add tensorrt_llm common and cutlass_extensions as 3rdparty (#3216)

Co-authored-by: BBuf <35585791+BBuf@users.noreply.github.com>

sgl-kernel/3rdparty/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/tile_interleaved_layout.h

+/***************************************************************************************************
+ * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-License-Identifier: BSD-3-Clause
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+ * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ **************************************************************************************************/
+/*! \file
+    \brief Defines new layouts needed for MoE
+*/
+#pragma once
+
+#include "cutlass/cutlass.h"
+#include "cutlass/fast_math.h"
+#include "cutlass/matrix_coord.h"
+#include "cutlass/pitch_linear_coord.h"
+
+namespace cutlass
+{
+namespace layout
+{
+
+template <int RowsPerTile, int ColumnsInterleaved>
+struct ColumnMajorTileInterleave
+{
+    static constexpr int kRowsPerTile = RowsPerTile;
+    static constexpr int kColumnsInterleaved = ColumnsInterleaved;
+};
+
+template <class T>
+struct IsColumnMajorTileInterleave
+{
+    static constexpr bool value = false;
+};
+
+template <int U, int V>
+struct IsColumnMajorTileInterleave<ColumnMajorTileInterleave<U, V>>
+{
+    static constexpr bool value = true;
+};
+
+} // namespace layout
+} // namespace cutlass

sgl-kernel/3rdparty/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/transform/threadblock/fine_grained_scale_zero_iterator.h

+/***************************************************************************************************
+ * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-License-Identifier: BSD-3-Clause
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+ * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ **************************************************************************************************/
+/*! \file
+    \brief Templates for visiting scales to be used when dequantizing the weights for weight-only GEMM
+           quantization.
+*/
+
+#pragma once
+
+#include "cutlass/array.h"
+#include "cutlass/coord.h"
+#include "cutlass/cutlass.h"
+#include "cutlass/layout/matrix.h"
+#include "cutlass/layout/pitch_linear.h"
+#include "cutlass/matrix_shape.h"
+#include "cutlass/predicate_vector.h"
+#include "cutlass/tensor_ref.h"
+#include "cutlass/tensor_view.h"
+#include "cutlass/transform/threadblock/predicated_tile_access_iterator_params.h"
+
+////////////////////////////////////////////////////////////////////////////////
+
+////////////////////////////////////////////////////////////////////////////////
+
+namespace cutlass
+{
+namespace transform
+{
+namespace threadblock
+{
+
+////////////////////////////////////////////////////////////////////////////////
+
+template <typename Shape, typename Element, typename Layout, int AdvanceRank, int Alignment>
+class FineGrainedScaleZeroIterator;
+
+template <typename Shape_, typename Element_, int Alignment_>
+class FineGrainedScaleZeroIterator<Shape_, Element_, layout::RowMajor, 0, Alignment_>
+{
+public:
+    using Shape = Shape_;
+    using Element = Element_;
+    using Layout = layout::RowMajor;
+    static int const kAdvanceRank = 0;
+    static int const kAlignment = Alignment_;
+
+    static int const kAccessesPerVector = 1;
+
+    /// Row index of scales corresponding to the groupsize of 64
+    int row_groupsize64_;
+    int group_size_;
+
+    using Index = typename Layout::Index;
+    using LongIndex = typename Layout::LongIndex;
+
+    using TensorRef = TensorRef<Element, Layout>;
+    using TensorView = TensorView<Element, Layout>;
+    using TensorCoord = typename Layout::TensorCoord;
+    using Pointer = Element*;
+    using NonConstPointer = typename platform::remove_const<Element>::type*;
+
+    using AccessType = AlignedArray<Element, kAlignment>;
+
+    using Fragment = cutlass::Array<Element, kAlignment>;
+
+    // For compatibility with existing iterator interface
+    struct Params
+    {
+        LongIndex stride_ = 0;
+
+        /// amount (in byte) to increment pointer from first access of current tile
+        /// to first access of next tile
+        LongIndex inc_advance_ = 0;
+
+        // Default ctor
+        CUTLASS_HOST_DEVICE
+        Params() {}
+
+        /// Construct the Params object given a pitch-linear tensor's layout
+        CUTLASS_HOST_DEVICE
+        Params(Layout const& layout)
+            : stride_(layout.stride(0))
+        {
+            inc_advance_ = Shape::kRow * stride_ * sizeof_bits<Element>::value / 8;
+        }
+    };
+
+private:
+    /// Internal pointer type permits fast address arithmetic
+    using BytePointer = char*;
+
+private:
+    //
+    // Data members
+    //
+
+    /// Parameters object with precomputed internal state
+    Params const params_;
+
+    /// Internal pointer to first access of tile
+    BytePointer pointer_scale_;
+    BytePointer pointer_zero_;
+
+    bool is_valid_ = false;
+
+public:
+    /// Constructs a TileIterator from its precomputed state, threadblock offset,
+    /// and thread ID
+    CUTLASS_DEVICE
+    FineGrainedScaleZeroIterator(
+        ///< Precomputed parameters object
+        Params const& params,
+        ///< Pointer to start of scale tensor
+        Pointer pointer_scale,
+        ///< Pointer to start of zero tensor
+        Pointer pointer_zero,
+        ///< Extent of the scale and bias
+        TensorCoord extent,
+        ///< ID of each participating thread
+        int thread_id,
+        ///< Initial offset of threadblock
+        TensorCoord const& threadblock_offset,
+        ///< Group size
+        int group_size)
+        : params_(params)
+        , pointer_scale_(reinterpret_cast<BytePointer>(const_cast<NonConstPointer>(pointer_scale)))
+        , pointer_zero_(reinterpret_cast<BytePointer>(const_cast<NonConstPointer>(pointer_zero)))
+    {
+        row_groupsize64_ = threadblock_offset.row();
+        group_size_ = group_size;
+
+        const LongIndex tb_row_byte_offset
+            = threadblock_offset.row() / (group_size / 64) * params_.stride_ * sizeof_bits<Element>::value / 8;
+        const LongIndex tb_col_byte_offset = threadblock_offset.column() * sizeof_bits<Element>::value / 8;
+        pointer_scale_ += (tb_row_byte_offset + tb_col_byte_offset);
+
+        if (pointer_zero_ != nullptr)
+        {
+            pointer_zero_ += (tb_row_byte_offset + tb_col_byte_offset);
+        }
+
+        static constexpr int THREADS_PER_ROW = Shape::kColumn / kAlignment;
+
+        int const thread_row = thread_id / THREADS_PER_ROW;
+        int const thread_col = thread_id % THREADS_PER_ROW;
+
+        const LongIndex thread_row_byte_offset = thread_row * params_.stride_ * sizeof_bits<Element>::value / 8;
+        const LongIndex thread_col_byte_offset = thread_col * kAlignment * sizeof_bits<Element>::value / 8;
+        pointer_scale_ += (thread_row_byte_offset + thread_col_byte_offset);
+        if (pointer_zero_ != nullptr)
+        {
+            pointer_zero_ += (thread_row_byte_offset + thread_col_byte_offset);
+        }
+
+        // For the rows, we must check that we are within the extent AND the tile to avoid extra reads on
+        // a given iteration. The same threads will be responsible for issues reads since the number of scales
+        // read in a given iteration is a constant. Therefore, we should never have to update is_valid_
+        // outside of the constructor.
+        int const global_row = threadblock_offset.row() + thread_row;
+        int const global_col = threadblock_offset.column() + thread_col * kAlignment;
+
+        bool const row_in_bounds = global_row < extent.row() && thread_row < Shape::kRow;
+        bool const col_in_bounds = global_col < extent.column();
+
+        is_valid_ = row_in_bounds && col_in_bounds;
+    }
+
+    /// Construct a PredicatedTileAccessIterator with zero threadblock offset
+    CUTLASS_HOST_DEVICE FineGrainedScaleZeroIterator(Params const& params, ///< Precomputed parameters object
+        Pointer pointer_scale,                                             ///< Pointer to start of scale tensor
+        Pointer pointer_zero,                                              ///< Pointer to start of zero tensor
+        TensorCoord extent,                                                ///< Extent of tensor
+        int thread_id,                                                     ///< ID of each participating thread
+        int group_size)
+        : FineGrainedScaleZeroIterator(
+            params, pointer_scale, pointer_zero, extent, thread_id, make_Coord(0, 0), group_size)
+    {
+    }
+
+    CUTLASS_DEVICE
+    void add_tile_offset(TensorCoord const& tile_offset)
+    {
+        const LongIndex row_byte_offset = tile_offset.row() * params_.inc_advance_;
+        const LongIndex col_byte_offset = tile_offset.column() * Shape::kColumn * sizeof_bits<Element>::value / 8;
+        pointer_scale_ += row_byte_offset + col_byte_offset;
+        if (pointer_zero_ != nullptr)
+        {
+            pointer_zero_ += row_byte_offset + col_byte_offset;
+        }
+    }
+
+    /// Clears the predicate set efficiently
+    CUTLASS_HOST_DEVICE void clear_mask(bool enable = true)
+    {
+        is_valid_ &= (!enable);
+    }
+
+    /// Returns whether access is valid or not
+    CUTLASS_HOST_DEVICE
+    bool valid() const
+    {
+        return is_valid_;
+    }
+
+    /// Returns a scale pointer
+    CUTLASS_HOST_DEVICE
+    AccessType* get_scale() const
+    {
+        return reinterpret_cast<AccessType*>(pointer_scale_);
+    }
+
+    /// Returns a zero pointer
+    CUTLASS_HOST_DEVICE
+    AccessType* get_zero() const
+    {
+        return reinterpret_cast<AccessType*>(pointer_zero_);
+    }
+};
+
+} // namespace threadblock
+} // namespace transform
+} // namespace cutlass

sgl-kernel/3rdparty/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/util/gather_tensor.hpp

+/***************************************************************************************************
+ * Copyright (c) 2023 - 2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-License-Identifier: BSD-3-Clause
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+ * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ **************************************************************************************************/
+#pragma once
+
+#include "cute/layout.hpp"
+#include "cute/tensor.hpp"
+#include "cute/util/print.hpp"
+
+using namespace cute;
+
+/// Function object that applies an index to its argument
+template <class Iter>
+struct IndexedGather
+{
+    CUTE_HOST_DEVICE constexpr IndexedGather(Iter indices = {})
+        : indices_(indices)
+    {
+    }
+
+    template <typename I>
+    CUTE_HOST_DEVICE constexpr auto operator()(I i) const
+    {
+        return indices_[i];
+    }
+
+    CUTE_HOST_DEVICE friend void print(IndexedGather const& s)
+    {
+        cute::print("Indexed{");
+        print(s.indices_);
+        print("}");
+    }
+
+    Iter indices_;
+};
+
+/// Custom stride object that applies a function followed by a stride
+template <class Func, class Stride>
+struct CustomStride
+{
+    CUTE_HOST_DEVICE constexpr CustomStride(Func const& func, Stride const& stride)
+        : func_(func)
+        , stride_(stride)
+    {
+    }
+
+    template <class I>
+    CUTE_HOST_DEVICE constexpr friend auto operator*(I i, CustomStride const& s)
+    {
+        return s.func_(i) * s.stride_;
+    }
+
+    template <class I>
+    CUTE_HOST_DEVICE constexpr friend auto operator*(CustomStride const& s, I i)
+    {
+        return s.func_(i) * s.stride_;
+    }
+
+    CUTE_HOST_DEVICE friend void print(CustomStride const& s)
+    {
+        cute::print("Custom{");
+        print(s.func_);
+        cute::print(",");
+        print(s.stride_);
+        cute::print("}");
+    }
+
+    template <class Div>
+    CUTE_HOST_DEVICE constexpr friend auto safe_div(CustomStride const& s, Div const& div)
+    {
+        return CustomStride<Func, decltype(safe_div(s.stride_, div))>(s.func_, safe_div(s.stride_, div));
+    }
+
+    // Circumvent the requirement on make_layout that shape and stride are integral
+    template <class Shape>
+    CUTE_HOST_DEVICE constexpr friend auto make_layout(Shape const& shape, CustomStride const& stride)
+    {
+        return Layout<Shape, CustomStride>(shape, stride);
+    }
+
+    Func func_;
+    Stride stride_;
+};
+
+template <class Stride, class Func>
+CUTLASS_HOST_DEVICE auto make_custom_stride_layout(Stride const& stride, Func&& func)
+{
+    // Use a dummy shape and replace the first non-unit and non-zero stride with a custom gather stride
+    auto idx = find_if(stride, [](auto x) { return !is_constant<1, decltype(x)>{} && !is_constant<0, decltype(x)>{}; });
+    constexpr int I = decltype(idx)::value;
+    return make_layout(
+        repeat_like(stride, _1{}), replace<I>(stride, CustomStride{static_cast<Func&&>(func), get<I>(stride)}));
+}
+
+/// Helper function to optionally create a gather tensor
+template <class Iterator, class Shape, class Stride, class Func>
+CUTLASS_HOST_DEVICE auto make_gather_tensor(Iterator iter, Shape const& shape, Stride const& stride, Func&& func)
+{
+    Layout matrix_layout = make_identity_layout(shape);
+    auto offset = as_arithmetic_tuple(repeat_like(shape, _0{}));
+    Layout gather_layout = make_custom_stride_layout(stride, static_cast<Func&&>(func));
+    return make_tensor(iter, ComposedLayout{gather_layout, offset, matrix_layout});
+}
+
+namespace cute
+{
+
+template <int N, int I, class Shape, class Stride>
+CUTE_HOST_DEVICE constexpr auto upcast(Shape const& shape, Stride const& stride)
+{
+    if constexpr (is_tuple<Shape>::value)
+    {
+        return transform_layout(shape, stride, [](auto const& s, auto const& d) { return upcast<N, I>(s, d); });
+    }
+    else if constexpr (is_scaled_basis<Stride>::value)
+    {
+        if constexpr (Stride::mode() == I)
+        {
+            return make_layout(shape_div(shape, Int<N>{}), shape_div(stride, Int<N>{}));
+        }
+        else
+        {
+            return make_layout(shape, stride);
+        }
+    }
+    else
+    {
+        return upcast<N>(shape, stride);
+    }
+
+    CUTE_GCC_UNREACHABLE;
+}
+
+template <int N, class OuterShape, class OuterStride, class Offset, class Shape, class Stride>
+CUTE_HOST_DEVICE constexpr auto upcast(
+    ComposedLayout<Layout<OuterShape, OuterStride>, Offset, Layout<Shape, Stride>> const& layout)
+{
+    // Find index of the stride-1 mode - that is the only one that requires updating inner shape and offset
+    auto idx = find_if(layout.layout_a().stride(), [](auto x) { return is_constant<1, decltype(x)>{}; });
+    constexpr int I = decltype(idx)::value;
+
+    // Upcast the outer layout (works as expected)
+    auto outer = upcast<N>(layout.layout_a());
+
+    // Upcast the accumulated offset along stride-1 mode
+    auto offset = as_arithmetic_tuple(replace<I>(layout.offset(), upcast<N>(get<I>(layout.offset()))));
+
+    // Upcast the inner layout's shape along stride-1 mode
+    auto inner = upcast<N, I>(layout.layout_b().shape(), layout.layout_b().stride());
+
+    return composition(outer, offset, inner);
+}
+
+} // namespace cute

sgl-kernel/3rdparty/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/weight_only_quant_op.h

+/***************************************************************************************************
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-License-Identifier: BSD-3-Clause
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+ * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ **************************************************************************************************/
+/*! \file
+  \brief Defines iterators used by warp-level matrix multiply operations targeting Tensor Cores.
+*/
+
+#pragma once
+
+namespace cutlass
+{
+
+enum class WeightOnlyQuantOp
+{
+    UNDEFINED,
+    PER_COLUMN_SCALE_ONLY,
+    FINEGRAINED_SCALE_ONLY,
+    FINEGRAINED_SCALE_AND_ZEROS
+};
+
+constexpr bool isFinegrained(WeightOnlyQuantOp op)
+{
+    return op == WeightOnlyQuantOp::FINEGRAINED_SCALE_AND_ZEROS || op == WeightOnlyQuantOp::FINEGRAINED_SCALE_ONLY;
+}
+
+constexpr bool hasZero(WeightOnlyQuantOp op)
+{
+    return op == WeightOnlyQuantOp::FINEGRAINED_SCALE_AND_ZEROS;
+}
+
+} // namespace cutlass

sgl-kernel/THIRDPARTYNOTICES.txt

@@ -223,3 +223,208 @@ BSD 3-Clause "New" License
 
 3rdparty/cutlass
 include/flashinfer/attention/hopper/block_sparse_gather.cuh
+
+Notice for NVIDIA/TensorRT-LLM
+-------------------------------
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

sgl-kernel/setup.py

@@ -39,6 +39,8 @@ cutlass_default = root / "3rdparty" / "cutlass"
 cutlass = Path(os.environ.get("CUSTOM_CUTLASS_SRC_DIR", default=cutlass_default))
 flashinfer = root / "3rdparty" / "flashinfer"
 turbomind = root / "3rdparty" / "turbomind"
+tensorrt_llm_parent = root / "3rdparty"
+tensorrt_llm = root / "3rdparty" / "tensorrt_llm"
 include_dirs = [
     cutlass.resolve() / "include",
     cutlass.resolve() / "tools" / "util" / "include",
@@ -51,6 +53,8 @@ include_dirs = [
     "cublasLt",
     turbomind.resolve(),
     turbomind.resolve() / "src",
+    tensorrt_llm_parent.resolve(),
+    tensorrt_llm.resolve() / "cutlass_extensions" / "include",
 ]
 
 nvcc_flags = [