe_sorting_dispatch_policy type annotation: bool -> int in fused_moe_fake and fused_moe_ - Remove moe_buf pass-through test from test_moe_sorting.py - Preserve legitimate fp4_utils usage (mxfp4_to_f32, e8m0_to_f32) with local imports in stage1/stage2 fallback functions * fix: restore fp4_utils.moe_mxfp4_sort for new code paths (different output layout than mxfp4_moe_sort_fwd) * style: fix Black formatting for local imports * fix: remove rejected W4A6 QuantType remap from fused_moe_dp_shared_expert Lingpeng explicitly rejected this change (from excluded PR #2457). Reverts the QuantType.No -> per_1x32 remap for fp4x2 weights. * fix: restore silently-reverted main features from bad merge resolution aiter/fused_moe.py: - Restore to origin/main. Per sunway513's own comment, #2457 and #2547 were excluded from this bulk merge; per valarLip, #2687 was rejected. No source PR should land changes in this file. The previous state (+110/-119 vs main) was collateral damage from auto-resolved conflicts taking older sides, which silently reverted #2262 (xbf16 asm fmoe path), #2726 (FlyDSL a8w4 MoE wrapper params + fuse_quant), #2658 (CK fp8 blockscale splitk tuner support), and #2620 (mxfp4_moe_sort_hip, flagged by valarLip). op_tests/test_gemm_a8w8_blockscale.py: - Replace with a clean 3-way merge of origin/main + #2541. Now +55/-0 vs main, matching #2541's actual contribution exactly. The previous state was silently reverting #2645 (CK GEMM multi-arch + test infra: TEST_NUM_ITERS, --csv/--output args, kernel_name= param). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore: remove #2464 from bulk merge per author request @xaguilar-amd asked to drop #2464 (CK MoE tuner bug fixes) from this bulk merge — they don't need it for the uplift. Verified that #2464 is the only PR in this bulk merge touching aiter/jit/core.py and aiter/utility/mp_tuner.py: the diff between the branch and origin/main on those files is exactly #2464's +9/-1 and +5/-0, with no other PR content mixed in. Restoring both files to origin/main therefore drops #2464 cleanly. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Signed-off-by: Markus Hartikainen <markus.hartikainen@amd.com> Co-authored-by: vecheruk-amd <vecheruk@amd.com> Co-authored-by: xaguilar-amd <xavier.aguilarfruto@amd.com> Co-authored-by: Sami Remes <samremes@amd.com> Co-authored-by: Li <chuali@amd.com> Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: samremes <181322991+samremes@users.noreply.github.com> Co-authored-by: hellozhuo <zhuo.su@amd.com> Co-authored-by: Tres Popp <tres.popp@amd.com> Co-authored-by: Juuso Korhonen <40278371+juuso-oskari@users.noreply.github.com> Co-authored-by: Niklas Holmberg <nholmber@users.noreply.github.com> Co-authored-by: Markus Hartikainen <markus.hartikainen@amd.com> Co-authored-by: frida-andersson <fanderss@amd.com> Co-authored-by: Aliasger Zaidy <aliasger.zaidy@amd.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

@xaguilar-amd

…3-Next, pa_mqa OOB) (#3005) * fix: remap QuantType.No to per_1x32 for fp4x2 MoE weights (W4A6 support) * Fixing two cascading bugs when running the MoE tuner * Enable split-K for block-scale A8W8 CK and CKTile GEMMs Propagate the splitK parameter (as KBatch = 2^splitK) through the block-scale GEMM kernel infrastructure so that the tuning scripts can sweep split-K values to improve occupancy on small-M shapes. CK path: add KBatch parameter to gemm_a8w8_blockscale_impl and call SetKBatch on the device argument. The CK invoker handles output zeroing and atomic accumulation internally. CKTile path: add k_batch parameter to gemm_a8w8_blockscale_cktile_impl, remove the "split-k is not supported yet" runtime guard, and add hipMemsetAsync to zero the output buffer before atomic accumulation. Non-tune entry points pass KBatch=1 (no split-K) to preserve existing behavior. Code generation scripts (gen_instances.py, gen_instances_cktile.py) updated to include the new parameter in generated wrappers and manifests. Made-with: Cursor * Wire splitK from tuning CSV through production blockscale GEMM dispatch The tuning infrastructure already sweeps splitK and writes it to the CSV, but the production dispatch ignored it and hardcoded KBatch=1. Add splitK as a runtime parameter to the non-tune entry points so tuned split-K values are used without compiling the full _tune instance set. Made-with: Cursor * fix: ck_moe_stage1 split-K output buffer overflow from padding scatter The CK kernel scatters output via sorted_token_ids using: token_offset = (fused_token & 0xffffff) * topk + (fused_token >> 24) Padding entries use the sentinel value (topk << 24 | token_num), which decodes to scatter position (token_num * topk + topk) -- beyond the valid output range [0, token_num * topk). The original buffer (token_num, topk, w1.shape[1]) only has token_num * topk rows, so the padding scatter writes out of bounds, causing "HIP runtime error: invalid argument" during CUDA graph capture (e.g. DeepSeek-R1 decode with token_num=1, topk=8, block_m=16). Fix: allocate (token_num * topk + topk + 1) rows -- the exact minimum needed to absorb all padding scatter writes. After the kernel, slice only the valid [0, token_num * topk) rows for the activation. Related: #2508 Made-with: Cursor * Address PR review feedback: validate splitK, fix hipMemset stride issue, add correctness test Agent-Logs-Url: https://github.com/ROCm/aiter/sessions/e3b37b0f-e151-4935-ad89-fd72436d41e2 Co-authored-by: samremes <181322991+samremes@users.noreply.github.com> * black format * fix splitk test dimensions * Add gdn fusions * style: fix ruff F841 and black-format Triton PR files Remove unused variable in rmsnorm FP8 test ref. Apply Black to kernels, launchers, tests, and gated_delta_rule decode __init__. Made-with: Cursor * Update fused_rearrange_sigmoid_gdr.py * Update op_tests * Fix BLACK format problem * Fix black check failure * Update test_fused_rearrange_sigmoid_gdr.py * Allow callers to pass pre-allocated moe_buf to avoid output copy Add an optional `moe_buf` parameter through the moe_sorting and fused_moe call chain. When provided, the sorting kernel writes directly into the caller's buffer instead of allocating a new one, eliminating a redundant copy on the output path. Made-with: Cursor * Add moe_buf pass-through test to existing test_moe_sorting Made-with: Cursor * Replace _fast with _single_token for causal conv1d update kernels for single token decoding * Fix blck format error * Add tuned a8w8 blockscale GEMM config for Qwen3-Next-80B-A3B on MI355X Tuned 1482 shapes (TP1/TP2/TP4) for Qwen/Qwen3-Next-80B-A3B-Instruct-FP8 on MI355X using CK + CK-TILE backends with splitK support. Depends on: - PR #2862 (CK bump for stride fix in CK-TILE blockscale) - PR #2541 (splitK support for CK/CK-TILE blockscale GEMMs) - PR #2487 (AQLayout tunable for CK-TILE blockscale 8-warp kernels) * refactor(triton): rename gated RMSNorm+FP8 op to fused_rms_gated_fp8_group_quant Colocate the gated RMSNorm + FP8 group quant path with the other fused FP8 ops. The Triton kernel is now _fused_rms_gated_fp8_group_quant_kernel in _triton_kernels/quant/fused_fp8_quant.py; the Python entry point is fused_rms_gated_fp8_group_quant in quant/fused_fp8_quant.py, with a docstring that contrasts it with fused_rms_fp8_group_quant. Remove the old rmsnorm_input_quant_fp8 module and rms_norm_input_quant_fp8 kernel file. Re-export the new symbol and helpers (get_fp8_min_max_bounds, calc_rows_per_block) from aiter.ops.triton.quant. Rename the test file to test_fused_rms_gated_fp8_group_quant.py and update test.sh. BREAKING CHANGE: rmsnorm_input_quant_fp8 is removed; use fused_rms_gated_fp8_group_quant instead. Made-with: Cursor * Retune blockscale GEMM configs to fix invalid kernelId+splitK combinations Full retune of all 1482 shapes on MI355X (gfx950, cu_num=256). Key changes: - SplitK usage dropped from 613 to 88 CK shapes (splitK > 0) - All shapes validated via --run_config (1482/1482 OK) - E2e perf: 2-8% output throughput improvement vs untuned heuristic * [Bug] pa_mqa_logits: mask OOB stores on OutLogits_buffer The gluon `_gluon_deepgemm_fp8_paged_mqa_logits_preshuffle` and `_gluon_deepgemm_fp8_paged_mqa_logits_preshuffle_varctx` kernels have 10 `buffer_store(ptr=OutLogits_buffer, ...)` call sites that are missing the upper-bound mask present on their sibling stores. When `context_length == max_model_len` (the last-token position in a long- context decode step), `split_context_length` is rounded UP to a `KVBlockSize` multiple at line 427 and the final prefix/suffix store then writes up to `ChunkKPerStage` float32 elements past the logical row end. With `stride_out_batch == max_model_len`, those writes cross into the next row / the next allocation, causing intermittent HIP memory-access faults on gfx950 during DeepSeek V3.2 MTP decoding. This change adds `mask=<offset> < max_model_len` to every unmasked `buffer_store` on `OutLogits_buffer` in both preshuffle kernels, matching the pattern of their already-masked neighbours. The existing `tl.where(..., -inf)` masking of the *values* is preserved; the only behavioural change is that out-of-row lanes no longer emit buffer stores. Hardware overhead is negligible: `buffer_store` with a predicate is the same SMEM descriptor path as the unmasked variant, just with a VCC mask setup. Repro + end-to-end fix evidence: see PR description. Signed-off-by: Markus Hartikainen <markus.hartikainen@amd.com> * style: fix Black formatting * style: fix Black formatting (Python 3.12 compatible) * ci: replace deprecated zmq package with pyzmq The `zmq` meta-package fails to install on some CI runners because it cannot resolve the `pyzmq` dependency. Use `pyzmq` directly, which is the actual package providing ZeroMQ bindings for Python. Fixes Triton Test Shard 7 setup failures. * ci: increase pip retries and timeout for CI reliability Set pip global retries=15 and timeout=120s in build_aiter_triton.sh to handle transient PyPI network failures on self-hosted runners. Shard 5/7 failures were caused by RemoteDisconnected during pip install. * ci: make pyzmq install non-blocking in triton test setup pyzmq is only used by aiter.dist.shm_broadcast, not by any triton test. When PyPI is unreachable on self-hosted runners, the pyzmq install failure should not block the entire CI shard. Split pyzmq into a separate pip install with || fallback so triton tests can proceed even when PyPI connectivity is degraded. * ci: retry pip install individually on batch failure When batch pip install fails (e.g., PyPI connectivity issues on self-hosted runners), retry each package individually. Only pyzmq is allowed to fail silently since it's only used by aiter.dist.shm_broadcast and not required by any CI test suite. Critical packages (pandas, einops, numpy) must still succeed. * [MLA] Fix nhead=32 non-persistent decode crash on gfx950 Commit c849fd5 ("Add bf16 MLA decode kernel for gqa_ratio=64, qseqlen=1 (non-persistent)") zeroed ptr_RP and out_16_nosplit for all non-persistent dispatch. The legacy QH16 ASM kernel used for nhead=32 (MLA_A16W16_1TG_4W_32mx1_16nx1_Coex0_Msk1_QH16.co) still writes directly to the output buffer via ptr_RP when kv_split==1. Dereferencing nullptr causes a GPU memory access fault during CUDA graph capture on MI355X (gfx950) with DeepSeek-V3.2 at TP4. Fix: - Conditionally restore ptr_RP and out_16_nosplit in the non-persistent path for legacy kernels (gqa_ratio * max_seqlen_q <= 64) while keeping nullptr for newer kernels (e.g. gqa_ratio=64). - Restore the bf16 nhead in [32,64] early-return after stage1 when num_kv_splits==1 to prevent stage2 from overwriting the kernel's direct output. Tested on MI355X TP4 with deepseek-ai/DeepSeek-V3.2 (nhead=32): - No crash during CUDA graph capture - Correct GSM8K accuracy Made-with: Cursor * revert: remove #2983 (MLA nhead=32 fix) — causes test_mla CI failures Reverting cherry-pick of #2983 from this bulk merge. The MLA nhead=32 non-persistent decode fix causes deterministic test_mla k_cache and mla_decode-absorb precision failures on CI MI35X runners (Shard 1 & 2). #2983 should go through its own PR with proper CI validation by the original author (frida-andersson). * fix: restore tuple unpack for FlyDSL fused-quant stage1 return flydsl_moe_stage1 returns (out, out_scale_sorted) when the kernel uses fused fp4/fp8 quantization. The tuple unpack logic was removed during earlier refactoring but the kernel behavior was not changed, causing fused_moe_2stages to crash with: AttributeError: 'tuple' object has no attribute 'view' Restore the unpack: detect tuple return, extract tensor and scale, handle fp4 byte-packing trim, and skip redundant Python-side requant when the kernel already produced sorted scales. * Revert leaked changes from excluded PRs #2457/#2547/#2687 in fused_moe.py - Restore import to match main: use `from aiter import fused_dynamic_mxfp4_quant_moe_sort, mxfp4_moe_sort_fwd` instead of importing from internal triton path and fp4_utils - Replace all fp4_utils.moe_mxfp4_sort() calls with mxfp4_moe_sort_fwd() using correct parameter names (cols= instead of block_size=) - Remove all moe_buf preallocated buffer additions (PR #2687 rejected): parameter defaults, if-guards, and pass-throughs in _moe_sorting_impl, moe_sorting, fused_moe, fused_moe_fake, and fused_moe_ - Fix moe_sorting_dispatch_policy type annotation: bool -> int in fused_moe_fake and fused_moe_ - Remove moe_buf pass-through test from test_moe_sorting.py - Preserve legitimate fp4_utils usage (mxfp4_to_f32, e8m0_to_f32) with local imports in stage1/stage2 fallback functions * fix: restore fp4_utils.moe_mxfp4_sort for new code paths (different output layout than mxfp4_moe_sort_fwd) * style: fix Black formatting for local imports * fix: remove rejected W4A6 QuantType remap from fused_moe_dp_shared_expert Lingpeng explicitly rejected this change (from excluded PR #2457). Reverts the QuantType.No -> per_1x32 remap for fp4x2 weights. * fix: restore silently-reverted main features from bad merge resolution aiter/fused_moe.py: - Restore to origin/main. Per sunway513's own comment, #2457 and #2547 were excluded from this bulk merge; per valarLip, #2687 was rejected. No source PR should land changes in this file. The previous state (+110/-119 vs main) was collateral damage from auto-resolved conflicts taking older sides, which silently reverted #2262 (xbf16 asm fmoe path), #2726 (FlyDSL a8w4 MoE wrapper params + fuse_quant), #2658 (CK fp8 blockscale splitk tuner support), and #2620 (mxfp4_moe_sort_hip, flagged by valarLip). op_tests/test_gemm_a8w8_blockscale.py: - Replace with a clean 3-way merge of origin/main + #2541. Now +55/-0 vs main, matching #2541's actual contribution exactly. The previous state was silently reverting #2645 (CK GEMM multi-arch + test infra: TEST_NUM_ITERS, --csv/--output args, kernel_name= param). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore: remove #2464 from bulk merge per author request @xaguilar-amd asked to drop #2464 (CK MoE tuner bug fixes) from this bulk merge — they don't need it for the uplift. Verified that #2464 is the only PR in this bulk merge touching aiter/jit/core.py and aiter/utility/mp_tuner.py: the diff between the branch and origin/main on those files is exactly #2464's +9/-1 and +5/-0, with no other PR content mixed in. Restoring both files to origin/main therefore drops #2464 cleanly. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Signed-off-by: Markus Hartikainen <markus.hartikainen@amd.com> Co-authored-by: vecheruk-amd <vecheruk@amd.com> Co-authored-by: xaguilar-amd <xavier.aguilarfruto@amd.com> Co-authored-by: Sami Remes <samremes@amd.com> Co-authored-by: Li <chuali@amd.com> Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: samremes <181322991+samremes@users.noreply.github.com> Co-authored-by: hellozhuo <zhuo.su@amd.com> Co-authored-by: Tres Popp <tres.popp@amd.com> Co-authored-by: Juuso Korhonen <40278371+juuso-oskari@users.noreply.github.com> Co-authored-by: Niklas Holmberg <nholmber@users.noreply.github.com> Co-authored-by: Markus Hartikainen <markus.hartikainen@amd.com> Co-authored-by: frida-andersson <fanderss@amd.com> Co-authored-by: Aliasger Zaidy <aliasger.zaidy@amd.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

junhaha666 added 2 commits April 5, 2026 14:20

add fused_dynamic_mxfp4_quant_moe_sort_hip

5abd9a9

use hip fused_dynamic_mxfp4_quant_moe_sort in fuse_moe

e735ecd

junhaha666 requested review from a team and Copilot April 5, 2026 14:24

Copilot started reviewing on behalf of junhaha666 April 5, 2026 14:26 View session

junhaha666 added the ci:atom label Apr 5, 2026

Copilot AI reviewed Apr 5, 2026

View reviewed changes

junhaha666 and others added 2 commits April 5, 2026 10:22

update

e7961a0

Merge branch 'main' into jun/fp4_moe_quant_sort

af38f34

junhaha666 marked this pull request as draft April 7, 2026 10:43

junhaha666 added 2 commits April 7, 2026 05:44

add mxfp4_moe_sort_hip

a5da769

add dispatch to choose the fused kernel or not in aiter.fused_dynamic…

b17d399

…_mxfp4_quant_moe_sort

junhaha666 marked this pull request as ready for review April 7, 2026 16:12

junhaha666 and others added 5 commits April 7, 2026 22:16

rm topk in api and use mxfp4_moe_sort_fwd instead of fp4_utils.moe_mx…

3b3257f

…fp4_sort in fused_moe

Merge remote-tracking branch 'origin/main' into jun/fp4_moe_quant_sort

8d16200

format

5684160

update

b3581aa

Merge branch 'main' into jun/fp4_moe_quant_sort

22ffb00

valarLip approved these changes Apr 8, 2026

View reviewed changes

valarLip merged commit 7d063b2 into main Apr 8, 2026
29 of 31 checks passed

valarLip deleted the jun/fp4_moe_quant_sort branch April 8, 2026 15:07

zejunchen-zejun reviewed Apr 16, 2026

View reviewed changes

azaidy mentioned this pull request May 4, 2026

[Silo] Bulk merge: kernel fixes and features (SplitK, MoE fixes, Qwen3-Next, pa_mqa OOB) #3005

Merged

3 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add fused_dynamic_mxfp4_quant_moe_sort_hip#2620

Add fused_dynamic_mxfp4_quant_moe_sort_hip#2620
valarLip merged 11 commits intomainfrom
jun/fp4_moe_quant_sort

junhaha666 commented Apr 5, 2026

Uh oh!

github-actions Bot commented Apr 5, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

zejunchen-zejun Apr 16, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

junhaha666 commented Apr 5, 2026

Motivation

Technical Details

Test Plan

Test Result

Submission Checklist

Uh oh!

github-actions Bot commented Apr 5, 2026

🏷️ CI Guide

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

zejunchen-zejun Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants