[Kernel] Marlin: deprecate VLLM_MARLIN_USE_ATOMIC_ADD and drop the bf16 sm90 build

CMakeLists.txt

@@ -651,7 +651,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA" OR VLLM_GPU_LANG STREQUAL "HIP")
     # Keep building Marlin for 9.0 as there are some group sizes and shapes that
     # are not supported by Machete yet.
 
-    # marlin arches for fp16 output
+    # marlin arches for fp16/bf16 output
     # Family-conditional 12.0f (one cubin for SM12x family) requires CUDA >= 13.0;
     # fall back to architecture-specific 12.0a;12.1a on CUDA < 13.0 (e.g. 12.8).
     if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
@@ -661,12 +661,6 @@ if(VLLM_GPU_LANG STREQUAL "CUDA" OR VLLM_GPU_LANG STREQUAL "HIP")
     endif()
     # marlin has limited support for turing
     cuda_archs_loose_intersection(MARLIN_SM75_ARCHS "7.5" "${CUDA_ARCHS}")
-    # marlin arches for bf16 output (we need 9.0 for bf16 atomicAdd PTX)
-    if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
-      cuda_archs_loose_intersection(MARLIN_BF16_ARCHS "8.0+PTX;9.0+PTX;12.0f" "${CUDA_ARCHS}")
-    else()
-      cuda_archs_loose_intersection(MARLIN_BF16_ARCHS "8.0+PTX;9.0+PTX;12.0a;12.1a" "${CUDA_ARCHS}")
-    endif()
     # marlin arches for fp8 input
     # - sm80 doesn't support fp8 computation
     # - sm90 and sm100 don't support QMMA.16832.F32.E4M3.E4M3 SAAS instruction
@@ -721,7 +715,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA" OR VLLM_GPU_LANG STREQUAL "HIP")
       endif()
 
       if (MARLIN_ARCHS)
-        file(GLOB MARLIN_TEMPLATE_KERNEL_SRC "csrc/quantization/marlin/sm80_kernel_*_float16.cu")
+        file(GLOB MARLIN_TEMPLATE_KERNEL_SRC "csrc/quantization/marlin/sm80_kernel_*.cu")
         set_gencode_flags_for_srcs(
           SRCS "${MARLIN_TEMPLATE_KERNEL_SRC}"
           CUDA_ARCHS "${MARLIN_ARCHS}")
@@ -730,16 +724,6 @@ if(VLLM_GPU_LANG STREQUAL "CUDA" OR VLLM_GPU_LANG STREQUAL "HIP")
             PROPERTIES COMPILE_FLAGS "-static-global-template-stub=false")
         endif()
         list(APPEND VLLM_STABLE_EXT_SRC ${MARLIN_TEMPLATE_KERNEL_SRC})
-
-        file(GLOB MARLIN_TEMPLATE_BF16_KERNEL_SRC "csrc/quantization/marlin/sm80_kernel_*_bfloat16.cu")
-        set_gencode_flags_for_srcs(
-          SRCS "${MARLIN_TEMPLATE_BF16_KERNEL_SRC}"
-          CUDA_ARCHS "${MARLIN_BF16_ARCHS}")
-        if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8)
-          set_source_files_properties(${MARLIN_TEMPLATE_BF16_KERNEL_SRC}
-            PROPERTIES COMPILE_FLAGS "-static-global-template-stub=false")
-        endif()
-        list(APPEND VLLM_STABLE_EXT_SRC ${MARLIN_TEMPLATE_BF16_KERNEL_SRC})
       endif()
 
       if (MARLIN_SM75_ARCHS)

recipes/models/MiniMaxAI/MiniMax-M3.yaml

@@ -185,9 +185,8 @@ variants:
           - "--max-model-len"
           - "131072"
       h200:
-        # Humming runs the dense MXFP8 linear layers; the atomic-add env var
-        # enables Marlin's atomic reduction on the MXFP8 MoE path. This image
-        # also supports the FP8 MSA indexer cache on H200.
+        # Humming runs the dense MXFP8 linear layers. This image also supports
+        # the FP8 MSA indexer cache on H200.
         docker_image: "vllm/vllm-openai:nightly-1a001d584244bd21829a5e5bd4a376ea141c0ce3"
         extra_args:
           - "--attention-backend"
@@ -199,8 +198,6 @@ variants:
           - "--gpu-memory-utilization"
           - "0.95"
           - "--enable-prefix-caching"
-        extra_env:
-          VLLM_MARLIN_USE_ATOMIC_ADD: "1"
       mi300x:
         docker_image: "vllm/vllm-openai-rocm:nightly"
         extra_args:

tests/config/test_config_generation.py

@@ -1,5 +1,7 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import logging
+
 import pytest
 
 from vllm.engine.arg_utils import EngineArgs
@@ -111,6 +113,15 @@ def test_unrecognized_env(monkeypatch):
     engine_args.create_engine_config()
 
 
+def test_deprecated_marlin_atomic_add_env_warns(monkeypatch, caplog_vllm):
+    monkeypatch.setenv("VLLM_MARLIN_USE_ATOMIC_ADD", "1")
+    with caplog_vllm.at_level(logging.WARNING, logger="vllm"):
+        EngineArgs().create_engine_config()
+    assert any(
+        "VLLM_MARLIN_USE_ATOMIC_ADD is deprecated" in m for m in caplog_vllm.messages
+    )
+
+
 def test_get_diff_sampling_param_includes_penalties():
     """``--override-generation-config`` must apply presence_penalty and
     frequency_penalty as server-side defaults, like the other sampling

vllm/engine/arg_utils.py

@@ -2274,6 +2274,11 @@ def create_engine_config(
         device_config = DeviceConfig(device=cast(Device, current_platform.device_type))
 
         current_platform.validate_environ(self.fail_on_environ_validation)
+        if envs.is_set("VLLM_MARLIN_USE_ATOMIC_ADD"):
+            logger.warning(
+                "VLLM_MARLIN_USE_ATOMIC_ADD is deprecated, has no effect, and "
+                "will be removed in v0.35.0."
+            )
 
         # Check if the model is a speculator and override model/tokenizer/config
         # BEFORE creating ModelConfig, so the config is created with the target model

vllm/envs.py

@@ -1596,7 +1596,7 @@ def _resolve_rust_cli_path() -> str | None:
     "VLLM_MODEL_REDIRECT_PATH": lambda: os.environ.get(
         "VLLM_MODEL_REDIRECT_PATH", None
     ),
-    # Whether to use atomicAdd reduce in gptq/awq marlin kernel.
+    # Deprecated and has no effect; will be removed in v0.35.0.
     "VLLM_MARLIN_USE_ATOMIC_ADD": lambda: (
         os.environ.get("VLLM_MARLIN_USE_ATOMIC_ADD", "0") == "1"
     ),

vllm/model_executor/layers/quantization/utils/marlin_utils.py

@@ -635,18 +635,6 @@ def moe_awq_to_marlin_zero_points(
     return output
 
 
-def maybe_warn_marlin_atomic_add(device, dtype):
-    if torch.compiler.is_dynamo_compiling():
-        return
-    device_capability = torch.cuda.get_device_capability(device)
-    if device_capability[0] < 9 and dtype == torch.bfloat16:
-        logger.info_once(
-            "You are running Marlin kernel with bf16 on GPUs before SM90. "
-            "You can consider change to fp16 to achieve better performance "
-            "if possible."
-        )
-
-
 def moe_packed_to_marlin_zero_points(
     q_zp_packed: torch.Tensor,
     size_k: int,
@@ -671,42 +659,6 @@ def moe_packed_to_marlin_zero_points(
     return output
 
 
-def maybe_warn_marlin_atomic_add_env():
-    if torch.compiler.is_dynamo_compiling():
-        return
-    if envs.VLLM_MARLIN_USE_ATOMIC_ADD:
-        return
-    logger.info_once(
-        "Marlin kernel can achieve better performance for small size_n "
-        "with experimental use_atomic_add feature. "
-        "You can consider set environment variable "
-        "VLLM_MARLIN_USE_ATOMIC_ADD to 1 if possible."
-    )
-
-
-def should_use_atomic_add_reduce(
-    m: int, n: int, k: int, device: torch.device, dtype: torch.dtype
-) -> bool:
-    # the performance of atomicAdd is better than global reduce
-    # only when m*n is small and k is large
-    if n >= 2048 or k < 2048 or device.type != "cuda":
-        return False
-
-    # disable atomicAdd reduce by default,
-    # one can enable it with VLLM_MARLIN_USE_ATOMIC_ADD=1
-    if not envs.VLLM_MARLIN_USE_ATOMIC_ADD:
-        maybe_warn_marlin_atomic_add_env()
-        return False
-
-    # sm8x doesn't support atomicAdd + bfloat16 natively
-    device_capability = torch.cuda.get_device_capability(device)
-    if device_capability[0] < 9 and dtype == torch.bfloat16:
-        maybe_warn_marlin_atomic_add(device, dtype)
-        return False
-
-    return True
-
-
 _quant_fp8_method: QuantFP8 | None = None
 
 
@@ -769,14 +721,6 @@ def apply_gptq_marlin_linear(
     padded_n, padded_k = marlin_repacked_nk(weight, wtype.size_bits)
     reshaped_x = marlin_pad_dim(reshaped_x, input_size_per_partition, padded_k)
 
-    use_atomic_add = should_use_atomic_add_reduce(
-        m=reshaped_x.size(0),
-        n=padded_n,
-        k=padded_k,
-        device=input.device,
-        dtype=input.dtype,
-    )
-
     a_scales = None
     if input_dtype == torch.int8:
         assert wtype == scalar_types.uint4b8, (
@@ -804,7 +748,7 @@ def apply_gptq_marlin_linear(
         size_m=reshaped_x.shape[0],
         size_n=padded_n,
         size_k=padded_k,
-        use_atomic_add=use_atomic_add,
+        use_atomic_add=False,
         use_fp32_reduce=use_fp32_reduce,
         is_zp_float=False,
     )

vllm/model_executor/layers/quantization/utils/marlin_utils_fp4.py

@@ -19,7 +19,6 @@
     marlin_quant_input,
     marlin_repacked_nk,
     marlin_unpad_output,
-    should_use_atomic_add_reduce,
 )
 from vllm.model_executor.utils import replace_parameter
 from vllm.platforms import current_platform
@@ -177,14 +176,6 @@ def apply_fp4_marlin_linear(
     padded_n, padded_k = marlin_repacked_nk(weight, num_bits=4)
     reshaped_x = marlin_pad_dim(reshaped_x, size_k, padded_k)
 
-    use_atomic_add = should_use_atomic_add_reduce(
-        m=reshaped_x.size(0),
-        n=padded_n,
-        k=padded_k,
-        device=input.device,
-        dtype=input.dtype,
-    )
-
     inputs = reshaped_x
     a_scales = None
     is_nvfp4 = weight_global_scale is not None
@@ -209,7 +200,7 @@ def apply_fp4_marlin_linear(
         size_m=reshaped_x.size(0),
         size_n=padded_n,
         size_k=padded_k,
-        use_atomic_add=use_atomic_add,
+        use_atomic_add=False,
         use_fp32_reduce=use_fp32_reduce,
     )
 

vllm/model_executor/layers/quantization/utils/marlin_utils_fp8.py

@@ -18,7 +18,6 @@
     marlin_permute_scales,
     marlin_repacked_nk,
     marlin_unpad_output,
-    should_use_atomic_add_reduce,
 )
 from vllm.model_executor.utils import replace_parameter
 from vllm.platforms import current_platform
@@ -65,14 +64,6 @@ def apply_fp8_marlin_linear(
     padded_n, padded_k = marlin_repacked_nk(weight, num_bits=8)
     reshaped_x = marlin_pad_dim(reshaped_x, size_k, padded_k)
 
-    use_atomic_add = should_use_atomic_add_reduce(
-        m=reshaped_x.size(0),
-        n=padded_n,
-        k=padded_k,
-        device=input.device,
-        dtype=input.dtype,
-    )
-
     inputs = reshaped_x
     a_scales = None
     if input_dtype is not None and input_dtype.itemsize == 1:
@@ -92,7 +83,7 @@ def apply_fp8_marlin_linear(
         size_m=reshaped_x.size(0),
         size_n=padded_n,
         size_k=padded_k,
-        use_atomic_add=use_atomic_add,
+        use_atomic_add=False,
         use_fp32_reduce=use_fp32_reduce,
     )
 
@@ -388,14 +379,6 @@ def apply_mxfp8_marlin_linear(
     padded_n, padded_k = marlin_repacked_nk(weight, num_bits=8)
     reshaped_x = marlin_pad_dim(reshaped_x, size_k, padded_k)
 
-    use_atomic_add = should_use_atomic_add_reduce(
-        m=reshaped_x.size(0),
-        n=padded_n,
-        k=padded_k,
-        device=input.device,
-        dtype=input.dtype,
-    )
-
     output = torch.ops.vllm.marlin_gemm(
         a=reshaped_x,
         b_q_weight=weight,
@@ -409,7 +392,7 @@ def apply_mxfp8_marlin_linear(
         size_m=reshaped_x.size(0),
         size_n=padded_n,
         size_k=padded_k,
-        use_atomic_add=use_atomic_add,
+        use_atomic_add=False,
         use_fp32_reduce=use_fp32_reduce,
     )