Merge pull request #40 from casper-hansen/new_kernel

[NEW] GEMV kernel implementation

awq_cuda/attention/decoder_masked_multihead_attention.h

+// Downloaded from from FasterTransformer v5.2.1
+// https://github.com/NVIDIA/FasterTransformer/blob/release/v5.2.1_tag/src/fastertransformer/kernels/decoder_masked_multihead_attention.h
+/*
+ * Copyright (c) 2020-2022, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * 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.
+ */
+
+#pragma once
+
+#include "cuda_bf16_wrapper.h"
+#include <cuda_fp16.h>
+#include <cuda_runtime_api.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#define CHECK_CUDA(call)                                                                                               \
+    do {                                                                                                               \
+        cudaError_t status_ = call;                                                                                    \
+        if (status_ != cudaSuccess) {                                                                                  \
+            fprintf(stderr, "CUDA error (%s:%d): %s\n", __FILE__, __LINE__, cudaGetErrorString(status_));              \
+            exit(1);                                                                                                   \
+        }                                                                                                              \
+    } while (0)
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+// The structure of parameters for the masked multihead attention kernel.
+//
+// We use the following terminology to describe the different dimensions.
+//
+// B:  Batch size (number of sequences),
+// L:  Sequence length,
+// D:  Hidden dimension,
+// H:  Number of heads,
+// Dh: Hidden dimension per head - Dh = D / H.
+
+template<typename T>
+struct Multihead_attention_params_base {
+
+    // The output buffer. Dimensions B x D.
+    T* out = nullptr;
+
+    // The input Qs and the associated bias. Dimensions B x D and D, resp.
+    const T *q = nullptr, *q_bias = nullptr;
+    // The input Ks and the associated bias. Dimensions B x D and D, resp.
+    const T *k = nullptr, *k_bias = nullptr;
+    // The input Vs and the associated bias. Dimensions B x D and D, resp.
+    const T *v = nullptr, *v_bias = nullptr;
+
+    // The cache for the Ks. The size must be at least B x L x D.
+    T* k_cache = nullptr;
+    // The cache for the Vs. The size must be at least B x L x D.
+    T* v_cache = nullptr;
+    // The indirections to use for cache when beam sampling.
+    const int* cache_indir = nullptr;
+
+    // Stride to handle the case when KQV is a single buffer
+    int stride = 0;
+
+    // The batch size.
+    int batch_size = 0;
+    // The beam width
+    int beam_width = 0;
+    // The sequence length.
+    int memory_max_len = 0;
+    // The number of heads (H).
+    int num_heads = 0;
+    // The number of heads for KV cache.
+    int num_kv_heads = 0;
+    // The hidden dimension per head (Dh).
+    int hidden_size_per_head = 0;
+    // The per-head latent space reserved for rotary embeddings.
+    int  rotary_embedding_dim = 0;
+    bool neox_rotary_style    = false;
+    float rotary_base = 0.0f;
+    // The maximum length of input sentences.
+    int max_input_length = 0;
+    // The current timestep. TODO(bhsueh) Check that do we only this param in cross attention?
+    int timestep = 0;
+    // The current timestep of each sentences (support different timestep for different sentences)
+
+    // The 1.f / sqrt(Dh). Computed on the host.
+    float inv_sqrt_dh = 0.0f;
+
+    // Used when we have some input context like gpt
+    const int* total_padding_tokens = nullptr;
+
+    const bool* masked_tokens            = nullptr;
+    const int*  prefix_prompt_lengths    = nullptr;
+    int         max_prefix_prompt_length = 0;
+
+    const T* relative_attention_bias        = nullptr;
+    int      relative_attention_bias_stride = 0;
+    // The slope per head of linear position bias to attention score (H).
+    const float* linear_bias_slopes = nullptr;
+
+    const T*   ia3_key_weights   = nullptr;
+    const T*   ia3_value_weights = nullptr;
+    const int* ia3_tasks         = nullptr;
+
+    const float* qkv_scale_out       = nullptr;
+    const float* attention_out_scale = nullptr;
+    int          int8_mode           = 0;
+};
+
+template<typename T, bool CROSS_ATTENTION>
+struct Multihead_attention_params: public Multihead_attention_params_base<T> {
+    // output cross attentions
+    float* cross_attention_out        = nullptr;
+    int    max_decoder_seq_len        = 0;
+    bool   is_return_cross_attentions = false;
+
+    // allows to exist attention eary
+    bool* finished = nullptr;
+
+    // required in case of cross attention
+    // will need it here till if constexpr in c++17
+    int* memory_length_per_sample = nullptr;
+
+    // required in case of masked attention with different length
+    const int* length_per_sample = nullptr;
+};
+
+template<typename T>
+struct Multihead_attention_params<T, true>: public Multihead_attention_params_base<T> {
+    // output cross attentions
+    float* cross_attention_out        = nullptr;
+    int    max_decoder_seq_len        = 0;
+    bool   is_return_cross_attentions = false;
+
+    // allows to exist attention eary
+    bool* finished = nullptr;
+
+    // required in case of cross attention
+    int* memory_length_per_sample = nullptr;
+
+    // required in case of masked attention with different length
+    const int* length_per_sample = nullptr;
+};
+
+template<class T>
+using Masked_multihead_attention_params = Multihead_attention_params<T, false>;
+
+template<class T>
+using Cross_multihead_attention_params = Multihead_attention_params<T, true>;
+
+template<typename T>
+struct outputCrossAttentionParam {
+    // max decoder output length
+    int  max_decoder_seq_len        = 0;
+    T*   cross_attention_out        = nullptr;
+    bool is_return_cross_attentions = false;
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+void masked_multihead_attention(const Masked_multihead_attention_params<float>& params, const cudaStream_t& stream);
+void masked_multihead_attention(const Masked_multihead_attention_params<uint16_t>& params, const cudaStream_t& stream);
+#ifdef ENABLE_BF16
+void masked_multihead_attention(const Masked_multihead_attention_params<__nv_bfloat16>& params,
+                                const cudaStream_t&                                     stream);
+#endif
+void cross_multihead_attention(const Cross_multihead_attention_params<float>& params, const cudaStream_t& stream);
+void cross_multihead_attention(const Cross_multihead_attention_params<uint16_t>& params, const cudaStream_t& stream);
+#ifdef ENABLE_BF16
+void cross_multihead_attention(const Cross_multihead_attention_params<__nv_bfloat16>& params,
+                               const cudaStream_t&                                    stream);
+#endif
+
+////////////////////////////////////////////////////////////////////////////////////////////////////

awq_cuda/attention/decoder_masked_multihead_attention_template.hpp

+// Downloaded from from FasterTransformer v5.2.1
+// https://github.com/NVIDIA/FasterTransformer/blob/release/v5.2.1_tag/src/fastertransformer/kernels/decoder_masked_multihead_attention/decoder_masked_multihead_attention_template.hpp
+/*
+ * Copyright (c) 2020-2022, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * 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.
+ */
+#pragma once
+
+#include "decoder_masked_multihead_attention.h"
+#include "decoder_masked_multihead_attention_utils.h"
+#include "cuda_bf16_wrapper.h"
+#include "cuda_bf16_fallbacks.cuh"
+#include <assert.h>
+#include <float.h>
+#include <type_traits>
+
+// #define MMHA_USE_HMMA_FOR_REDUCTION
+
+// Below are knobs to extend FP32 accumulation for higher FP16 accuracy
+
+// Does not seem to affect the accuracy that much
+#define MMHA_USE_FP32_ACUM_FOR_FMA
+
+// Seems to slightly improve the accuracy
+#define MMHA_USE_FP32_ACUM_FOR_OUT
+
+#if 0 && defined(MMHA_USE_FP32_ACUM_FOR_OUT)
+ // Does not seem to improve the accuracy
+ //#define MMHA_USE_FP32_ACUM_FOR_LOGITS
+#endif
+
+namespace mmha {
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+//
+// We use the following terminology to describe the different dimensions.
+//
+// B:  Batch size (number of sequences),
+// L:  Sequence length,
+// D:  Hidden dimension,
+// H:  Number of heads,
+// Dh: Hidden dimension per head - Dh = D / H.
+//
+// The different kernels assign a threadblock for B x H pair. The grid has size (1, B, H). We use
+// 64, 128 and 256 threads per block.
+//
+// Each threadblock loads Dh values from Q and its associated bias. The kernels run a loop to
+// compute Q * K^T where K is loaded from a cache buffer -- except for the current timestep. The
+// cache buffer helps with memory accesses and contains keys with bias.
+//
+// The layout of the cache buffer for the keys is [B, H, Dh/x, L, x] where x == 8 for FP16 and
+// x == 4 for FP32 where the fastest moving dimension (contiguous data) is the rightmost one. The
+// values for x are chosen to create chunks of 16 bytes.
+//
+// The different kernels use 1, 2 or 4 threads per key (THREADS_PER_KEY). The size of the LDGs
+// depends on the number of threads per key. Each thread sums Dh / THREADS_PER_KEY elements. At
+// the end of each iteration of the Q * K^T loop, we perform a reduction between lanes using an
+// HMMA instruction (Tensor Core). Each Q * K^T valuey is stored in shared memory in FP32.
+//
+// After that loop, a parallel softmax is computed across the different Q * K^T values stored in
+// shared memory.
+//
+// The kernel ends with a loop over the values in V. We use THREADS_PER_VALUE to control how many
+// timesteps are computed by loop iteration. As with the keys, the values are read from a cache
+// except for the current timestep. The layout of the cache buffer for the values is much simpler
+// as it is [B, H, L, Dh].
+//
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T, int Dh>
+struct Qk_vec_ {
+};
+
+template<>
+struct Qk_vec_<float, 32> {
+    using Type = float;
+};
+template<>
+struct Qk_vec_<float, 64> {
+    using Type = float2;
+};
+template<>
+struct Qk_vec_<float, 128> {
+    using Type = float4;
+};
+template<>
+struct Qk_vec_<float, 256> {
+    using Type = float4;
+};
+template<>
+struct Qk_vec_<uint16_t, 32> {
+    using Type = uint32_t;
+};
+template<>
+struct Qk_vec_<uint16_t, 64> {
+    using Type = uint32_t;
+};
+template<>
+struct Qk_vec_<uint16_t, 128> {
+    using Type = uint2;
+};
+template<>
+struct Qk_vec_<uint16_t, 256> {
+    using Type = uint4;
+};
+#ifdef ENABLE_BF16
+template<>
+struct Qk_vec_<__nv_bfloat16, 32> {
+    using Type = __nv_bfloat162;
+};
+template<>
+struct Qk_vec_<__nv_bfloat16, 64> {
+    using Type = __nv_bfloat162;
+};
+template<>
+struct Qk_vec_<__nv_bfloat16, 128> {
+    using Type = bf16_4_t;
+};
+template<>
+struct Qk_vec_<__nv_bfloat16, 256> {
+    using Type = bf16_8_t;
+};
+#endif  // ENABLE_BF16
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T, int THREADS_PER_KEY>
+struct K_vec_ {
+};
+
+template<>
+struct K_vec_<float, 4> {
+    using Type = float;
+};
+template<>
+struct K_vec_<float, 2> {
+    using Type = float2;
+};
+template<>
+struct K_vec_<float, 1> {
+    using Type = float4;
+};
+template<>
+struct K_vec_<uint16_t, 4> {
+    using Type = uint32_t;
+};
+template<>
+struct K_vec_<uint16_t, 2> {
+    using Type = uint2;
+};
+template<>
+struct K_vec_<uint16_t, 1> {
+    using Type = uint4;
+};
+#ifdef ENABLE_BF16
+template<>
+struct K_vec_<__nv_bfloat16, 4> {
+    using Type = __nv_bfloat162;
+};
+template<>
+struct K_vec_<__nv_bfloat16, 2> {
+    using Type = bf16_4_t;
+};
+template<>
+struct K_vec_<__nv_bfloat16, 1> {
+    using Type = bf16_8_t;
+};
+#endif  // ENABLE_BF16
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T, int V_VEC_SIZE>
+struct V_vec_ {
+};
+
+template<>
+struct V_vec_<float, 1> {
+    using Type = float;
+};
+template<>
+struct V_vec_<float, 2> {
+    using Type = float2;
+};
+template<>
+struct V_vec_<float, 4> {
+    using Type = float4;
+};
+template<>
+struct V_vec_<uint16_t, 2> {
+    using Type = uint32_t;
+};
+template<>
+struct V_vec_<uint16_t, 4> {
+    using Type = uint2;
+};
+template<>
+struct V_vec_<uint16_t, 8> {
+    using Type = uint4;
+};
+#ifdef ENABLE_BF16
+template<>
+struct V_vec_<__nv_bfloat16, 2> {
+    using Type = __nv_bfloat162;
+};
+template<>
+struct V_vec_<__nv_bfloat16, 4> {
+    using Type = bf16_4_t;
+};
+template<>
+struct V_vec_<__nv_bfloat16, 8> {
+    using Type = bf16_8_t;
+};
+#endif  // ENABLE_BF16
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifdef MMHA_USE_FP32_ACUM_FOR_FMA
+template<typename T>
+struct Qk_vec_acum_fp32_ {
+};
+
+template<>
+struct Qk_vec_acum_fp32_<float> {
+    using Type = float;
+};
+template<>
+struct Qk_vec_acum_fp32_<float2> {
+    using Type = float2;
+};
+template<>
+struct Qk_vec_acum_fp32_<float4> {
+    using Type = float4;
+};
+// template<> struct Qk_vec_acum_fp32_<uint16_t> { using Type = float;        };
+template<>
+struct Qk_vec_acum_fp32_<uint32_t> {
+    using Type = float2;
+};
+template<>
+struct Qk_vec_acum_fp32_<uint2> {
+    using Type = Float4_;
+};
+template<>
+struct Qk_vec_acum_fp32_<uint4> {
+    using Type = Float8_;
+};
+template<>
+struct Qk_vec_acum_fp32_<__nv_bfloat16> {
+    using Type = float;
+};
+template<>
+struct Qk_vec_acum_fp32_<__nv_bfloat162> {
+    using Type = float2;
+};
+template<>
+struct Qk_vec_acum_fp32_<bf16_4_t> {
+    using Type = Float4_;
+};
+template<>
+struct Qk_vec_acum_fp32_<bf16_8_t> {
+    using Type = Float8_;
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T>
+struct K_vec_acum_fp32_ {
+};
+
+template<>
+struct K_vec_acum_fp32_<float> {
+    using Type = float;
+};
+template<>
+struct K_vec_acum_fp32_<float2> {
+    using Type = float2;
+};
+template<>
+struct K_vec_acum_fp32_<float4> {
+    using Type = float4;
+};
+template<>
+struct K_vec_acum_fp32_<uint32_t> {
+    using Type = float2;
+};
+template<>
+struct K_vec_acum_fp32_<uint2> {
+    using Type = Float4_;
+};
+template<>
+struct K_vec_acum_fp32_<uint4> {
+    using Type = Float8_;
+};
+template<>
+struct K_vec_acum_fp32_<__nv_bfloat16> {
+    using Type = float;
+};
+template<>
+struct K_vec_acum_fp32_<__nv_bfloat162> {
+    using Type = float2;
+};
+template<>
+struct K_vec_acum_fp32_<bf16_4_t> {
+    using Type = Float4_;
+};
+template<>
+struct K_vec_acum_fp32_<bf16_8_t> {
+    using Type = Float8_;
+};
+#endif
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifdef MMHA_USE_FP32_ACUM_FOR_OUT
+template<typename T>
+struct V_vec_acum_fp32_ {
+};
+
+template<>
+struct V_vec_acum_fp32_<float> {
+    using Type = float;
+};
+template<>
+struct V_vec_acum_fp32_<float2> {
+    using Type = float2;
+};
+template<>
+struct V_vec_acum_fp32_<float4> {
+    using Type = float4;
+};
+template<>
+struct V_vec_acum_fp32_<uint32_t> {
+    using Type = float2;
+};
+template<>
+struct V_vec_acum_fp32_<uint2> {
+    using Type = Float4_;
+};
+template<>
+struct V_vec_acum_fp32_<uint4> {
+    using Type = Float8_;
+};
+#ifdef ENABLE_BF16
+template<>
+struct V_vec_acum_fp32_<__nv_bfloat162> {
+    using Type = float2;
+};
+template<>
+struct V_vec_acum_fp32_<bf16_4_t> {
+    using Type = Float4_;
+};
+template<>
+struct V_vec_acum_fp32_<bf16_8_t> {
+    using Type = Float8_;
+};
+#endif  // ENABLE_BF16
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<int THREADS_PER_KEY, typename K_vec, int N>
+inline __device__ float qk_dot_(const K_vec (&q)[N], const K_vec (&k)[N])
+{
+#ifdef MMHA_USE_FP32_ACUM_FOR_FMA
+    using K_vec_acum = typename K_vec_acum_fp32_<K_vec>::Type;
+#else
+    using K_vec_acum = K_vec;
+#endif
+    // Compute the parallel products for Q*K^T (treat vector lanes separately).
+    K_vec_acum qk_vec = mul<K_vec_acum, K_vec, K_vec>(q[0], k[0]);
+#pragma unroll
+    for (int ii = 1; ii < N; ++ii) {
+        qk_vec = fma(q[ii], k[ii], qk_vec);
+    }
+
+    // Finalize the reduction across lanes.
+    float qk = sum(qk_vec);
+#pragma unroll
+    for (int mask = THREADS_PER_KEY / 2; mask >= 1; mask /= 2) {
+        qk += __shfl_xor_sync(uint32_t(-1), qk, mask);
+    }
+    return qk;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T, int THREADS_PER_KEY>
+struct Qk_dot {
+    template<typename K_vec, int N>
+    static inline __device__ float dot(const K_vec (&q)[N], const K_vec (&k)[N])
+    {
+        return qk_dot_<THREADS_PER_KEY>(q, k);
+    }
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float4 hmma_fp32(const uint2& a, uint32_t b)
+{
+    float4 c;
+    float  zero = 0.f;
+    asm volatile("mma.sync.aligned.m16n8k8.row.col.f32.f16.f16.f32 \n"
+                 "    {%0, %1, %2, %3}, \n"
+                 "    {%4, %5}, \n"
+                 "    {%6}, \n"
+                 "    {%7, %7, %7, %7}; \n"
+
+                 : "=f"(c.x), "=f"(c.y), "=f"(c.z), "=f"(c.w)
+                 : "r"(a.x) "r"(a.y), "r"(b), "f"(zero));
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<int N>
+inline __device__ float qk_hmma_dot_(const uint32_t (&q)[N], const uint32_t (&k)[N])
+{
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 750
+#ifdef MMHA_USE_FP32_ACUM_FOR_FMA
+    using K_vec_acum = typename K_vec_acum_fp32_<uint32_t>::Type;
+#else
+    using K_vec_acum = uint32_t;
+#endif
+    K_vec_acum qk_vec = mul<K_vec_acum, uint32_t, uint32_t>(q[0], k[0]);
+#pragma unroll
+    for (int ii = 1; ii < N; ++ii) {
+        qk_vec = fma(q[ii], k[ii], qk_vec);
+    }
+#ifdef MMHA_USE_FP32_ACUM_FOR_FMA
+    uint32_t qk_vec_ = float2_to_half2(qk_vec);
+    return hmma_fp32(make_uint2(qk_vec_, 0u), 0x3c003c00u).x;
+#else
+    return hmma_fp32(make_uint2(qk_vec, 0u), 0x3c003c00u).x;
+#endif
+#else
+    return 0.f;
+#endif
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+struct Qk_dot<uint16_t, 4> {
+    template<int N>
+    static inline __device__ float dot(const uint32_t (&q)[N], const uint32_t (&k)[N])
+    {
+#if __CUDA_ARCH__ >= 750 && defined(MMHA_USE_HMMA_FOR_REDUCTION)
+        return qk_hmma_dot_(q, k);
+#else
+        return qk_dot_<4>(q, k);
+#endif  // defined MMHA_USE_HMMA_FOR_REDUCTION
+    }
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<int WARPS_PER_BLOCK, int WARP_SIZE = 32>
+inline __device__ float block_sum(float* red_smem, float sum)
+{
+
+    // Decompose the thread index into warp / lane.
+    int warp = threadIdx.x / WARP_SIZE;
+    int lane = threadIdx.x % WARP_SIZE;
+
+// Compute the sum per warp.
+#pragma unroll
+    for (int mask = WARP_SIZE / 2; mask >= 1; mask /= 2) {
+        sum += __shfl_xor_sync(uint32_t(-1), sum, mask);
+    }
+
+    // Warp leaders store the data to shared memory.
+    if (lane == 0) {
+        red_smem[warp] = sum;
+    }
+
+    // Make sure the data is in shared memory.
+    __syncthreads();
+
+    // The warps compute the final sums.
+    if (lane < WARPS_PER_BLOCK) {
+        sum = red_smem[lane];
+    }
+
+// Parallel reduction inside the warp.
+#pragma unroll
+    for (int mask = WARPS_PER_BLOCK / 2; mask >= 1; mask /= 2) {
+        sum += __shfl_xor_sync(uint32_t(-1), sum, mask);
+    }
+
+    // Broadcast to other threads.
+    return __shfl_sync(uint32_t(-1), sum, 0);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(float& dst, float src)
+{
+    dst = src;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(uint16_t& dst, float src)
+{
+    dst = float_to_half(src);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(uint32_t& dst, float2 src)
+{
+    dst = float2_to_half2(src);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+#ifdef ENABLE_BF16
+inline __device__ void convert_from_float(__nv_bfloat16& dst, float src)
+{
+    dst = __float2bfloat16(src);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(__nv_bfloat162& dst, float2 src)
+{
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 800
+    dst = __float22bfloat162_rn(src);
+#else
+    dst   = __floats2bfloat162_rn(src.x, src.y);
+#endif
+}
+#endif  // ENABLE_BF16
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(uint2& dst, Float4_ src)
+{
+    dst.x = float2_to_half2(src.x);
+    dst.y = float2_to_half2(src.y);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(uint2& dst, float4 src)
+{
+    convert_from_float(dst, Float4_{make_float2(src.x, src.y), make_float2(src.z, src.w)});
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(uint4& dst, Float8_ src)
+{
+    dst.x = float2_to_half2(src.x);
+    dst.y = float2_to_half2(src.y);
+    dst.z = float2_to_half2(src.z);
+    dst.w = float2_to_half2(src.w);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifdef ENABLE_BF16
+inline __device__ void convert_from_float(bf16_4_t& dst, Float4_ src)
+{
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 800
+    dst.x = __float22bfloat162_rn(src.x);
+    dst.y = __float22bfloat162_rn(src.y);
+#else
+    dst.x = __floats2bfloat162_rn(src.x.x, src.x.y);
+    dst.y = __floats2bfloat162_rn(src.y.x, src.y.y);
+#endif
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(bf16_4_t& dst, float4 src)
+{
+    convert_from_float(dst, Float4_{make_float2(src.x, src.y), make_float2(src.z, src.w)});
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(bf16_8_t& dst, Float8_ src)
+{
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 800
+    dst.x = __float22bfloat162_rn(src.x);
+    dst.y = __float22bfloat162_rn(src.y);
+    dst.z = __float22bfloat162_rn(src.z);
+    dst.w = __float22bfloat162_rn(src.w);
+#else
+    dst.x = __floats2bfloat162_rn(src.x.x, src.x.y);
+    dst.y = __floats2bfloat162_rn(src.y.x, src.y.y);
+    dst.z = __floats2bfloat162_rn(src.z.x, src.z.y);
+    dst.w = __floats2bfloat162_rn(src.w.x, src.w.y);
+#endif
+}
+#endif  // ENABLE_BF16
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(float2& dst, float2 src)
+{
+    dst = src;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void convert_from_float(float4& dst, float4 src)
+{
+    dst = src;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float convert_to_float(float4 u)
+{
+    return u.x;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float convert_to_float(uint4 u)
+{
+    float2 tmp = half2_to_float2(u.x);
+    return tmp.x;
+}
+
+#if defined(MMHA_USE_FP32_ACUM_FOR_LOGITS)
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float cast_to_float(float u)
+{
+    return u;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 cast_to_float(float2 u)
+{
+    return u;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float4 cast_to_float(float4 u)
+{
+    return u;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float4_ cast_to_float(Float4_ u)
+{
+    return u;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float8_ cast_to_float(Float8_ u)
+{
+    return u;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 cast_to_float(uint32_t u)
+{
+    return half2_to_float2(u);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float4_ cast_to_float(uint2 u)
+{
+    Float4_ tmp;
+    tmp.x = half2_to_float2(u.x);
+    tmp.y = half2_to_float2(u.y);
+    return tmp;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float8_ cast_to_float(uint4 u)
+{
+    Float8_ tmp;
+    tmp.x = half2_to_float2(u.x);
+    tmp.y = half2_to_float2(u.y);
+    tmp.z = half2_to_float2(u.z);
+    tmp.w = half2_to_float2(u.w);
+    return tmp;
+}
+
+#endif
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float float_from_int8(int8_t u)
+{
+    return u;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 float_from_int8(int16_t u)
+{
+    union {
+        int16_t int16;
+        int8_t  int8[2];
+    };
+    int16 = u;
+    return make_float2(int8[0], int8[1]);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float4 float_from_int8(int32_t u)
+{
+    union {
+        int32_t int32;
+        int8_t  int8[4];
+    };
+    int32 = u;
+    return make_float4(int8[0], int8[1], int8[2], int8[3]);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+// clang-format off
+inline __device__ Float8_ float_from_int8(int64_t u)
+{
+    union {
+        int64_t int64;
+        int16_t int16[4];
+    };
+    int64 = u;
+    return Float8_ {float_from_int8(int16[0]),
+                    float_from_int8(int16[1]),
+                    float_from_int8(int16[2]),
+                    float_from_int8(int16[3])};
+}
+// clang-format on
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ int8_t cast_to_int8(float val)
+{
+    union {
+        int8_t  int8[2];
+        int16_t int16;
+    };
+    asm volatile("cvt.rni.sat.s8.f32 %0, %1;" : "=h"(int16) : "f"(val));
+    return int8[0];
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ int32_t cast_to_int8(float4 val)
+{
+    union {
+        int8_t  int8[4];
+        int32_t int32;
+    };
+    int8[0] = cast_to_int8(val.x);
+    int8[1] = cast_to_int8(val.y);
+    int8[2] = cast_to_int8(val.z);
+    int8[3] = cast_to_int8(val.w);
+    return int32;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ int64_t cast_to_int8(Float8_ val)
+{
+    union {
+        int8_t  int8[8];
+        int64_t int64;
+    };
+    int8[0] = cast_to_int8(val.x.x);
+    int8[1] = cast_to_int8(val.x.y);
+    int8[2] = cast_to_int8(val.y.x);
+    int8[3] = cast_to_int8(val.y.y);
+    int8[4] = cast_to_int8(val.z.x);
+    int8[5] = cast_to_int8(val.z.y);
+    int8[6] = cast_to_int8(val.w.x);
+    int8[7] = cast_to_int8(val.w.y);
+    return int64;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T>
+inline __device__ __host__ T div_up(T m, T n)
+{
+    return (m + n - 1) / n;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T, bool DO_CROSS_ATTENTION>
+inline size_t smem_size_in_bytes(const Multihead_attention_params<T, DO_CROSS_ATTENTION>& params,
+                                 int                                                      threads_per_value,
+                                 int                                                      threads_per_block)
+{
+    // The amount of shared memory needed to store the Q*K^T values in float.
+    const int max_timesteps = min(params.timestep, params.memory_max_len);
+    size_t qk_sz = (DO_CROSS_ATTENTION) ? div_up(params.memory_max_len + 1, 4) * 16 : div_up(max_timesteps + 1, 4) * 16;
+
+    // The extra memory needed if we are not using floats for the final logits.
+    size_t logits_sz = 0;
+#ifndef MMHA_USE_FP32_ACUM_FOR_LOGITS
+    if (sizeof(T) != 4) {
+        // TDOD
+        logits_sz = (DO_CROSS_ATTENTION) ? div_up(params.memory_max_len + 1, 4) * 4 * sizeof(T) :
+                                           div_up(max_timesteps + 1, 4) * 4 * sizeof(T);
+    }
+#endif
+
+    // The total size needed during softmax.
+    size_t softmax_sz = qk_sz + logits_sz;
+
+    // The number of partial rows to reduce in the final reduction.
+    int rows_per_red = threads_per_block / threads_per_value;
+    // The amount of storage needed to finalize the outputs.
+    size_t red_sz = rows_per_red * params.hidden_size_per_head * sizeof(T) / 2;
+
+    size_t transpose_rotary_size = 0;
+    if (params.rotary_embedding_dim > 0 && params.neox_rotary_style) {
+        transpose_rotary_size = 2 * params.rotary_embedding_dim * sizeof(T);
+    }
+
+    // The max.
+    return max(max(softmax_sz, red_sz), transpose_rotary_size);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ constexpr uint32_t shfl_mask(int threads)
+{
+    return threads == 32 ? uint32_t(-1) : (1u << threads) - 1u;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<
+    // The type of the inputs. Supported types: float and half.
+    typename T,
+    // The hidden dimension per head.
+    int Dh,
+    int Dh_MAX,
+    // The number of threads per key.
+    int THREADS_PER_KEY,
+    // The number of threads per value.
+    int THREADS_PER_VALUE,
+    // The number of threads in a threadblock.
+    int  THREADS_PER_BLOCK,
+    bool DO_CROSS_ATTENTION>
+__global__ void masked_multihead_attention_kernel(Multihead_attention_params<T, DO_CROSS_ATTENTION> params)
+{
+
+    // Make sure the hidden dimension per head is a multiple of the number of threads per key.
+    static_assert(Dh_MAX % THREADS_PER_KEY == 0, "");
+    // Make sure the hidden dimension per head is a multiple of the number of threads per value.
+    static_assert(Dh_MAX % THREADS_PER_VALUE == 0, "");
+
+    // The size of a warp.
+    constexpr int WARP_SIZE = 32;
+    // The number of warps in a threadblock.
+    constexpr int WARPS_PER_BLOCK = THREADS_PER_BLOCK / WARP_SIZE;
+
+    // Use smem_size_in_bytes (above) to determine the amount of shared memory.
+    extern __shared__ char smem_[];
+
+    // The shared memory for the Q*K^T values and partial logits in softmax.
+    float* qk_smem = reinterpret_cast<float*>(smem_);
+
+    // The shared memory for the logits. For FP32, that's the same buffer as qk_smem.
+    char* logits_smem_ = smem_;
+#ifndef MMHA_USE_FP32_ACUM_FOR_LOGITS
+    if (sizeof(T) != 4) {
+        // TODO - change to tlength
+        const int max_timesteps = min(params.timestep, params.memory_max_len);
+        logits_smem_ +=
+            (DO_CROSS_ATTENTION) ? div_up(params.memory_max_len + 1, 4) * 16 : div_up(max_timesteps + 1, 4) * 16;
+    }
+    T* logits_smem = reinterpret_cast<T*>(logits_smem_);
+#else
+    float* logits_smem = reinterpret_cast<float*>(logits_smem_);
+#endif
+
+    // The shared memory to do the final reduction for the output values. Reuse qk_smem.
+    T* out_smem = reinterpret_cast<T*>(smem_);
+
+    // The shared memory buffers for the block-wide reductions. One for max, one for sum.
+    __shared__ float red_smem[WARPS_PER_BLOCK * 2];
+
+    // A vector of Q or K elements for the current timestep.
+    using Qk_vec = typename Qk_vec_<T, Dh_MAX>::Type;
+
+    // Use alignment for safely casting the shared buffers as Qk_vec.
+    // Shared memory to store Q inputs.
+    __shared__ __align__(sizeof(Qk_vec)) T q_smem[Dh_MAX];
+
+    // This is one of the reasons we should have a separate kernel for cross attention
+    __shared__ __align__(sizeof(Qk_vec)) T bias_smem[DO_CROSS_ATTENTION ? Dh_MAX : 1];
+
+    // A vector of Q or K elements for the current timestep.
+    using Qk_vec = typename Qk_vec_<T, Dh_MAX>::Type;
+    // The number of elements per vector.
+    constexpr int QK_VEC_SIZE = sizeof(Qk_vec) / sizeof(T);
+    // Make sure the hidden size per head is a multiple of the vector size.
+    static_assert(Dh_MAX % QK_VEC_SIZE == 0, "");
+    // We will use block wide reduction if needed
+    // static_assert(Dh_MAX / QK_VEC_SIZE <= WARP_SIZE, "");
+    // The number of vectors per warp.
+    constexpr int QK_VECS_PER_WARP = Dh_MAX / QK_VEC_SIZE;
+
+    // The layout of the cache is [B, H, Dh/x, L, x] with x == 4/8 for FP32/FP16. Since each thread
+    // owns x elements, we have to decompose the linear index into chunks of x values and the posi-
+    // tion of the thread in that chunk.
+
+    // The number of elements in a chunk of 16B (that's the x in the above formula).
+    constexpr int QK_ELTS_IN_16B = 16 / sizeof(T);
+    // The number of K vectors in 16B.
+    constexpr int QK_VECS_IN_16B = 16 / sizeof(Qk_vec);
+
+    // The batch/beam idx
+    const int bi = blockIdx.y;
+    if (params.finished != nullptr && params.finished[bi] == true) {
+        return;
+    }
+    // The beam idx
+    const int beami = bi % params.beam_width;
+    // The "beam-aware" batch idx
+    const int bbi = bi / params.beam_width;
+    // The head.
+    const int num_kv_heads = params.num_kv_heads;
+    const int kv_rep = (params.num_heads / num_kv_heads);
+    const int hi = blockIdx.x;
+    const int hi_kv = hi / kv_rep;
+    
+    // Combine the batch and the head indices.
+    const int bhi = bi * params.num_heads + hi;
+    const int bhi_kv = bi * (params.num_heads / kv_rep) + hi_kv;
+    // Combine the "beam-aware" batch idx and the head indices.
+    const int bbhi = bbi * params.beam_width * params.num_heads + hi;
+    const int bbhi_kv = bbi * params.beam_width * (params.num_heads / kv_rep) + hi_kv;
+    // The thread in the block.
+    const int tidx = threadIdx.x;
+
+    const bool handle_kv = !DO_CROSS_ATTENTION || (DO_CROSS_ATTENTION && params.timestep == 0);
+    // Every kv_rep threads have the same kv_cache values. So only the first one writes back.
+    const int write_kv_cache = handle_kv && (hi % kv_rep == 0);
+
+    // While doing the product Q*K^T for the different keys we track the max.
+    float qk_max = -FLT_MAX;
+
+    float qk = 0.0F;
+
+    // int qkv_base_offset = (params.stride == 0) ? bhi * Dh : bi * params.stride + hi * Dh;
+    const int q_base_offset = bi * params.stride + hi * Dh;
+    const int k_base_offset = bi * params.stride + hi_kv * Dh;
+    const int v_base_offset = k_base_offset;
+
+    const size_t bi_seq_len_offset = bi * params.memory_max_len;
+
+    // int tlength = (DO_CROSS_ATTENTION)? params.memory_length_per_sample[bi] - 1 : params.timestep;
+    int       tlength      = (DO_CROSS_ATTENTION) ? params.memory_length_per_sample[bi] - 1 :
+                             (params.length_per_sample == nullptr) ?
+                                                    params.timestep :
+                                                    params.length_per_sample[bi] + params.max_prefix_prompt_length;
+    const int first_step   = max(0, tlength + 1 - params.memory_max_len);
+    const int tlength_circ = tlength % params.memory_max_len;
+
+    // First QK_VECS_PER_WARP load Q and K + the bias values for the current timestep.
+    const bool is_masked = tidx >= QK_VECS_PER_WARP;
+
+    // The offset in the Q and K buffer also accounts for the batch.
+    // int qk_offset = qkv_base_offset + tidx * QK_VEC_SIZE;
+    int q_offset = q_base_offset + tidx * QK_VEC_SIZE;
+    int k_offset = k_base_offset + tidx * QK_VEC_SIZE;
+    int v_offset = k_offset;
+
+    // The offset in the bias buffer.
+    // int qk_bias_offset = hi * Dh + tidx * QK_VEC_SIZE;
+    int q_bias_offset = hi * Dh + tidx * QK_VEC_SIZE;
+    int k_bias_offset = hi_kv * Dh + tidx * QK_VEC_SIZE;
+    int v_bias_offset = k_bias_offset;
+
+    const bool do_ia3      = handle_kv && params.ia3_tasks != nullptr;
+    const int  ia3_task_id = do_ia3 ? params.ia3_tasks[bbi] : 0;
+
+    // Trigger the loads from the Q and K buffers.
+    Qk_vec q;
+    zero(q);
+    if (!is_masked && (Dh == Dh_MAX || tidx * QK_VEC_SIZE < Dh)) {
+        if (params.int8_mode == 2) {
+            using Packed_Int8_t  = typename packed_type<int8_t, num_elems<Qk_vec>::value>::type;
+            using Packed_Float_t = typename packed_type<float, num_elems<Qk_vec>::value>::type;
+            const auto q_scaling = params.qkv_scale_out[0];
+            const auto q_quant =
+                *reinterpret_cast<const Packed_Int8_t*>(&reinterpret_cast<const int8_t*>(params.q)[q_offset]);
+
+            convert_from_float(q, mul<Packed_Float_t, float>(q_scaling, float_from_int8(q_quant)));
+        }
+        else {
+            q = *reinterpret_cast<const Qk_vec*>(&params.q[q_offset]);
+        }
+    }
+
+    Qk_vec k;
+    zero(k);
+    if (DO_CROSS_ATTENTION) {
+        // The 16B chunk written by the thread.
+        int co = tidx / QK_VECS_IN_16B;
+        // The position of the thread in that 16B chunk.
+        int ci = tidx % QK_VECS_IN_16B * QK_VEC_SIZE;
+
+        // Two chunks are separated by L * x elements. A thread write QK_VEC_SIZE elements.
+        int offset = bhi_kv * params.memory_max_len * Dh + co * params.memory_max_len * QK_ELTS_IN_16B +
+                     // params.timestep*QK_ELTS_IN_16B +
+                     tlength * QK_ELTS_IN_16B + ci;
+        k = !is_masked && (Dh == Dh_MAX || tidx * QK_VEC_SIZE < Dh) ?
+                *reinterpret_cast<const Qk_vec*>(&params.k_cache[offset]) :
+                k;
+    }
+    else {
+        if (!is_masked && (Dh == Dh_MAX || tidx * QK_VEC_SIZE < Dh)) {
+            if (params.int8_mode == 2) {
+                using Packed_Int8_t  = typename packed_type<int8_t, num_elems<Qk_vec>::value>::type;
+                using Packed_Float_t = typename packed_type<float, num_elems<Qk_vec>::value>::type;
+                const auto k_scaling = params.qkv_scale_out[1];
+                const auto k_quant =
+                    *reinterpret_cast<const Packed_Int8_t*>(&reinterpret_cast<const int8_t*>(params.k)[k_offset]);
+
+                convert_from_float(k, mul<Packed_Float_t, float>(k_scaling, float_from_int8(k_quant)));
+            }
+            else {
+                k = *reinterpret_cast<const Qk_vec*>(&params.k[k_offset]);
+            }
+        }
+    }
+
+    // Trigger the loads from the Q and K bias buffers.
+    Qk_vec q_bias;
+    zero(q_bias);
+    q_bias = (!is_masked && Dh == Dh_MAX || tidx * QK_VEC_SIZE < Dh) && params.q_bias != nullptr ?
+                 *reinterpret_cast<const Qk_vec*>(&params.q_bias[q_bias_offset]) :
+                 q_bias;
+
+    Qk_vec k_bias;
+    zero(k_bias);
+    if (handle_kv) {
+        k_bias = !is_masked && (Dh == Dh_MAX || tidx * QK_VEC_SIZE < Dh) && params.k_bias != nullptr ?
+                     *reinterpret_cast<const Qk_vec*>(&params.k_bias[k_bias_offset]) :
+                     k_bias;
+    }
+
+    // Computes the Q/K values with bias.
+    q = add(q, q_bias);
+    if (handle_kv) {
+        k = add(k, k_bias);
+    }
+    if (do_ia3 && !is_masked) {
+        k = mul<Qk_vec, Qk_vec, Qk_vec>(
+            k,
+            *reinterpret_cast<const Qk_vec*>(
+                &params.ia3_key_weights[(ia3_task_id * params.num_heads + hi) * Dh + tidx * QK_VEC_SIZE]));
+    }
+
+    // Padded len
+    const int padd_len = (params.total_padding_tokens == nullptr) ? 0 : params.total_padding_tokens[bi];
+    if (params.rotary_embedding_dim > 0 && !params.neox_rotary_style) {
+        if (handle_kv) {
+            apply_rotary_embedding(q, k, tidx, params.rotary_embedding_dim, tlength - padd_len, params.rotary_base);
+        }
+        else {
+            apply_rotary_embedding(q, tidx, params.rotary_embedding_dim, tlength - padd_len, params.rotary_base);
+        }
+    }
+    else if (params.rotary_embedding_dim > 0 && params.neox_rotary_style) {
+        const bool do_rotary = !is_masked && QK_VEC_SIZE * tidx < params.rotary_embedding_dim;
+
+        T* q_smem = reinterpret_cast<T*>(smem_);
+        T* k_smem = q_smem + params.rotary_embedding_dim;
+
+        const int half_rotary_dim = params.rotary_embedding_dim / 2;
+        const int half_idx        = (tidx * QK_VEC_SIZE) / half_rotary_dim;
+        const int intra_half_idx  = (tidx * QK_VEC_SIZE) % half_rotary_dim;
+        const int smem_pitch      = half_rotary_dim;  // TODO: adjust for bank conflicts
+
+        assert(half_rotary_dim % QK_VEC_SIZE == 0);
+
+        if (do_rotary) {
+            *reinterpret_cast<Qk_vec*>(q_smem + half_idx * smem_pitch + intra_half_idx) = q;
+
+            if (handle_kv) {
+                *reinterpret_cast<Qk_vec*>(k_smem + half_idx * smem_pitch + intra_half_idx) = k;
+            }
+        }
+
+        __syncthreads();
+
+        const int     transpose_idx = half_idx * (half_rotary_dim / 2) + intra_half_idx / 2;
+        constexpr int tidx_factor   = (QK_VEC_SIZE > 1) ? QK_VEC_SIZE / 2 : 1;
+        if (do_rotary) {
+            mmha::vec_from_smem_transpose(q, q_smem, transpose_idx, smem_pitch);
+
+            if (handle_kv) {
+                mmha::vec_from_smem_transpose(k, k_smem, transpose_idx, smem_pitch);
+
+                mmha::apply_rotary_embedding(
+                    q, k, transpose_idx / tidx_factor, params.rotary_embedding_dim, tlength - padd_len, params.rotary_base);
+
+                mmha::write_smem_transpose(k, k_smem, transpose_idx, smem_pitch);
+            }
+            else {
+                mmha::apply_rotary_embedding(
+                    q, transpose_idx / tidx_factor, params.rotary_embedding_dim, tlength, params.rotary_base);
+            }
+            mmha::write_smem_transpose(q, q_smem, transpose_idx, smem_pitch);
+        }
+
+        __syncthreads();
+
+        if (do_rotary) {
+            q = *reinterpret_cast<Qk_vec*>(q_smem + half_idx * smem_pitch + intra_half_idx);
+            if (handle_kv) {
+                k = *reinterpret_cast<Qk_vec*>(k_smem + half_idx * smem_pitch + intra_half_idx);
+            }
+        }
+
+        __syncthreads();
+    }
+
+    if (!is_masked) {
+        // Store the Q values to shared memory.
+        *reinterpret_cast<Qk_vec*>(&q_smem[tidx * QK_VEC_SIZE]) = q;
+
+        // Store Dh values of k_bias into smem, since will need to add later
+        // if params.timestep == 0
+        if (DO_CROSS_ATTENTION && params.timestep == 0) {
+            *reinterpret_cast<Qk_vec*>(&bias_smem[tidx * QK_VEC_SIZE]) = k_bias;
+        }
+
+        // Write the K values to the global memory cache.
+        //
+        // NOTE: The stores are uncoalesced as we have multiple chunks of 16B spread across the memory
+        // system. We designed it this way as it allows much better memory loads (and there are many
+        // more loads) + the stores are really "write and forget" since we won't need the ack before
+        // the end of the kernel. There's plenty of time for the transactions to complete.
+
+        // The 16B chunk written by the thread.
+        int co = tidx / QK_VECS_IN_16B;
+        // The position of the thread in that 16B chunk.
+        int ci = tidx % QK_VECS_IN_16B * QK_VEC_SIZE;
+
+        // Two chunks are separated by L * x elements. A thread write QK_VEC_SIZE elements.
+        int offset = bhi_kv * params.memory_max_len * Dh + co * params.memory_max_len * QK_ELTS_IN_16B +
+                     // params.timestep*QK_ELTS_IN_16B +
+                     tlength_circ * QK_ELTS_IN_16B + ci;
+
+        if (write_kv_cache) {
+            // Trigger the stores to global memory.
+            if (Dh == Dh_MAX || co < Dh / QK_ELTS_IN_16B) {
+                *reinterpret_cast<Qk_vec*>(&params.k_cache[offset]) = k;
+            }
+        }
+
+        // Compute \sum_i Q[i] * K^T[i] for the current timestep.
+#ifdef MMHA_USE_FP32_ACUM_FOR_FMA
+        using Qk_vec_acum = typename Qk_vec_acum_fp32_<Qk_vec>::Type;
+#else
+        using Qk_vec_acum = Qk_vec;
+#endif
+        qk = dot<Qk_vec_acum, Qk_vec>(q, k);
+        if (QK_VECS_PER_WARP <= WARP_SIZE) {
+#pragma unroll
+            for (int mask = QK_VECS_PER_WARP / 2; mask >= 1; mask /= 2) {
+                qk += __shfl_xor_sync(shfl_mask(QK_VECS_PER_WARP), qk, mask);
+            }
+        }
+    }
+
+    if (QK_VECS_PER_WARP > WARP_SIZE) {
+        constexpr int WARPS_PER_RED = (QK_VECS_PER_WARP + WARP_SIZE - 1) / WARP_SIZE;
+        qk                          = block_sum<WARPS_PER_RED>(&red_smem[WARPS_PER_RED], qk);
+    }
+
+    // Store that value in shared memory. Keep the Q*K^T value in register for softmax.
+    if (tidx == 0) {
+        // Normalize qk.
+        qk *= params.inv_sqrt_dh;
+        if (params.relative_attention_bias != nullptr) {
+            // TODO (Haotian): check whether we should replace hi with hi_kv,
+            // although params.relative_attention_bias is usually not used.
+            qk = add(qk,
+                     params.relative_attention_bias[hi * params.relative_attention_bias_stride
+                                                        * params.relative_attention_bias_stride
+                                                    + (tlength - padd_len) * params.relative_attention_bias_stride
+                                                    + (tlength - padd_len)]);
+        }
+        // Add alibi positional encoding
+        // qk += (alibi_slope != 0) ? alibi_slope * (params.timestep - params.memory_max_len) : 0;
+        // We don't need to apply the linear position bias here since qi - ki = 0 yields the position bias 0.
+
+        qk_max                        = qk;
+        qk_smem[tlength - first_step] = qk;
+        // qk_smem[params.timestep] = qk;
+    }
+
+    // Make sure the data is in shared memory.
+    __syncthreads();
+
+    // The type of queries and keys for the math in the Q*K^T product.
+    using K_vec = typename K_vec_<T, THREADS_PER_KEY>::Type;
+    // The number of elements per vector.
+    constexpr int K_VEC_SIZE = sizeof(K_vec) / sizeof(T);
+    // Make sure the hidden size per head is a multiple of the vector size.
+    static_assert(Dh_MAX % K_VEC_SIZE == 0, "");
+    // The number of elements per thread.
+    constexpr int K_ELTS_PER_THREAD = Dh_MAX / THREADS_PER_KEY;
+    // The number of vectors per thread.
+    constexpr int K_VECS_PER_THREAD = K_ELTS_PER_THREAD / K_VEC_SIZE;
+
+    // The position the first key loaded by each thread from the cache buffer (for this B * H).
+    int ko = tidx / THREADS_PER_KEY;
+    // The position of the thread in the chunk of keys.
+    int ki = tidx % THREADS_PER_KEY * K_VEC_SIZE;
+
+    static_assert(Dh_MAX == THREADS_PER_KEY * K_VEC_SIZE * K_VECS_PER_THREAD);
+
+    // Load the Q values from shared memory. The values are reused during the loop on K.
+    K_vec q_vec[K_VECS_PER_THREAD];
+#pragma unroll
+    for (int ii = 0; ii < K_VECS_PER_THREAD; ++ii) {
+        q_vec[ii] = *reinterpret_cast<const K_vec*>(&q_smem[ki + ii * THREADS_PER_KEY * K_VEC_SIZE]);
+    }
+
+    K_vec k_bias_vec[DO_CROSS_ATTENTION ? K_VECS_PER_THREAD : 1];
+    if (DO_CROSS_ATTENTION && params.timestep == 0) {
+#pragma unroll
+        for (int ii = 0; ii < K_VECS_PER_THREAD; ++ii) {
+            k_bias_vec[ii] = *reinterpret_cast<const K_vec*>(&bias_smem[ki + ii * THREADS_PER_KEY * K_VEC_SIZE]);
+        }
+    }
+
+    // The number of timesteps loaded per iteration.
+    constexpr int K_PER_ITER = THREADS_PER_BLOCK / THREADS_PER_KEY;
+    // The number of keys per warp.
+    constexpr int K_PER_WARP = WARP_SIZE / THREADS_PER_KEY;
+
+    // The base pointer for the key in the cache buffer.
+    T* k_cache = &params.k_cache[bhi_kv * params.memory_max_len * Dh + ki];
+    // Base pointer for the beam's batch, before offsetting with indirection buffer
+    T* k_cache_batch = &params.k_cache[bbhi_kv * params.memory_max_len * Dh + ki];
+
+    // Pick a number of keys to make sure all the threads of a warp enter (due to shfl_sync).
+    // int ti_end = div_up(params.timestep, K_PER_WARP) * K_PER_WARP;
+    int ti_end = div_up(tlength - first_step, K_PER_WARP) * K_PER_WARP + first_step;
+
+    // prefix prompt length if has
+    const int prefix_prompt_length = (params.prefix_prompt_lengths == nullptr) ? 0 : params.prefix_prompt_lengths[bi];
+
+    // Iterate over the keys/timesteps to compute the various (Q*K^T)_{ti} values.
+    const bool has_beams    = params.cache_indir != nullptr;
+    const int* beam_indices = has_beams ? &params.cache_indir[bi_seq_len_offset] : nullptr;
+
+    for (int ti = first_step + ko; ti < ti_end; ti += K_PER_ITER) {
+        const int ti_circ = ti % params.memory_max_len;
+
+        // The keys loaded from the key cache.
+        K_vec k[K_VECS_PER_THREAD];
+        K_vec k_vec_zero;
+        zero(k_vec_zero);
+#pragma unroll
+        for (int ii = 0; ii < K_VECS_PER_THREAD; ++ii) {
+            int jj = ii * params.memory_max_len + ti_circ;
+            // if( ti < params.timestep ) {
+            const bool within_bounds = (Dh == Dh_MAX || jj * QK_ELTS_IN_16B < Dh * params.memory_max_len);
+            if (ti < tlength) {
+                if (!within_bounds) {
+                    k[ii] = k_vec_zero;
+                }
+                else {
+                    if (has_beams) {
+                        const int beam_offset = beam_indices[ti_circ] * params.num_heads * params.memory_max_len * Dh;
+                        k[ii] = *reinterpret_cast<const K_vec*>(&k_cache_batch[beam_offset + jj * QK_ELTS_IN_16B]);
+                    }
+                    else {
+                        k[ii] = *reinterpret_cast<const K_vec*>(&k_cache_batch[jj * QK_ELTS_IN_16B]);
+                    }
+                }
+                // add bias and update k_cache
+                if (DO_CROSS_ATTENTION && params.timestep == 0) {
+                    k[ii] = add(k[ii], k_bias_vec[ii]);
+
+                    if (do_ia3) {
+                        k[ii] = mul<K_vec, K_vec, K_vec>(
+                            k[ii],
+                            *reinterpret_cast<const K_vec*>(
+                                &params.ia3_key_weights[(ia3_task_id * params.num_heads + hi) * Dh + ki
+                                                        + ii * THREADS_PER_KEY * K_VEC_SIZE]));
+                    }
+
+                    if (Dh == Dh_MAX || jj * QK_ELTS_IN_16B < Dh * params.memory_max_len) {
+                        *reinterpret_cast<K_vec*>(&k_cache[jj * QK_ELTS_IN_16B]) = k[ii];
+                    }
+                }
+            }
+        }
+
+        // Perform the dot product and normalize qk.
+        //
+        // WARNING: ALL THE THREADS OF A WARP MUST ENTER!!!
+        float qk      = Qk_dot<T, THREADS_PER_KEY>::dot(q_vec, k) * params.inv_sqrt_dh;
+        bool  is_mask = (params.masked_tokens != nullptr) && params.masked_tokens[bi_seq_len_offset + ti];
+
+        // Store the product to shared memory. There's one qk value per timestep. Update the max.
+        // if( ti < params.timestep && tidx % THREADS_PER_KEY == 0 ) {
+        if (ti < tlength && tidx % THREADS_PER_KEY == 0) {
+            if (params.relative_attention_bias != nullptr) {
+                qk = add(qk,
+                         params.relative_attention_bias[hi * params.relative_attention_bias_stride
+                                                            * params.relative_attention_bias_stride
+                                                        + tlength * params.relative_attention_bias_stride + ti]);
+            }
+            if (params.linear_bias_slopes != nullptr) {
+                // Apply the linear position bias: (ki - qi) * slope[hi].
+                // The padding token locates between the input context and the generated tokens.
+                // We need to remove the number of padding tokens in the distance computation.
+                //   ti   : 0 1 2 3 4 5 6 7 8 9(tlength)
+                //   token: i i i i p p p o o o where i=input, p=pad, o=output.
+                // e.g. ti = 2, dist = (9 - 3) - 2 = 4.
+                int   max_context_length = params.max_prefix_prompt_length + params.max_input_length;
+                float dist               = (ti < max_context_length ? ti + padd_len : ti) - tlength;
+
+                qk += mul<float, float, float>(params.linear_bias_slopes[hi], dist);
+            }
+            // Add alibi positional encoding
+            // qk += (alibi_slope != 0) ? alibi_slope * (params.timestep - params.memory_max_len) : 0;
+            qk_max                   = is_mask ? qk_max : fmaxf(qk_max, qk);
+            qk_smem[ti - first_step] = qk;
+        }
+    }
+
+// Perform the final reduction to compute the max inside each warp.
+//
+// NOTE: In a group of THREADS_PER_KEY threads, the leader already has the max value for the
+// group so it's not needed to run the reduction inside the group (again).
+#pragma unroll
+    for (int mask = WARP_SIZE / 2; mask >= THREADS_PER_KEY; mask /= 2) {
+        qk_max = fmaxf(qk_max, __shfl_xor_sync(uint32_t(-1), qk_max, mask));
+    }
+
+    // Decompose the thread index into warp and lane.
+    const int warp = tidx / WARP_SIZE;
+    const int lane = tidx % WARP_SIZE;
+
+    // The warp leader writes the max to shared memory.
+    if (lane == 0) {
+        red_smem[warp] = qk_max;
+    }
+
+    // Make sure the products are in shared memory.
+    __syncthreads();
+
+    // The warps finalize the reduction.
+    qk_max = lane < WARPS_PER_BLOCK ? red_smem[lane] : -FLT_MAX;
+#pragma unroll
+    for (int mask = WARPS_PER_BLOCK / 2; mask >= 1; mask /= 2) {
+        qk_max = fmaxf(qk_max, __shfl_xor_sync(uint32_t(-1), qk_max, mask));
+    }
+
+    // Broadcast to all the threads in the warp.
+    qk_max = __shfl_sync(uint32_t(-1), qk_max, 0);
+
+    // Compute the logits and start the sum.
+    float sum = 0.f;
+    // for( int ti = tidx; ti <= params.timestep; ti += THREADS_PER_BLOCK ) {
+    for (int ti = first_step + tidx; ti <= tlength; ti += THREADS_PER_BLOCK) {
+        bool  is_mask = (params.masked_tokens != nullptr) && params.masked_tokens[bi_seq_len_offset + ti];
+        float logit   = is_mask ? 0.f : __expf(qk_smem[ti - first_step] - qk_max);
+        sum += logit;
+        qk_smem[ti - first_step] = logit;
+    }
+
+    // Compute the sum.
+    sum = block_sum<WARPS_PER_BLOCK>(&red_smem[WARPS_PER_BLOCK], sum);
+
+    // Normalize the logits.
+    float inv_sum = __fdividef(1.f, sum + 1.e-6f);
+    // for( int ti = tidx; ti <= params.timestep; ti += THREADS_PER_BLOCK ) {
+    const size_t cross_attention_out_offset =
+        params.is_return_cross_attentions ?
+            bhi_kv * params.max_decoder_seq_len * params.memory_max_len + params.timestep * params.memory_max_len :
+            0;
+    for (int ti = first_step + tidx; ti <= tlength; ti += THREADS_PER_BLOCK) {
+        float logit = qk_smem[ti - first_step] * inv_sum;
+        if (params.is_return_cross_attentions) {
+            params.cross_attention_out[cross_attention_out_offset + ti] = logit;
+        }
+        convert_from_float(logits_smem[ti - first_step], logit);
+    }
+
+    // Put Values part below so we leverage __syncthreads
+    // from the previous step
+
+    // The number of elements per vector.
+    constexpr int V_VEC_SIZE = Dh_MAX / THREADS_PER_VALUE;
+    // A vector of V elements for the current timestep.
+    using V_vec = typename V_vec_<T, V_VEC_SIZE>::Type;
+
+    // The value computed by this thread.
+    int vo = tidx / THREADS_PER_VALUE;
+    // The hidden dimensions computed by this particular thread.
+    int vi = tidx % THREADS_PER_VALUE * V_VEC_SIZE;
+
+    // The base pointer for the value in the cache buffer.
+    T* v_cache = &params.v_cache[bhi_kv * params.memory_max_len * Dh + vi];
+    // Base pointer for the beam's batch, before offsetting with indirection buffer
+    T* v_cache_batch = &params.v_cache[bbhi_kv * params.memory_max_len * Dh + vi];
+
+    // The number of values processed per iteration of the loop.
+    constexpr int V_PER_ITER = THREADS_PER_BLOCK / THREADS_PER_VALUE;
+
+    // One group of threads computes the product(s) for the current timestep.
+    V_vec v_bias;
+    zero(v_bias);
+    // if( vo == params.timestep % V_PER_ITER ) {
+    if (Dh == Dh_MAX || vi < Dh) {
+        if (handle_kv) {
+            if (vo == tlength % V_PER_ITER) {
+                // Trigger the loads from the V bias buffer.
+                if (params.v_bias != nullptr) {
+                    v_bias = *reinterpret_cast<const V_vec*>(&params.v_bias[hi_kv * Dh + vi]);
+                }
+                if (DO_CROSS_ATTENTION) {
+                    *reinterpret_cast<V_vec*>(&bias_smem[vi]) = v_bias;
+                }
+            }
+        }
+    }
+
+    // From previous, before values, step
+    // Also make sure the logits are in shared memory.
+    __syncthreads();
+
+    // Values continued
+#ifdef MMHA_USE_FP32_ACUM_FOR_OUT
+    using V_vec_acum = typename V_vec_acum_fp32_<V_vec>::Type;
+#else
+    using V_vec_acum = V_vec;
+#endif
+    // The partial outputs computed by each thread.
+    V_vec_acum out;
+    zero(out);
+
+    // Loop over the timesteps to compute the partial outputs.
+    // for( int ti = vo; ti < params.timestep; ti += V_PER_ITER ) {
+    if (Dh == Dh_MAX || vi < Dh) {
+        for (int ti = first_step + vo; ti < tlength; ti += V_PER_ITER) {
+            const int ti_circ = ti % params.memory_max_len;
+
+            // Fetch offset based on cache_indir when beam sampling
+            const int beam_src = (params.cache_indir != nullptr) ? params.cache_indir[bi_seq_len_offset + ti_circ] : 0;
+            const int beam_offset = beam_src * params.num_heads * params.memory_max_len * Dh;
+            // Load the values from the cache.
+            V_vec v = *reinterpret_cast<const V_vec*>(&v_cache_batch[beam_offset + ti_circ * Dh]);
+            if (DO_CROSS_ATTENTION && params.timestep == 0) {
+                v = add(v, *reinterpret_cast<V_vec*>(&bias_smem[vi]));
+                if (do_ia3) {
+                    v = mul<V_vec, V_vec, V_vec>(
+                        v,
+                        *reinterpret_cast<const V_vec*>(
+                            &params.ia3_value_weights[(ia3_task_id * params.num_heads + hi) * Dh + vi]));
+                }
+                *reinterpret_cast<V_vec*>(&v_cache[ti * Dh]) = v;
+            }
+            // Load the logits from shared memory.
+#if defined(MMHA_USE_FP32_ACUM_FOR_LOGITS)
+            float logit = logits_smem[ti - first_step];
+            out         = fma(logit, cast_to_float(v), out);
+#else
+            T logit = logits_smem[ti - first_step];
+
+            // Update the partial sums.
+            out = fma(logit, v, out);
+#endif
+        }
+    }
+
+    // One group of threads computes the product(s) for the current timestep.
+    // if( vo == params.timestep % V_PER_ITER ) {
+    if (vo == tlength % V_PER_ITER && (Dh == Dh_MAX || vi < Dh)) {
+
+        V_vec v;
+        if (DO_CROSS_ATTENTION) {
+            v = *reinterpret_cast<const V_vec*>(&v_cache[tlength * Dh]);
+        }
+        else {
+            // Trigger the loads from the V buffer.
+            const auto v_offset = v_base_offset + vi;
+            if (params.int8_mode == 2) {
+                using Packed_Int8_t  = typename packed_type<int8_t, num_elems<V_vec>::value>::type;
+                using Packed_Float_t = typename packed_type<float, num_elems<V_vec>::value>::type;
+                const auto v_scaling = params.qkv_scale_out[2];
+                const auto v_quant =
+                    *reinterpret_cast<const Packed_Int8_t*>(&reinterpret_cast<const int8_t*>(params.v)[v_offset]);
+
+                convert_from_float(v, mul<Packed_Float_t, float>(v_scaling, float_from_int8(v_quant)));
+            }
+            else {
+                v = *reinterpret_cast<const V_vec*>(&params.v[v_offset]);
+            }
+            // Trigger the loads from the V bias buffer.
+            // V_vec v_bias = *reinterpret_cast<const V_vec*>(&params.v_bias[hi*Dh + vi]);
+        }
+
+        // Compute the V values with bias.
+        v = add(v, v_bias);
+        if (write_kv_cache) {
+
+            if (do_ia3) {
+                v = mul<V_vec, V_vec, V_vec>(
+                    v,
+                    *reinterpret_cast<const V_vec*>(
+                        &params.ia3_value_weights[(ia3_task_id * params.num_heads + hi) * Dh + vi]));
+            }
+
+            // Store the values with bias back to global memory in the cache for V.
+            //*reinterpret_cast<V_vec*>(&v_cache[params.timestep*Dh]) = v;
+            *reinterpret_cast<V_vec*>(&v_cache[tlength_circ * Dh]) = v;
+        }
+
+        // Initialize the output value with the current timestep.
+#if defined(MMHA_USE_FP32_ACUM_FOR_LOGITS)
+        // out = fma(logits_smem[params.timestep], cast_to_float(v), out);
+        out = fma(logits_smem[tlength - first_step], cast_to_float(v), out);
+#else
+        // out = fma(logits_smem[params.timestep], v, out);
+        out = fma(logits_smem[tlength - first_step], v, out);
+#endif
+    }
+
+    // Make sure we can start writing to shared memory.
+    __syncthreads();
+
+    // Run the final reduction amongst the different groups computing different partial outputs.
+    if (Dh == Dh_MAX || vi < Dh) {
+#pragma unroll
+        for (int active_groups = V_PER_ITER; active_groups >= 2; active_groups /= 2) {
+
+            // The midpoint in the number of active groups.
+            int midpoint = active_groups / 2;
+
+            // The upper part of active threads store to shared memory.
+            if (vo >= midpoint && vo < active_groups && (Dh == Dh_MAX || vi < Dh)) {
+#ifdef MMHA_USE_FP32_ACUM_FOR_OUT
+                convert_from_float(*reinterpret_cast<V_vec*>(&out_smem[(vo - midpoint) * Dh + vi]), out);
+#else
+                *reinterpret_cast<V_vec*>(&out_smem[(vo - midpoint) * Dh + vi]) = out;
+#endif
+            }
+            __syncthreads();
+
+            // The bottom warps update their values.
+            if (vo < midpoint && (Dh == Dh_MAX || vi < Dh)) {
+                out = add(*reinterpret_cast<const V_vec*>(&out_smem[vo * Dh + vi]), out);
+            }
+            __syncthreads();
+        }
+    }
+
+    // Output the final values.
+    if (vo == 0 && (Dh == Dh_MAX || vi < Dh)) {
+#ifdef MMHA_USE_FP32_ACUM_FOR_OUT
+        if (params.int8_mode == 2) {
+            using Packed_Int8_t = typename packed_type<int8_t, num_elems<V_vec_acum>::value>::type;
+            out                 = mul<V_vec_acum, float>(*params.attention_out_scale, out);
+            *reinterpret_cast<Packed_Int8_t*>(&(reinterpret_cast<int8_t*>(params.out)[bhi * Dh + vi])) =
+                cast_to_int8(out);
+        }
+        else {
+            convert_from_float(*reinterpret_cast<V_vec*>(&params.out[bhi * Dh + vi]), out);
+        }
+#else
+        // TODO: support int8_mode?
+        *reinterpret_cast<V_vec*>(&params.out[bhi * Dh + vi]) = out;
+#endif
+    }
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+}  // namespace mmha
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T, int Dh, int Dh_MAX, typename KERNEL_PARAMS_TYPE>
+void mmha_launch_kernel(const KERNEL_PARAMS_TYPE& params, const cudaStream_t& stream);

awq_cuda/attention/decoder_masked_multihead_attention_utils.h

+// Downloaded from from FasterTransformer v5.2.1
+// https://github.com/NVIDIA/FasterTransformer/blob/release/v5.2.1_tag/src/fastertransformer/kernels/decoder_masked_multihead_attention_utils.h
+/*
+ * Copyright (c) 2020-2022, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * 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.
+ */
+
+#pragma once
+
+#include "cuda_bf16_wrapper.h"
+#include "cuda_bf16_fallbacks.cuh"
+#include <stdint.h>
+
+using namespace fastertransformer;
+
+namespace mmha {
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+struct Float8_ {
+    float2 x;
+    float2 y;
+    float2 z;
+    float2 w;
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+struct Float4_ {
+    float2 x;
+    float2 y;
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifdef ENABLE_BF16
+struct bf16_4_t {
+    __nv_bfloat162 x;
+    __nv_bfloat162 y;
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+struct bf16_8_t {
+    __nv_bfloat162 x;
+    __nv_bfloat162 y;
+    __nv_bfloat162 z;
+    __nv_bfloat162 w;
+};
+#endif
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T>
+struct num_elems;
+template<>
+struct num_elems<float> {
+    static constexpr int value = 1;
+};
+template<>
+struct num_elems<float2> {
+    static constexpr int value = 2;
+};
+template<>
+struct num_elems<float4> {
+    static constexpr int value = 4;
+};
+template<>
+struct num_elems<Float4_> {
+    static constexpr int value = 4;
+};
+template<>
+struct num_elems<Float8_> {
+    static constexpr int value = 8;
+};
+
+template<>
+struct num_elems<uint32_t> {
+    static constexpr int value = 2;
+};
+template<>
+struct num_elems<uint2> {
+    static constexpr int value = 4;
+};
+template<>
+struct num_elems<uint4> {
+    static constexpr int value = 8;
+};
+
+#ifdef ENABLE_BF16
+template<>
+struct num_elems<__nv_bfloat162> {
+    static constexpr int value = 2;
+};
+template<>
+struct num_elems<bf16_4_t> {
+    static constexpr int value = 4;
+};
+template<>
+struct num_elems<bf16_8_t> {
+    static constexpr int value = 8;
+};
+#endif
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T, int N>
+struct packed_type;
+template<typename T>
+struct packed_type<T, 1> {
+    using type = T;
+};
+template<>
+struct packed_type<int8_t, 2> {
+    using type = int16_t;
+};
+template<>
+struct packed_type<int8_t, 4> {
+    using type = int32_t;
+};
+template<>
+struct packed_type<int8_t, 8> {
+    using type = int64_t;
+};
+
+template<>
+struct packed_type<float, 2> {
+    using type = float2;
+};
+template<>
+struct packed_type<float, 4> {
+    using type = float4;
+};
+template<>
+struct packed_type<float, 8> {
+    using type = Float8_;
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float add(float a, float b)
+{
+    return a + b;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 add(float2 a, float2 b)
+{
+    float2 c;
+    c.x = add(a.x, b.x);
+    c.y = add(a.y, b.y);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float4 add(float4 a, float4 b)
+{
+    float4 c;
+    c.x = add(a.x, b.x);
+    c.y = add(a.y, b.y);
+    c.z = add(a.z, b.z);
+    c.w = add(a.w, b.w);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifdef ENABLE_BF16
+inline __device__ __nv_bfloat16 add(__nv_bfloat16 a, __nv_bfloat16 b)
+{
+    return a + b;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ __nv_bfloat162 add(__nv_bfloat162 a, __nv_bfloat162 b)
+{
+    return bf16hadd2(a, b);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ bf16_4_t add(bf16_4_t a, bf16_4_t b)
+{
+    bf16_4_t c;
+    c.x = add(a.x, b.x);
+    c.y = add(a.y, b.y);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ bf16_8_t add(bf16_8_t a, bf16_8_t b)
+{
+    bf16_8_t c;
+    c.x = add(a.x, b.x);
+    c.y = add(a.y, b.y);
+    c.z = add(a.z, b.z);
+    c.w = add(a.w, b.w);
+    return c;
+}
+#endif  // ENABLE_BF16
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint16_t add(uint16_t a, uint16_t b)
+{
+    uint16_t c;
+    asm volatile("add.f16 %0, %1, %2;\n" : "=h"(c) : "h"(a), "h"(b));
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint32_t add(uint32_t a, uint32_t b)
+{
+    uint32_t c;
+    asm volatile("add.f16x2 %0, %1, %2;\n" : "=r"(c) : "r"(a), "r"(b));
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint2 add(uint2 a, uint2 b)
+{
+    uint2 c;
+    c.x = add(a.x, b.x);
+    c.y = add(a.y, b.y);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint4 add(uint4 a, uint4 b)
+{
+    uint4 c;
+    c.x = add(a.x, b.x);
+    c.y = add(a.y, b.y);
+    c.z = add(a.z, b.z);
+    c.w = add(a.w, b.w);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint16_t float_to_half(float f)
+{
+    union {
+        uint32_t u32;
+        uint16_t u16[2];
+    } tmp;
+#if 0 && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 800  // Is it better?
+  float zero = 0.f;
+  asm volatile("cvt.rn.f16x2.f32 %0, %1, %2;\n" : "=r"(tmp.u32) : "f"(zero), "f"(f));
+#else
+    asm volatile("cvt.rn.f16.f32 %0, %1;\n" : "=h"(tmp.u16[0]) : "f"(f));
+#endif
+    return tmp.u16[0];
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint32_t float2_to_half2(float2 f)
+{
+    union {
+        uint32_t u32;
+        uint16_t u16[2];
+    } tmp;
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 800
+    asm volatile("cvt.rn.f16x2.f32 %0, %1, %2;\n" : "=r"(tmp.u32) : "f"(f.y), "f"(f.x));
+#else
+    asm volatile("cvt.rn.f16.f32 %0, %1;\n" : "=h"(tmp.u16[0]) : "f"(f.x));
+    asm volatile("cvt.rn.f16.f32 %0, %1;\n" : "=h"(tmp.u16[1]) : "f"(f.y));
+#endif
+    return tmp.u32;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float half_to_float(uint16_t h)
+{
+    float f;
+    asm volatile("cvt.f32.f16 %0, %1;\n" : "=f"(f) : "h"(h));
+    return f;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 half2_to_float2(uint32_t v)
+{
+    uint16_t lo, hi;
+    asm volatile("mov.b32 {%0, %1}, %2;\n" : "=h"(lo), "=h"(hi) : "r"(v));
+    return make_float2(half_to_float(lo), half_to_float(hi));
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float add(float a, uint16_t b)
+{
+    return a + half_to_float(b);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifdef ENABLE_BF16
+inline __device__ float add(float a, __nv_bfloat16 b)
+{
+    return a + __bfloat162float(b);
+}
+#endif
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 add(uint32_t a, float2 fb)
+{
+    float2 fa = half2_to_float2(a);
+    return add(fa, fb);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float4_ add(uint2 a, Float4_ fb)
+{
+    Float4_ fc;
+    fc.x = add(a.x, fb.x);
+    fc.y = add(a.y, fb.y);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float8_ add(uint4 a, Float8_ fb)
+{
+    Float8_ fc;
+    fc.x = add(a.x, fb.x);
+    fc.y = add(a.y, fb.y);
+    fc.z = add(a.z, fb.z);
+    fc.w = add(a.w, fb.w);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint32_t h0_h0(uint16_t a)
+{
+    uint32_t b;
+    asm volatile("mov.b32 %0, {%1, %1};" : "=r"(b) : "h"(a));
+    return b;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float fma(float a, float b, float c)
+{
+    return a * b + c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 fma(float2 a, float2 b, float2 c)
+{
+    float2 d;
+    d.x = fma(a.x, b.x, c.x);
+    d.y = fma(a.y, b.y, c.y);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 fma(float a, float2 b, float2 c)
+{
+    float2 d;
+    d.x = fma(a, b.x, c.x);
+    d.y = fma(a, b.y, c.y);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float4 fma(float4 a, float4 b, float4 c)
+{
+    float4 d;
+    d.x = fma(a.x, b.x, c.x);
+    d.y = fma(a.y, b.y, c.y);
+    d.z = fma(a.z, b.z, c.z);
+    d.w = fma(a.w, b.w, c.w);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float4 fma(float a, float4 b, float4 c)
+{
+    float4 d;
+    d.x = fma(a, b.x, c.x);
+    d.y = fma(a, b.y, c.y);
+    d.z = fma(a, b.z, c.z);
+    d.w = fma(a, b.w, c.w);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float4_ fma(float a, Float4_ b, Float4_ c)
+{
+    Float4_ d;
+    d.x = fma(a, b.x, c.x);
+    d.y = fma(a, b.y, c.y);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float8_ fma(float a, Float8_ b, Float8_ c)
+{
+    Float8_ d;
+    d.x = fma(a, b.x, c.x);
+    d.y = fma(a, b.y, c.y);
+    d.z = fma(a, b.z, c.z);
+    d.w = fma(a, b.w, c.w);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifdef ENABLE_BF16
+inline __device__ float2 add(__nv_bfloat162 a, float2 fb)
+{
+    float2 fa = bf1622float2(a);
+    return add(fa, fb);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float4_ add(bf16_4_t a, Float4_ fb)
+{
+    Float4_ fc;
+    fc.x = add(a.x, fb.x);
+    fc.y = add(a.y, fb.y);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float8_ add(bf16_8_t a, Float8_ fb)
+{
+    Float8_ fc;
+    fc.x = add(a.x, fb.x);
+    fc.y = add(a.y, fb.y);
+    fc.z = add(a.z, fb.z);
+    fc.w = add(a.w, fb.w);
+    return fc;
+}
+#endif  // ENABLE_BF16
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint32_t fma(uint32_t a, uint32_t b, uint32_t c)
+{
+    uint32_t d;
+    asm volatile("fma.rn.f16x2 %0, %1, %2, %3;\n" : "=r"(d) : "r"(a), "r"(b), "r"(c));
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint32_t fma(uint16_t a, uint32_t b, uint32_t c)
+{
+    return fma(h0_h0(a), b, c);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint2 fma(uint2 a, uint2 b, uint2 c)
+{
+    uint2 d;
+    d.x = fma(a.x, b.x, c.x);
+    d.y = fma(a.y, b.y, c.y);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint2 fma(uint16_t a, uint2 b, uint2 c)
+{
+    uint32_t s = h0_h0(a);
+    uint2    d;
+    d.x = fma(s, b.x, c.x);
+    d.y = fma(s, b.y, c.y);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint4 fma(uint4 a, uint4 b, uint4 c)
+{
+    uint4 d;
+    d.x = fma(a.x, b.x, c.x);
+    d.y = fma(a.y, b.y, c.y);
+    d.z = fma(a.z, b.z, c.z);
+    d.w = fma(a.w, b.w, c.w);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ uint4 fma(uint16_t a, uint4 b, uint4 c)
+{
+    uint32_t s = h0_h0(a);
+    uint4    d;
+    d.x = fma(s, b.x, c.x);
+    d.y = fma(s, b.y, c.y);
+    d.z = fma(s, b.z, c.z);
+    d.w = fma(s, b.w, c.w);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float fma(uint16_t a, uint16_t b, float fc)
+{
+    float fa = half_to_float(a);
+    float fb = half_to_float(b);
+    return fa * fb + fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 fma(uint32_t a, uint32_t b, float2 fc)
+{
+    float2 fa = half2_to_float2(a);
+    float2 fb = half2_to_float2(b);
+    return fma(fa, fb, fc);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 fma(uint16_t a, uint32_t b, float2 fc)
+{
+    return fma(h0_h0(a), b, fc);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float4_ fma(uint2 a, uint2 b, Float4_ fc)
+{
+    Float4_ fd;
+    fd.x = fma(a.x, b.x, fc.x);
+    fd.y = fma(a.y, b.y, fc.y);
+    return fd;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float4_ fma(uint16_t a, uint2 b, Float4_ fc)
+{
+    uint32_t s = h0_h0(a);
+    Float4_  fd;
+    fd.x = fma(s, b.x, fc.x);
+    fd.y = fma(s, b.y, fc.y);
+    return fd;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float8_ fma(uint4 a, uint4 b, Float8_ fc)
+{
+    Float8_ fd;
+    fd.x = fma(a.x, b.x, fc.x);
+    fd.y = fma(a.y, b.y, fc.y);
+    fd.z = fma(a.z, b.z, fc.z);
+    fd.w = fma(a.w, b.w, fc.w);
+    return fd;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float8_ fma(uint16_t a, uint4 b, Float8_ fc)
+{
+    uint32_t s = h0_h0(a);
+    Float8_  fd;
+    fd.x = fma(s, b.x, fc.x);
+    fd.y = fma(s, b.y, fc.y);
+    fd.z = fma(s, b.z, fc.z);
+    fd.w = fma(s, b.w, fc.w);
+    return fd;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+#ifdef ENABLE_BF16
+inline __device__ __nv_bfloat162 fma(__nv_bfloat162 a, __nv_bfloat162 b, __nv_bfloat162 c)
+{
+    return bf16hfma2(a, b, c);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ __nv_bfloat162 fma(__nv_bfloat16 a, __nv_bfloat162 b, __nv_bfloat162 c)
+{
+    return bf16hfma2(bf162bf162(a), b, c);
+}
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ bf16_4_t fma(bf16_4_t a, bf16_4_t b, bf16_4_t c)
+{
+    bf16_4_t d;
+    d.x = fma(a.x, b.x, c.x);
+    d.y = fma(a.y, b.y, c.y);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ bf16_4_t fma(__nv_bfloat16 a, bf16_4_t b, bf16_4_t c)
+{
+    __nv_bfloat162 s = bf162bf162(a);
+    bf16_4_t       d;
+    d.x = fma(s, b.x, c.x);
+    d.y = fma(s, b.y, c.y);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ bf16_8_t fma(bf16_8_t a, bf16_8_t b, bf16_8_t c)
+{
+    bf16_8_t d;
+    d.x = fma(a.x, b.x, c.x);
+    d.y = fma(a.y, b.y, c.y);
+    d.z = fma(a.z, b.z, c.z);
+    d.w = fma(a.w, b.w, c.w);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ bf16_8_t fma(__nv_bfloat16 a, bf16_8_t b, bf16_8_t c)
+{
+    __nv_bfloat162 s = bf162bf162(a);
+    bf16_8_t       d;
+    d.x = fma(s, b.x, c.x);
+    d.y = fma(s, b.y, c.y);
+    d.z = fma(s, b.z, c.z);
+    d.w = fma(s, b.w, c.w);
+    return d;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float fma(__nv_bfloat16 a, __nv_bfloat16 b, float fc)
+{
+    return __bfloat162float(a) * __bfloat162float(b) + fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 fma(__nv_bfloat162 a, __nv_bfloat162 b, float2 fc)
+{
+    float2 fa = bf1622float2(a);
+    float2 fb = bf1622float2(b);
+    return fma(fa, fb, fc);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 fma(__nv_bfloat16 a, __nv_bfloat162 b, float2 fc)
+{
+    return fma(bf162bf162(a), b, fc);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float4_ fma(bf16_4_t a, bf16_4_t b, Float4_ fc)
+{
+    Float4_ fd;
+    fd.x = fma(a.x, b.x, fc.x);
+    fd.y = fma(a.y, b.y, fc.y);
+    return fd;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float4_ fma(__nv_bfloat16 a, bf16_4_t b, Float4_ fc)
+{
+    __nv_bfloat162 s = bf162bf162(a);
+    Float4_        fd;
+    fd.x = fma(s, b.x, fc.x);
+    fd.y = fma(s, b.y, fc.y);
+    return fd;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float8_ fma(bf16_8_t a, bf16_8_t b, Float8_ fc)
+{
+    Float8_ fd;
+    fd.x = fma(a.x, b.x, fc.x);
+    fd.y = fma(a.y, b.y, fc.y);
+    fd.z = fma(a.z, b.z, fc.z);
+    fd.w = fma(a.w, b.w, fc.w);
+    return fd;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ Float8_ fma(__nv_bfloat16 a, bf16_8_t b, Float8_ fc)
+{
+    __nv_bfloat162 s = bf162bf162(a);
+    Float8_        fd;
+    fd.x = fma(s, b.x, fc.x);
+    fd.y = fma(s, b.y, fc.y);
+    fd.z = fma(s, b.z, fc.z);
+    fd.w = fma(s, b.w, fc.w);
+    return fd;
+}
+#endif  // ENABLE_BF16
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename Acc, typename A, typename B>
+inline __device__ Acc mul(A a, B b)
+{
+    return a * b;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float mul<float, float>(float a, float b)
+{
+    return a * b;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float2 mul(float2 a, float2 b)
+{
+    float2 c;
+    c.x = a.x * b.x;
+    c.y = a.y * b.y;
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float2 mul(float a, float2 b)
+{
+    float2 c;
+    c.x = a * b.x;
+    c.y = a * b.y;
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float4 mul(float4 a, float4 b)
+{
+    float4 c;
+    c.x = a.x * b.x;
+    c.y = a.y * b.y;
+    c.z = a.z * b.z;
+    c.w = a.w * b.w;
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float4 mul(float a, float4 b)
+{
+    float4 c;
+    c.x = a * b.x;
+    c.y = a * b.y;
+    c.z = a * b.z;
+    c.w = a * b.w;
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ Float8_ mul(float a, Float8_ b)
+{
+    Float8_ c;
+    c.x = make_float2(a * b.x.x, a * b.x.y);
+    c.y = make_float2(a * b.y.x, a * b.y.y);
+    c.z = make_float2(a * b.z.x, a * b.z.y);
+    c.w = make_float2(a * b.w.x, a * b.w.y);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ uint16_t mul(uint16_t a, uint16_t b)
+{
+    uint16_t c;
+    asm volatile("mul.f16 %0, %1, %2;\n" : "=h"(c) : "h"(a), "h"(b));
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ uint32_t mul(uint32_t a, uint32_t b)
+{
+    uint32_t c;
+    asm volatile("mul.f16x2 %0, %1, %2;\n" : "=r"(c) : "r"(a), "r"(b));
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ uint32_t mul(uint16_t a, uint32_t b)
+{
+    return mul<uint32_t, uint32_t, uint32_t>(h0_h0(a), b);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ uint2 mul(uint2 a, uint2 b)
+{
+    uint2 c;
+    c.x = mul<uint32_t, uint32_t, uint32_t>(a.x, b.x);
+    c.y = mul<uint32_t, uint32_t, uint32_t>(a.y, b.y);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ uint2 mul(uint16_t a, uint2 b)
+{
+    uint32_t s = h0_h0(a);
+    uint2    c;
+    c.x = mul<uint32_t, uint32_t, uint32_t>(s, b.x);
+    c.y = mul<uint32_t, uint32_t, uint32_t>(s, b.y);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ uint4 mul(uint4 a, uint4 b)
+{
+    uint4 c;
+    c.x = mul<uint32_t, uint32_t, uint32_t>(a.x, b.x);
+    c.y = mul<uint32_t, uint32_t, uint32_t>(a.y, b.y);
+    c.z = mul<uint32_t, uint32_t, uint32_t>(a.z, b.z);
+    c.w = mul<uint32_t, uint32_t, uint32_t>(a.w, b.w);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ uint4 mul(uint16_t a, uint4 b)
+{
+    uint32_t s = h0_h0(a);
+    uint4    c;
+    c.x = mul<uint32_t, uint32_t, uint32_t>(s, b.x);
+    c.y = mul<uint32_t, uint32_t, uint32_t>(s, b.y);
+    c.z = mul<uint32_t, uint32_t, uint32_t>(s, b.z);
+    c.w = mul<uint32_t, uint32_t, uint32_t>(s, b.w);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float mul(uint16_t a, uint16_t b)
+{
+    float fa = half_to_float(a);
+    float fb = half_to_float(b);
+    return fa * fb;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float mul(uint16_t a, float b)
+{
+    return half_to_float(a) * b;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float2 mul(uint32_t a, uint32_t b)
+{
+    float2 fa = half2_to_float2(a);
+    float2 fb = half2_to_float2(b);
+    return mul<float2, float2, float2>(fa, fb);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float2 mul(uint16_t a, uint32_t b)
+{
+    return mul<float2, uint32_t, uint32_t>(h0_h0(a), b);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ Float4_ mul(uint2 a, uint2 b)
+{
+    Float4_ fc;
+    fc.x = mul<float2, uint32_t, uint32_t>(a.x, b.x);
+    fc.y = mul<float2, uint32_t, uint32_t>(a.y, b.y);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ Float4_ mul(uint16_t a, uint2 b)
+{
+    uint32_t s = h0_h0(a);
+    Float4_  fc;
+    fc.x = mul<float2, uint32_t, uint32_t>(s, b.x);
+    fc.y = mul<float2, uint32_t, uint32_t>(s, b.y);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ Float8_ mul(uint4 a, uint4 b)
+{
+    Float8_ fc;
+    fc.x = mul<float2, uint32_t, uint32_t>(a.x, b.x);
+    fc.y = mul<float2, uint32_t, uint32_t>(a.y, b.y);
+    fc.z = mul<float2, uint32_t, uint32_t>(a.z, b.z);
+    fc.w = mul<float2, uint32_t, uint32_t>(a.w, b.w);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ Float8_ mul(uint16_t a, uint4 b)
+{
+    uint32_t s = h0_h0(a);
+    Float8_  fc;
+    fc.x = mul<float2, uint32_t, uint32_t>(s, b.x);
+    fc.y = mul<float2, uint32_t, uint32_t>(s, b.y);
+    fc.z = mul<float2, uint32_t, uint32_t>(s, b.z);
+    fc.w = mul<float2, uint32_t, uint32_t>(s, b.w);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifdef ENABLE_BF16
+template<>
+inline __device__ __nv_bfloat16 mul(__nv_bfloat16 a, __nv_bfloat16 b)
+{
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 800
+    return __hmul(a, b);
+#else
+    return bf16hmul(a, b);
+#endif
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ __nv_bfloat162 mul(__nv_bfloat162 a, __nv_bfloat162 b)
+{
+    return bf16hmul2(a, b);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ __nv_bfloat162 mul(__nv_bfloat16 a, __nv_bfloat162 b)
+{
+    return mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(bf162bf162(a), b);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ bf16_4_t mul(bf16_4_t a, bf16_4_t b)
+{
+    bf16_4_t c;
+    c.x = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(a.x, b.x);
+    c.y = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(a.y, b.y);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ bf16_4_t mul(__nv_bfloat16 a, bf16_4_t b)
+{
+    __nv_bfloat162 s = bf162bf162(a);
+    bf16_4_t       c;
+    c.x = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(s, b.x);
+    c.y = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(s, b.y);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ bf16_8_t mul(bf16_8_t a, bf16_8_t b)
+{
+    bf16_8_t c;
+    c.x = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(a.x, b.x);
+    c.y = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(a.y, b.y);
+    c.z = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(a.z, b.z);
+    c.w = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(a.w, b.w);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ bf16_8_t mul(__nv_bfloat16 a, bf16_8_t b)
+{
+    __nv_bfloat162 s = bf162bf162(a);
+    bf16_8_t       c;
+    c.x = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(s, b.x);
+    c.y = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(s, b.y);
+    c.z = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(s, b.z);
+    c.w = mul<__nv_bfloat162, __nv_bfloat162, __nv_bfloat162>(s, b.w);
+    return c;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float mul(__nv_bfloat16 a, __nv_bfloat16 b)
+{
+    float fa = (float)a;
+    float fb = (float)b;
+    return fa * fb;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float mul(__nv_bfloat16 a, float b)
+{
+    return __bfloat162float(a) * b;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float2 mul(__nv_bfloat162 a, __nv_bfloat162 b)
+{
+    float2 fa = bf1622float2(a);
+    float2 fb = bf1622float2(b);
+    return mul<float2, float2, float2>(fa, fb);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ float2 mul(__nv_bfloat16 a, __nv_bfloat162 b)
+{
+    return mul<float2, __nv_bfloat162, __nv_bfloat162>(bf162bf162(a), b);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ Float4_ mul(bf16_4_t a, bf16_4_t b)
+{
+    Float4_ fc;
+    fc.x = mul<float2, __nv_bfloat162, __nv_bfloat162>(a.x, b.x);
+    fc.y = mul<float2, __nv_bfloat162, __nv_bfloat162>(a.y, b.y);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ Float4_ mul(__nv_bfloat16 a, bf16_4_t b)
+{
+    __nv_bfloat162 s = bf162bf162(a);
+    Float4_        fc;
+    fc.x = mul<float2, __nv_bfloat162, __nv_bfloat162>(s, b.x);
+    fc.y = mul<float2, __nv_bfloat162, __nv_bfloat162>(s, b.y);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ Float8_ mul(bf16_8_t a, bf16_8_t b)
+{
+    Float8_ fc;
+    fc.x = mul<float2, __nv_bfloat162, __nv_bfloat162>(a.x, b.x);
+    fc.y = mul<float2, __nv_bfloat162, __nv_bfloat162>(a.y, b.y);
+    fc.z = mul<float2, __nv_bfloat162, __nv_bfloat162>(a.z, b.z);
+    fc.w = mul<float2, __nv_bfloat162, __nv_bfloat162>(a.w, b.w);
+    return fc;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<>
+inline __device__ Float8_ mul(__nv_bfloat16 a, bf16_8_t b)
+{
+    __nv_bfloat162 s = bf162bf162(a);
+    Float8_        fc;
+    fc.x = mul<float2, __nv_bfloat162, __nv_bfloat162>(s, b.x);
+    fc.y = mul<float2, __nv_bfloat162, __nv_bfloat162>(s, b.y);
+    fc.z = mul<float2, __nv_bfloat162, __nv_bfloat162>(s, b.z);
+    fc.w = mul<float2, __nv_bfloat162, __nv_bfloat162>(s, b.w);
+    return fc;
+}
+#endif  // ENABLE_BF16
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(float v)
+{
+    return v;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(float2 v)
+{
+    return v.x + v.y;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(float4 v)
+{
+    return v.x + v.y + v.z + v.w;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifdef ENABLE_BF16
+inline __device__ float sum(__nv_bfloat162 v)
+{
+    float2 vf = bf1622float2(v);
+    return vf.x + vf.y;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(bf16_4_t v)
+{
+    return sum(v.x) + sum(v.y);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(bf16_8_t v)
+{
+    return sum(v.x) + sum(v.y) + sum(v.z) + sum(v.w);
+}
+#endif  // ENABLE_BF16
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(uint16_t v)
+{
+    return half_to_float(v);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(uint32_t v)
+{
+    float2 tmp = half2_to_float2(v);
+    return tmp.x + tmp.y;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(uint2 v)
+{
+    uint32_t c = add(v.x, v.y);
+    return sum(c);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(uint4 v)
+{
+#if 1
+    uint32_t c = add(v.x, v.y);
+    c          = add(c, v.z);
+    c          = add(c, v.w);
+#else
+    uint32_t c = add(v.x, v.y);
+    uint32_t d = add(v.z, v.w);
+    c          = add(c, d);
+#endif
+    return sum(c);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(Float4_ v)
+{
+    return v.x.x + v.x.y + v.y.x + v.y.y;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float sum(Float8_ v)
+{
+    return v.x.x + v.x.y + v.y.x + v.y.y + v.z.x + v.z.y + v.w.x + v.w.y;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T>
+inline __device__ float dot(T a, T b)
+{
+    return sum(mul<T, T, T>(a, b));
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename A, typename T>
+inline __device__ float dot(T a, T b)
+{
+    return sum(mul<A, T, T>(a, b));
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ void zero(uint16_t& dst)
+{
+    dst = uint16_t(0);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template<typename T>
+inline __device__ void zero(T& dst)
+{
+    constexpr int WORDS = sizeof(T) / 4;
+    union {
+        T        raw;
+        uint32_t words[WORDS];
+    } tmp;
+#pragma unroll
+    for (int ii = 0; ii < WORDS; ++ii) {
+        tmp.words[ii] = 0u;
+    }
+    dst = tmp.raw;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+inline __device__ float2 rotary_embedding_coefficient(const int zid, const int rot_embed_dim, const float t_step, const float base)
+{
+    const float inv_freq = t_step / pow(base, zid / (float)rot_embed_dim);
+    return {cos(inv_freq), sin(inv_freq)};
+}
+
+inline __device__ float2 rotary_embedding_transform(const float2 v, const float2 coef)
+{
+    float2 rot_v;
+    rot_v.x = coef.x * v.x - coef.y * v.y;
+    rot_v.y = coef.x * v.y + coef.y * v.x;
+    return rot_v;
+}
+
+inline __device__ uint32_t rotary_embedding_transform(const uint32_t v, const float2 coef)
+{
+    float2 fv     = half2_to_float2(v);
+    float2 rot_fv = rotary_embedding_transform(fv, coef);
+    return float2_to_half2(rot_fv);
+}
+
+#ifdef ENABLE_BF16
+inline __device__ __nv_bfloat162 rotary_embedding_transform(const __nv_bfloat162 v, const float2 coef)
+{
+    float2 fv     = bf1622float2(v);
+    float2 rot_fv = rotary_embedding_transform(fv, coef);
+    return __floats2bfloat162_rn(rot_fv.x, rot_fv.y);
+}
+#endif
+
+inline __device__ void apply_rotary_embedding(float& q, int zid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    return;
+}
+
+inline __device__ void apply_rotary_embedding(float& q, float& k, int zid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    return;
+}
+
+inline __device__ void apply_rotary_embedding(float2& q, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (2 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef = rotary_embedding_coefficient(2 * tid, rot_embed_dim, t_step, base);
+    q               = rotary_embedding_transform(q, coef);
+}
+
+inline __device__ void apply_rotary_embedding(float2& q, float2& k, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (2 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef = rotary_embedding_coefficient(2 * tid, rot_embed_dim, t_step, base);
+    q               = rotary_embedding_transform(q, coef);
+    k               = rotary_embedding_transform(k, coef);
+}
+
+inline __device__ void apply_rotary_embedding(float4& q, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (4 * tid >= rot_embed_dim) {
+        return;
+    }
+
+    Float4_&   q_    = *reinterpret_cast<Float4_*>(&q);
+    const auto coef0 = rotary_embedding_coefficient(4 * tid, rot_embed_dim, t_step, base);
+    q_.x             = rotary_embedding_transform(q_.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(4 * tid + 2, rot_embed_dim, t_step, base);
+    q_.y             = rotary_embedding_transform(q_.y, coef1);
+}
+
+inline __device__ void apply_rotary_embedding(float4& q, float4& k, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (4 * tid >= rot_embed_dim) {
+        return;
+    }
+
+    Float4_&   q_    = *reinterpret_cast<Float4_*>(&q);
+    Float4_&   k_    = *reinterpret_cast<Float4_*>(&k);
+    const auto coef0 = rotary_embedding_coefficient(4 * tid, rot_embed_dim, t_step, base);
+    q_.x             = rotary_embedding_transform(q_.x, coef0);
+    k_.x             = rotary_embedding_transform(k_.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(4 * tid + 2, rot_embed_dim, t_step, base);
+    q_.y             = rotary_embedding_transform(q_.y, coef1);
+    k_.y             = rotary_embedding_transform(k_.y, coef1);
+}
+
+inline __device__ void apply_rotary_embedding(uint32_t& q, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (2 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef = rotary_embedding_coefficient(2 * tid, rot_embed_dim, t_step, base);
+    q               = rotary_embedding_transform(q, coef);
+}
+
+inline __device__ void apply_rotary_embedding(uint32_t& q, uint32_t& k, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (2 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef = rotary_embedding_coefficient(2 * tid, rot_embed_dim, t_step, base);
+    q               = rotary_embedding_transform(q, coef);
+    k               = rotary_embedding_transform(k, coef);
+}
+
+inline __device__ void apply_rotary_embedding(uint2& q, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (4 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef0 = rotary_embedding_coefficient(4 * tid, rot_embed_dim, t_step, base);
+    q.x              = rotary_embedding_transform(q.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(4 * tid + 2, rot_embed_dim, t_step, base);
+    q.y              = rotary_embedding_transform(q.y, coef1);
+}
+
+inline __device__ void apply_rotary_embedding(uint2& q, uint2& k, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (4 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef0 = rotary_embedding_coefficient(4 * tid, rot_embed_dim, t_step, base);
+    q.x              = rotary_embedding_transform(q.x, coef0);
+    k.x              = rotary_embedding_transform(k.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(4 * tid + 2, rot_embed_dim, t_step, base);
+    q.y              = rotary_embedding_transform(q.y, coef1);
+    k.y              = rotary_embedding_transform(k.y, coef1);
+}
+
+inline __device__ void apply_rotary_embedding(uint4& q, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (8 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef0 = rotary_embedding_coefficient(8 * tid, rot_embed_dim, t_step, base);
+    q.x              = rotary_embedding_transform(q.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(8 * tid + 2, rot_embed_dim, t_step, base);
+    q.y              = rotary_embedding_transform(q.y, coef1);
+    const auto coef2 = rotary_embedding_coefficient(8 * tid + 4, rot_embed_dim, t_step, base);
+    q.z              = rotary_embedding_transform(q.z, coef2);
+    const auto coef3 = rotary_embedding_coefficient(8 * tid + 6, rot_embed_dim, t_step, base);
+    q.w              = rotary_embedding_transform(q.w, coef3);
+}
+
+inline __device__ void apply_rotary_embedding(uint4& q, uint4& k, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (8 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef0 = rotary_embedding_coefficient(8 * tid, rot_embed_dim, t_step, base);
+    q.x              = rotary_embedding_transform(q.x, coef0);
+    k.x              = rotary_embedding_transform(k.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(8 * tid + 2, rot_embed_dim, t_step, base);
+    q.y              = rotary_embedding_transform(q.y, coef1);
+    k.y              = rotary_embedding_transform(k.y, coef1);
+    const auto coef2 = rotary_embedding_coefficient(8 * tid + 4, rot_embed_dim, t_step, base);
+    q.z              = rotary_embedding_transform(q.z, coef2);
+    k.z              = rotary_embedding_transform(k.z, coef2);
+    const auto coef3 = rotary_embedding_coefficient(8 * tid + 6, rot_embed_dim, t_step, base);
+    q.w              = rotary_embedding_transform(q.w, coef3);
+    k.w              = rotary_embedding_transform(k.w, coef3);
+}
+
+#ifdef ENABLE_BF16
+inline __device__ void apply_rotary_embedding(__nv_bfloat162& q, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (2 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef = rotary_embedding_coefficient(2 * tid, rot_embed_dim, t_step, base);
+    q               = rotary_embedding_transform(q, coef);
+}
+
+inline __device__ void
+apply_rotary_embedding(__nv_bfloat162& q, __nv_bfloat162& k, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (2 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef = rotary_embedding_coefficient(2 * tid, rot_embed_dim, t_step, base);
+    q               = rotary_embedding_transform(q, coef);
+    k               = rotary_embedding_transform(k, coef);
+}
+
+inline __device__ void apply_rotary_embedding(bf16_4_t& q, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (4 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef0 = rotary_embedding_coefficient(4 * tid, rot_embed_dim, t_step, base);
+    q.x              = rotary_embedding_transform(q.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(4 * tid + 2, rot_embed_dim, t_step, base);
+    q.y              = rotary_embedding_transform(q.y, coef1);
+}
+
+inline __device__ void apply_rotary_embedding(bf16_4_t& q, bf16_4_t& k, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (4 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef0 = rotary_embedding_coefficient(4 * tid, rot_embed_dim, t_step, base);
+    q.x              = rotary_embedding_transform(q.x, coef0);
+    k.x              = rotary_embedding_transform(k.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(4 * tid + 2, rot_embed_dim, t_step, base);
+    q.y              = rotary_embedding_transform(q.y, coef1);
+    k.y              = rotary_embedding_transform(k.y, coef1);
+}
+
+inline __device__ void apply_rotary_embedding(bf16_8_t& q, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (8 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef0 = rotary_embedding_coefficient(8 * tid, rot_embed_dim, t_step, base);
+    q.x              = rotary_embedding_transform(q.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(8 * tid + 2, rot_embed_dim, t_step, base);
+    q.y              = rotary_embedding_transform(q.y, coef1);
+    const auto coef2 = rotary_embedding_coefficient(8 * tid + 4, rot_embed_dim, t_step, base);
+    q.z              = rotary_embedding_transform(q.z, coef2);
+    const auto coef3 = rotary_embedding_coefficient(8 * tid + 6, rot_embed_dim, t_step, base);
+    q.w              = rotary_embedding_transform(q.w, coef3);
+}
+
+inline __device__ void apply_rotary_embedding(bf16_8_t& q, bf16_8_t& k, int tid, int rot_embed_dim, int t_step, const float base=10000.0f)
+{
+    if (8 * tid >= rot_embed_dim) {
+        return;
+    }
+    const auto coef0 = rotary_embedding_coefficient(8 * tid, rot_embed_dim, t_step, base);
+    q.x              = rotary_embedding_transform(q.x, coef0);
+    k.x              = rotary_embedding_transform(k.x, coef0);
+    const auto coef1 = rotary_embedding_coefficient(8 * tid + 2, rot_embed_dim, t_step, base);
+    q.y              = rotary_embedding_transform(q.y, coef1);
+    k.y              = rotary_embedding_transform(k.y, coef1);
+    const auto coef2 = rotary_embedding_coefficient(8 * tid + 4, rot_embed_dim, t_step, base);
+    q.z              = rotary_embedding_transform(q.z, coef2);
+    k.z              = rotary_embedding_transform(k.z, coef2);
+    const auto coef3 = rotary_embedding_coefficient(8 * tid + 6, rot_embed_dim, t_step, base);
+    q.w              = rotary_embedding_transform(q.w, coef3);
+    k.w              = rotary_embedding_transform(k.w, coef3);
+}
+#endif  // ENABLE_BF16
+
+template<typename Vec_T, typename T>
+__device__ __inline__ void vec_from_smem_transpose(Vec_T& vec, T* smem, int transpose_idx, int smem_pitch);
+
+template<>
+__device__ __inline__ void vec_from_smem_transpose(float& vec, float* smem, int transpose_idx, int smem_pitch)
+{
+    return;
+}
+
+template<>
+__device__ __inline__ void vec_from_smem_transpose(uint32_t& vec, uint16_t* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint32_t u32;
+        uint16_t u16[2];
+    } tmp;
+    tmp.u16[0] = smem[transpose_idx];
+    tmp.u16[1] = smem[smem_pitch + transpose_idx];
+
+    vec = tmp.u32;
+}
+
+template<>
+__device__ __inline__ void vec_from_smem_transpose(uint2& vec, uint16_t* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint32_t u32;
+        uint16_t u16[2];
+    } tmp_1, tmp_2;
+    tmp_1.u32 = *reinterpret_cast<uint32_t*>(&smem[transpose_idx]);
+    tmp_2.u32 = *reinterpret_cast<uint32_t*>(&smem[smem_pitch + transpose_idx]);
+
+    union {
+        uint2    u32x2;
+        uint16_t u16[4];
+    } tmp_3;
+    tmp_3.u16[0] = tmp_1.u16[0];
+    tmp_3.u16[1] = tmp_2.u16[0];
+    tmp_3.u16[2] = tmp_1.u16[1];
+    tmp_3.u16[3] = tmp_2.u16[1];
+
+    vec = tmp_3.u32x2;
+}
+
+template<>
+__device__ __inline__ void vec_from_smem_transpose(uint4& vec, uint16_t* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint64_t u64;
+        uint16_t u16[4];
+    } tmp_1, tmp_2;
+    tmp_1.u64 = *reinterpret_cast<uint64_t*>(&smem[transpose_idx]);
+    tmp_2.u64 = *reinterpret_cast<uint64_t*>(&smem[smem_pitch + transpose_idx]);
+
+    union {
+        uint4    u32x4;
+        uint16_t u16[8];
+    } tmp_3;
+    tmp_3.u16[0] = tmp_1.u16[0];
+    tmp_3.u16[1] = tmp_2.u16[0];
+    tmp_3.u16[2] = tmp_1.u16[1];
+    tmp_3.u16[3] = tmp_2.u16[1];
+    tmp_3.u16[4] = tmp_1.u16[2];
+    tmp_3.u16[5] = tmp_2.u16[2];
+    tmp_3.u16[6] = tmp_1.u16[3];
+    tmp_3.u16[7] = tmp_2.u16[3];
+
+    vec = tmp_3.u32x4;
+}
+
+#ifdef ENABLE_BF16
+template<>
+__device__ __inline__ void
+vec_from_smem_transpose(bf16_4_t& vec, __nv_bfloat16* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint32_t      u32;
+        __nv_bfloat16 bf16[2];
+    } tmp_1, tmp_2;
+    tmp_1.u32 = *reinterpret_cast<uint32_t*>(&smem[transpose_idx]);
+    tmp_2.u32 = *reinterpret_cast<uint32_t*>(&smem[smem_pitch + transpose_idx]);
+
+    vec.x = __nv_bfloat162{tmp_1.bf16[0], tmp_2.bf16[0]};
+    vec.y = __nv_bfloat162{tmp_1.bf16[1], tmp_2.bf16[1]};
+}
+
+template<>
+__device__ __inline__ void
+vec_from_smem_transpose(bf16_8_t& vec, __nv_bfloat16* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint64_t      u64;
+        __nv_bfloat16 bf16[4];
+    } tmp_1, tmp_2;
+    tmp_1.u64 = *reinterpret_cast<uint64_t*>(&smem[transpose_idx]);
+    tmp_2.u64 = *reinterpret_cast<uint64_t*>(&smem[smem_pitch + transpose_idx]);
+
+    vec.x = __nv_bfloat162{tmp_1.bf16[0], tmp_2.bf16[0]};
+    vec.y = __nv_bfloat162{tmp_1.bf16[1], tmp_2.bf16[1]};
+    vec.z = __nv_bfloat162{tmp_1.bf16[2], tmp_2.bf16[2]};
+    vec.w = __nv_bfloat162{tmp_1.bf16[3], tmp_2.bf16[3]};
+}
+#endif  // ENABLE_BF16
+
+template<>
+__device__ __inline__ void vec_from_smem_transpose(float4& vec, float* smem, int transpose_idx, int smem_pitch)
+{
+    vec.x = smem[transpose_idx];
+    vec.z = smem[transpose_idx + 1];
+    vec.y = smem[smem_pitch + transpose_idx];
+    vec.w = smem[smem_pitch + transpose_idx + 1];
+}
+
+template<>
+__device__ __inline__ void vec_from_smem_transpose(uint32_t& vec, half* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint32_t u32;
+        half     u16[2];
+    } tmp;
+    tmp.u16[0] = smem[transpose_idx];
+    tmp.u16[1] = smem[smem_pitch + transpose_idx];
+
+    vec = tmp.u32;
+}
+
+#ifdef ENABLE_BF16
+template<>
+__device__ __inline__ void
+vec_from_smem_transpose(__nv_bfloat162& vec, __nv_bfloat16* smem, int transpose_idx, int smem_pitch)
+{
+    vec.x = smem[transpose_idx];
+    vec.y = smem[smem_pitch + transpose_idx];
+}
+#endif
+
+template<>
+__device__ __inline__ void vec_from_smem_transpose(float2& vec, float* smem, int transpose_idx, int smem_pitch)
+{
+    vec.x = smem[transpose_idx];
+    vec.y = smem[smem_pitch + transpose_idx];
+}
+
+template<typename Vec_T, typename T>
+__device__ __inline__ void write_smem_transpose(const Vec_T& vec, T* smem, int transpose_idx, int smem_pitch);
+
+template<>
+__device__ __inline__ void write_smem_transpose(const float& vec, float* smem, int transpose_idx, int smem_pitch)
+{
+    return;
+}
+
+template<>
+__device__ __inline__ void write_smem_transpose(const uint4& vec, uint16_t* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint64_t u64;
+        uint16_t u16[4];
+    } tmp_1, tmp_2;
+
+    union {
+        uint4    u32x4;
+        uint16_t u16[8];
+    } tmp_3;
+    tmp_3.u32x4  = vec;
+    tmp_1.u16[0] = tmp_3.u16[0];
+    tmp_2.u16[0] = tmp_3.u16[1];
+    tmp_1.u16[1] = tmp_3.u16[2];
+    tmp_2.u16[1] = tmp_3.u16[3];
+    tmp_1.u16[2] = tmp_3.u16[4];
+    tmp_2.u16[2] = tmp_3.u16[5];
+    tmp_1.u16[3] = tmp_3.u16[6];
+    tmp_2.u16[3] = tmp_3.u16[7];
+
+    *reinterpret_cast<uint64_t*>(&smem[transpose_idx])              = tmp_1.u64;
+    *reinterpret_cast<uint64_t*>(&smem[smem_pitch + transpose_idx]) = tmp_2.u64;
+}
+
+template<>
+__device__ __inline__ void write_smem_transpose(const uint2& vec, uint16_t* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint32_t u32;
+        uint16_t u16[2];
+    } tmp_1, tmp_2;
+
+    union {
+        uint2    u32x2;
+        uint16_t u16[4];
+    } tmp_3;
+    tmp_3.u32x2  = vec;
+    tmp_1.u16[0] = tmp_3.u16[0];
+    tmp_2.u16[0] = tmp_3.u16[1];
+    tmp_1.u16[1] = tmp_3.u16[2];
+    tmp_2.u16[1] = tmp_3.u16[3];
+
+    *reinterpret_cast<uint32_t*>(&smem[transpose_idx])              = tmp_1.u32;
+    *reinterpret_cast<uint32_t*>(&smem[smem_pitch + transpose_idx]) = tmp_2.u32;
+}
+
+template<>
+__device__ __inline__ void write_smem_transpose(const uint32_t& vec, uint16_t* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint32_t u32;
+        uint16_t u16[2];
+    } tmp;
+    tmp.u32 = vec;
+
+    smem[transpose_idx]              = tmp.u16[0];
+    smem[smem_pitch + transpose_idx] = tmp.u16[1];
+}
+
+template<>
+__device__ __inline__ void write_smem_transpose(const float4& vec, float* smem, int transpose_idx, int smem_pitch)
+{
+    smem[transpose_idx]                  = vec.x;
+    smem[transpose_idx + 1]              = vec.z;
+    smem[smem_pitch + transpose_idx]     = vec.y;
+    smem[smem_pitch + transpose_idx + 1] = vec.w;
+}
+
+template<>
+__device__ __inline__ void write_smem_transpose(const uint32_t& vec, half* smem, int transpose_idx, int smem_pitch)
+{
+    union {
+        uint32_t u32;
+        half     u16[2];
+    } tmp;
+
+    tmp.u32                          = vec;
+    smem[transpose_idx]              = tmp.u16[0];
+    smem[smem_pitch + transpose_idx] = tmp.u16[1];
+}
+
+#ifdef ENABLE_BF16
+template<>
+__device__ __inline__ void
+write_smem_transpose(const __nv_bfloat162& vec, __nv_bfloat16* smem, int transpose_idx, int smem_pitch)
+{
+    smem[transpose_idx]              = vec.x;
+    smem[smem_pitch + transpose_idx] = vec.y;
+}
+
+template<>
+__device__ __inline__ void
+write_smem_transpose(const bf16_4_t& vec, __nv_bfloat16* smem, int transpose_idx, int smem_pitch)
+{
+    write_smem_transpose(reinterpret_cast<const uint2&>(vec), reinterpret_cast<uint16_t*>(smem), transpose_idx, smem_pitch);
+}
+
+template<>
+__device__ __inline__ void
+write_smem_transpose(const bf16_8_t& vec, __nv_bfloat16* smem, int transpose_idx, int smem_pitch)
+{
+    write_smem_transpose(reinterpret_cast<const uint4&>(vec), reinterpret_cast<uint16_t*>(smem), transpose_idx, smem_pitch);
+}
+#endif
+
+template<>
+__device__ __inline__ void write_smem_transpose(const float2& vec, float* smem, int transpose_idx, int smem_pitch)
+{
+    smem[transpose_idx]              = vec.x;
+    smem[smem_pitch + transpose_idx] = vec.y;
+}
+
+}  // namespace mmha

awq_cuda/attention/ft_attention.cpp

+// Adapted from NVIDIA/FasterTransformer and FlashAttention
+
+#include <torch/extension.h>
+#include "ATen/cuda/CUDAContext.h"
+#include <c10/cuda/CUDAGuard.h>
+
+#include "ft_attention.h"
+#include "decoder_masked_multihead_attention.h"
+
+#define CHECK_DEVICE(x) TORCH_CHECK(x.device().type() == torch::kCUDA, #x " must be on CUDA")
+#define CHECK_SHAPE(x, ...) TORCH_CHECK(x.sizes() == torch::IntArrayRef({__VA_ARGS__}), #x " must have shape (" #__VA_ARGS__ ")")
+#define CHECK_CONTIGUOUS(x) TORCH_CHECK(x.is_contiguous(), #x " must be contiguous")
+
+#define DISPATCH_FLOAT_AND_HALF_AND_BF16(TYPE, NAME, ...)                  \
+  if (TYPE == at::ScalarType::Half) {                                      \
+    using scalar_t = at::Half;                                             \
+    __VA_ARGS__();                                                         \
+  } else if (TYPE == at::ScalarType::BFloat16) {                           \
+    using scalar_t = at::BFloat16;                                         \
+    __VA_ARGS__();                                                         \
+  } else if (TYPE == at::ScalarType::Float)  {                             \
+    using scalar_t = float;                                                \
+    __VA_ARGS__();                                                         \
+  } else {                                                                 \
+    AT_ERROR(#NAME, " not implemented for type '", toString(TYPE), "'"); \
+  }
+
+template<typename T>
+void masked_multihead_attention(const Masked_multihead_attention_params<T>& params,
+                                const cudaStream_t& stream);
+
+template<typename T>
+void cross_multihead_attention(const Masked_multihead_attention_params<T>& params,
+                               const cudaStream_t& stream);
+
+template<typename T>
+struct SATypeConverter {
+    using Type = T;
+};
+
+template<>
+struct SATypeConverter<at::Half> {
+    using Type = uint16_t;
+};
+
+template<>
+struct SATypeConverter<at::BFloat16> {
+    using Type = __nv_bfloat16;
+};
+
+template <typename T>
+void set_params(Masked_multihead_attention_params<T> &params,
+                const size_t batch_size,
+                const size_t nheads,
+                const size_t nheads_kv,
+                const size_t memory_max_seqlen,
+                const size_t headdim,
+                const int timestep,
+                const int rotary_embedding_dim,
+                const float rotary_base,
+                const bool neox_rotary_style,
+                const int qkv_batch_stride,
+                T *q_ptr,
+                T *k_ptr,
+                T *v_ptr,
+                T *k_cache_ptr,
+                T *v_cache_ptr,
+                int *length_per_sample,
+                float *alibi_slopes_ptr,
+                T *out_ptr) {
+    // Reset the parameters
+    memset(&params, 0, sizeof(params));
+    params.q = q_ptr;
+    params.k = k_ptr;
+    params.v = v_ptr;
+    params.q_bias = nullptr;
+    params.k_bias = nullptr;
+    params.v_bias = nullptr;
+    params.k_cache = k_cache_ptr;
+    params.v_cache = v_cache_ptr;
+    params.linear_bias_slopes = alibi_slopes_ptr;
+    params.out = out_ptr;
+    params.cache_indir = nullptr;
+    params.stride = qkv_batch_stride;
+    params.batch_size = batch_size;
+    params.beam_width = 1;
+    params.memory_max_len = memory_max_seqlen;
+    params.num_heads = nheads;
+    params.num_kv_heads = nheads_kv;
+    params.hidden_size_per_head = headdim;
+    params.rotary_embedding_dim = rotary_embedding_dim;
+    params.rotary_base = rotary_base;
+    params.neox_rotary_style = neox_rotary_style;
+    params.timestep = timestep;
+    params.inv_sqrt_dh = 1.f / sqrt(float(headdim));
+    params.total_padding_tokens = nullptr;
+    params.masked_tokens = nullptr;
+    params.prefix_prompt_lengths = nullptr;
+    params.max_prefix_prompt_length = 0;
+    params.relative_attention_bias = nullptr;
+    params.relative_attention_bias_stride = 0;
+    params.cross_attention_out = nullptr;
+    params.max_decoder_seq_len = 0;
+    params.is_return_cross_attentions = false;
+    params.finished = nullptr;
+    params.memory_length_per_sample = nullptr;
+    params.length_per_sample = length_per_sample;
+}
+
+torch::Tensor single_query_attention(const torch::Tensor q,
+                                     const torch::Tensor k,
+                                     const torch::Tensor v,
+                                     torch::Tensor k_cache,
+                                     torch::Tensor v_cache,
+                                     c10::optional<const torch::Tensor> length_per_sample_,
+                                     c10::optional<const torch::Tensor> alibi_slopes_,
+                                     const int timestep,
+                                     const int rotary_embedding_dim,
+                                     const float rotary_base,
+                                     // neox_rotary_style = not interleaved
+                                     const bool neox_rotary_style) {
+    CHECK_DEVICE(q); CHECK_DEVICE(k); CHECK_DEVICE(v); CHECK_DEVICE(k_cache); CHECK_DEVICE(v_cache);
+    int batch_size = v_cache.size(0);
+    int nheads = q.size(1);
+    int nheads_kv = v_cache.size(1);
+    int memory_max_seqlen = v_cache.size(2);
+    int headdim = v_cache.size(3);
+    CHECK_SHAPE(q, batch_size, nheads, headdim);
+    CHECK_SHAPE(k, batch_size, nheads_kv, headdim);
+    CHECK_SHAPE(v, batch_size, nheads_kv, headdim);
+    CHECK_SHAPE(v_cache, batch_size, nheads_kv, memory_max_seqlen, headdim);
+    // k_cache shape: [B, H, Dh/x, L, x] where x=8 for fp16 and x=4 for fp32
+    int packsize = k_cache.dtype() == torch::kFloat32 ? 4 : 8;
+    CHECK_SHAPE(k_cache, batch_size, nheads_kv, headdim / packsize, memory_max_seqlen, packsize);
+    TORCH_CHECK(q.stride(2) == 1 && q.stride(1) == headdim);
+    TORCH_CHECK(k.stride(2) == 1 && k.stride(1) == headdim);
+    TORCH_CHECK(v.stride(2) == 1 && v.stride(1) == headdim);
+    // TORCH_CHECK(q.stride(0) == k.stride(0) && q.stride(0) == v.stride(0));
+    CHECK_CONTIGUOUS(v_cache); CHECK_CONTIGUOUS(k_cache);
+
+    if (length_per_sample_.has_value()) {
+        auto length_per_sample = length_per_sample_.value();
+        CHECK_DEVICE(length_per_sample);
+        CHECK_SHAPE(length_per_sample, batch_size);
+        CHECK_CONTIGUOUS(length_per_sample);
+        TORCH_CHECK(length_per_sample.dtype() == torch::kInt32);
+    }
+
+    if (alibi_slopes_.has_value()) {
+      auto alibi_slopes = alibi_slopes_.value();
+      CHECK_DEVICE(alibi_slopes);
+      CHECK_SHAPE(alibi_slopes, nheads);
+      CHECK_CONTIGUOUS(alibi_slopes); 
+      TORCH_CHECK(alibi_slopes.dtype() == torch::kFloat32);
+    }
+
+    // Otherwise the kernel will be launched from cuda:0 device
+    // Cast to char to avoid compiler warning about narrowing
+    at::cuda::CUDAGuard device_guard{(char)q.get_device()};
+
+    torch::Tensor out = torch::empty_like(q);
+
+    DISPATCH_FLOAT_AND_HALF_AND_BF16(q.scalar_type(), "single_query_attention", [&] {
+        using DataType = typename SATypeConverter<scalar_t>::Type;
+        Masked_multihead_attention_params<DataType> params;
+        set_params(params, batch_size, nheads, nheads_kv, memory_max_seqlen, headdim, 
+                   timestep, rotary_embedding_dim, rotary_base, neox_rotary_style, q.stride(0),
+                   reinterpret_cast<DataType*>(q.data_ptr()),
+                   reinterpret_cast<DataType*>(k.data_ptr()),
+                   reinterpret_cast<DataType*>(v.data_ptr()),
+                   reinterpret_cast<DataType*>(k_cache.data_ptr()),
+                   reinterpret_cast<DataType*>(v_cache.data_ptr()),
+                   length_per_sample_.has_value()
+                       ? length_per_sample_.value().data_ptr<int>() : nullptr,
+                   alibi_slopes_.has_value() 
+                       ? alibi_slopes_.value().data_ptr<float>(): nullptr,
+                   reinterpret_cast<DataType*>(out.data_ptr()));
+        auto stream = at::cuda::getCurrentCUDAStream();
+        masked_multihead_attention(params, stream);
+    });
+    return out;
+}
\ No newline at end of file

awq_cuda/attention/ft_attention.h

+#pragma once
+#include <torch/extension.h>
+
+
+torch::Tensor single_query_attention(const torch::Tensor q,
+                                     const torch::Tensor k,
+                                     const torch::Tensor v,
+                                     torch::Tensor k_cache,
+                                     torch::Tensor v_cache,
+                                     c10::optional<const torch::Tensor> length_per_sample_,
+                                     c10::optional<const torch::Tensor> alibi_slopes_,
+                                     const int timestep,
+                                     const int rotary_embedding_dim = 0,
+                                     const float rotary_base = 10000.0f,
+                                     const bool neox_rotary_style=true);
\ No newline at end of file

awq_cuda/position_embedding/pos_encoding.h

 #pragma once
 #include <torch/extension.h>
 
-void rotary_embedding(
+void rotary_embedding_neox(
   torch::Tensor& positions,
   torch::Tensor& query,
   torch::Tensor& key,
   int head_size,
-  torch::Tensor& cos_sin_cache,
-  bool is_neox);
\ No newline at end of file
+  torch::Tensor& cos_sin_cache);
\ No newline at end of file

awq_cuda/position_embedding/pos_encoding_kernels.cu

@@ -9,56 +9,15 @@ https://github.com/vllm-project/vllm/blob/main/csrc/pos_encoding_kernels.cu
 #include <ATen/cuda/CUDAContext.h>
 #include "pos_encoding.h"
 
-#define VLLM_DISPATCH_CASE_FLOATING_TYPES(...)              \
-  AT_DISPATCH_CASE(at::ScalarType::Float, __VA_ARGS__)      \
-  AT_DISPATCH_CASE(at::ScalarType::Half, __VA_ARGS__)       \
-  AT_DISPATCH_CASE(at::ScalarType::BFloat16, __VA_ARGS__)
-
-#define VLLM_DISPATCH_FLOATING_TYPES(TYPE, NAME, ...)             \
-  AT_DISPATCH_SWITCH(                                             \
-    TYPE, NAME, VLLM_DISPATCH_CASE_FLOATING_TYPES(__VA_ARGS__))
-
-template<typename scalar_t, bool IS_NEOX>
-inline __device__ void apply_rotary_embedding(
-  scalar_t* __restrict__ arr,
-  const scalar_t* __restrict__ cos_ptr,
-  const scalar_t* __restrict__ sin_ptr,
-  int rot_offset,
-  int embed_dim)
-{
-  int x_index, y_index;
-  scalar_t cos, sin;
-  if (IS_NEOX) {
-    // GPT-NeoX style rotary embedding.
-    x_index = rot_offset;
-    y_index = embed_dim + rot_offset;
-    cos = __ldg(cos_ptr + x_index);
-    sin = __ldg(sin_ptr + x_index);
-  } else {
-    // GPT-J style rotary embedding.
-    x_index = 2 * rot_offset;
-    y_index = 2 * rot_offset + 1;
-    cos = __ldg(cos_ptr + x_index / 2);
-    sin = __ldg(sin_ptr + x_index / 2);
-  }
-
-  const scalar_t x = arr[x_index];
-  const scalar_t y = arr[y_index];
-  arr[x_index] = x * cos - y * sin;
-  arr[y_index] = y * cos + x * sin;
-}
-
-template<typename scalar_t, bool IS_NEOX>
-__global__ void rotary_embedding_kernel(
+template<typename scalar_t>
+__global__ void rotary_embedding_neox_kernel(
   const int64_t* __restrict__ positions,        // [num_tokens]
   scalar_t* __restrict__ query,                 // [num_tokens, num_heads, head_size]
-  scalar_t* __restrict__ key,                   // [num_tokens, num_kv_heads, head_size]
+  scalar_t* __restrict__ key,                   // [num_tokens, num_heads, head_size]
   const scalar_t* __restrict__ cos_sin_cache,   // [max_position, 2, rot_dim // 2]
   const int rot_dim,
-  const int query_stride,
-  const int key_stride,
+  const int stride,
   const int num_heads,
-  const int num_kv_heads,
   const int head_size) {
   // Each thread block is responsible for one token.
   const int token_idx = blockIdx.x;
@@ -66,72 +25,64 @@ __global__ void rotary_embedding_kernel(
   const scalar_t* cache_ptr = cos_sin_cache + pos * rot_dim;
 
   const int embed_dim = rot_dim / 2;
-  const scalar_t* cos_ptr = cache_ptr;
-  const scalar_t* sin_ptr = cache_ptr + embed_dim;
-
-  const int nq = num_heads * embed_dim;
-  for (int i = threadIdx.x; i < nq; i += blockDim.x) {
+  const int n = num_heads * embed_dim;
+  for (int i = threadIdx.x; i < n; i += blockDim.x) {
     const int head_idx = i / embed_dim;
-    const int token_head = token_idx * query_stride + head_idx * head_size;
-    const int rot_offset = i % embed_dim;
-    apply_rotary_embedding<scalar_t, IS_NEOX>(query + token_head, cos_ptr,
-                                              sin_ptr, rot_offset, embed_dim);
-  }
+    const int token_head = token_idx * stride + head_idx * head_size;
 
-  const int nk = num_kv_heads * embed_dim;
-  for (int i = threadIdx.x; i < nk; i += blockDim.x) {
-    const int head_idx = i / embed_dim;
-    const int token_head = token_idx * key_stride + head_idx * head_size;
     const int rot_offset = i % embed_dim;
-    apply_rotary_embedding<scalar_t, IS_NEOX>(key + token_head, cos_ptr,
-                                              sin_ptr, rot_offset, embed_dim);
+    const int x_index = rot_offset;
+    const int y_index = embed_dim + rot_offset;
+
+    const int out_x = token_idx * stride + head_idx * head_size + x_index;
+    const int out_y = token_idx * stride + head_idx * head_size + y_index;
+
+    const scalar_t cos = __ldg(cache_ptr + x_index);
+    const scalar_t sin = __ldg(cache_ptr + y_index);
+
+    const scalar_t q_x = query[token_head + x_index];
+    const scalar_t q_y = query[token_head + y_index];
+    query[out_x] = q_x * cos - q_y * sin;
+    query[out_y] = q_y * cos + q_x * sin;
+
+    const scalar_t k_x = key[token_head + x_index];
+    const scalar_t k_y = key[token_head + y_index];
+    key[out_x] = k_x * cos - k_y * sin;
+    key[out_y] = k_y * cos + k_x * sin;
   }
 }
-void rotary_embedding(
-  torch::Tensor& positions,         // [num_tokens]
-  torch::Tensor& query,             // [num_tokens, num_heads * head_size]
-  torch::Tensor& key,               // [num_tokens, num_kv_heads * head_size]
+
+void rotary_embedding_neox(
+  torch::Tensor& positions,         // [b, num_tokens]
+  torch::Tensor& query,             // [b, num_tokens, 1, num_heads, head_size]
+  torch::Tensor& key,               // [b, num_tokens, 1, num_heads, head_size]
   int head_size,
-  torch::Tensor& cos_sin_cache,     // [max_position, rot_dim]
-  bool is_neox) {
-  int num_tokens = query.size(0);
+  torch::Tensor& cos_sin_cache)     // [max_position, rot_dim]
+{
+  int num_tokens = query.size(0) * query.size(1);
   int rot_dim = cos_sin_cache.size(1);
-  int num_heads = query.size(1) / head_size;
-  int num_kv_heads = key.size(1) / head_size;
-  int query_stride = query.stride(0);
-  int key_stride = key.stride(0);
+  int num_heads = query.size(-2);
+  int stride = num_heads * head_size;
+  // TORCH_CHECK(stride == key.stride(0));
 
   dim3 grid(num_tokens);
   dim3 block(std::min(num_heads * rot_dim / 2, 512));
   const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
-  VLLM_DISPATCH_FLOATING_TYPES(
+  AT_DISPATCH_FLOATING_TYPES_AND2(
+    at::ScalarType::Half,
+    at::ScalarType::BFloat16,
     query.scalar_type(),
-    "rotary_embedding",
+    "rotary_embedding_neox",
     [&] {
-      if (is_neox) {
-        rotary_embedding_kernel<scalar_t, true><<<grid, block, 0, stream>>>(
-          positions.data_ptr<int64_t>(),
-          query.data_ptr<scalar_t>(),
-          key.data_ptr<scalar_t>(),
-          cos_sin_cache.data_ptr<scalar_t>(),
-          rot_dim,
-          query_stride,
-          key_stride,
-          num_heads,
-          num_kv_heads,
-          head_size);
-      } else {
-        rotary_embedding_kernel<scalar_t, false><<<grid, block, 0, stream>>>(
-          positions.data_ptr<int64_t>(),
-          query.data_ptr<scalar_t>(),
-          key.data_ptr<scalar_t>(),
-          cos_sin_cache.data_ptr<scalar_t>(),
-          rot_dim,
-          query_stride,
-          key_stride,
-          num_heads,
-          num_kv_heads,
-          head_size);
-      }
+      rotary_embedding_neox_kernel<scalar_t><<<grid, block, 0, stream>>>(
+        positions.data_ptr<int64_t>(),
+        query.data_ptr<scalar_t>(),
+        key.data_ptr<scalar_t>(),
+        cos_sin_cache.data_ptr<scalar_t>(),
+        rot_dim,
+        stride,
+        num_heads,
+        head_size);
     });
-}
\ No newline at end of file
+}
+

awq_cuda/pybind.cpp

 #include <pybind11/pybind11.h>
 #include <torch/extension.h>
+#include "attention/ft_attention.h"
 #include "layernorm/layernorm.h"
 #include "quantization/gemm_cuda.h"
+#include "quantization/gemv_cuda.h"
 #include "position_embedding/pos_encoding.h"
 
 PYBIND11_MODULE(TORCH_EXTENSION_NAME, m)
 {
     m.def("layernorm_forward_cuda", &layernorm_forward_cuda, "FasterTransformer layernorm kernel");
     m.def("gemm_forward_cuda", &gemm_forward_cuda, "Quantized GEMM kernel.");
-    m.def("rotary_embedding", &rotary_embedding, "Apply rotary embedding to query and key");
-}
+    m.def("gemv_forward_cuda", &gemv_forward_cuda, "Quantized GEMV kernel.");
+    m.def("rotary_embedding_neox", &rotary_embedding_neox, "Apply GPT-NeoX style rotary embedding to query and key");
+    m.def("single_query_attention", &single_query_attention, "Attention with a single query",
+          py::arg("q"), py::arg("k"), py::arg("v"), py::arg("k_cache"), py::arg("v_cache"),
+          py::arg("length_per_sample_"), py::arg("alibi_slopes_"), py::arg("timestep"), py::arg("rotary_embedding_dim")=0,
+          py::arg("rotary_base")=10000.0f, py::arg("neox_rotary_style")=true);
+}
\ No newline at end of file

awq_cuda/quantization/gemm_cuda_gen.cu

@@ -10,6 +10,7 @@
  */
 
 #include <torch/extension.h>
+#include <ATen/cuda/CUDAContext.h>
 #include "gemm_cuda.h"
 #include "dequantize.cuh"
 #include <cuda_fp16.h>
@@ -439,6 +440,7 @@ torch::Tensor gemm_forward_cuda(
     auto scaling_factors = reinterpret_cast<half*>(_scaling_factors.data_ptr<at::Half>());
     auto zeros = reinterpret_cast<int*>(_zeros.data_ptr<int>());
     int group_size = num_in_channels / _scaling_factors.size(0);
+    const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
 
     if (num_out_channels % 64 != 0)
         throw std::invalid_argument("OC is not multiple of cta_N = 64");
@@ -456,7 +458,7 @@ torch::Tensor gemm_forward_cuda(
         // threadIdx.x: 32
         // threadIdx.y: i_factors[2] * j_factors[2]
         dim3 threads_per_block(32, 2);
-        gemm_forward_4bit_cuda_m16n128k32<<<num_blocks, threads_per_block>>>(
+        gemm_forward_4bit_cuda_m16n128k32<<<num_blocks, threads_per_block, 0, stream>>>(
             group_size, split_k_iters, in_feats, kernel, scaling_factors, zeros, num_in_feats, num_in_channels, num_out_channels, out_feats);
     }
     else if (num_out_channels % 64 == 0)
@@ -467,7 +469,7 @@ torch::Tensor gemm_forward_cuda(
         // threadIdx.x: 32
         // threadIdx.y: i_factors[2] * j_factors[2]
         dim3 threads_per_block(32, 2);
-        gemm_forward_4bit_cuda_m16n64k32<<<num_blocks, threads_per_block>>>(
+        gemm_forward_4bit_cuda_m16n64k32<<<num_blocks, threads_per_block, 0, stream>>>(
             group_size, split_k_iters, in_feats, kernel, scaling_factors, zeros, num_in_feats, num_in_channels, num_out_channels, out_feats);
     }
     return _out_feats.sum(0);

awq_cuda/quantization/gemv_cuda.cu

+// Inspired by https://github.com/ankan-ban/llama_cu_awq
+/*
+
+@article{lin2023awq,
+  title={AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration},
+  author={Lin, Ji and Tang, Jiaming and Tang, Haotian and Yang, Shang and Dang, Xingyu and Han, Song},
+  journal={arXiv},
+  year={2023}
+}
+
+*/
+
+#include <cuda_fp16.h>
+#include <stdio.h>
+#include <torch/extension.h>
+#include <ATen/cuda/CUDAContext.h>
+#include "gemv_cuda.h"
+#define VECTORIZE_FACTOR 8
+#define Q_VECTORIZE_FACTOR 8
+#define PACK_FACTOR 8
+#define WARP_SIZE 32
+
+
+// Reduce sum within the warp using the tree reduction algorithm.
+__device__ __forceinline__ float warp_reduce_sum(float sum) {
+  #pragma unroll
+  for(int i = 4; i >= 0; i--){
+    sum += __shfl_down_sync(0xffffffff, sum, 1<<i);
+  }
+  /*
+  // Equivalent to the following tree reduction implementation:
+  sum += __shfl_down_sync(0xffffffff, sum, 16);
+  sum += __shfl_down_sync(0xffffffff, sum, 8);
+  sum += __shfl_down_sync(0xffffffff, sum, 4);
+  sum += __shfl_down_sync(0xffffffff, sum, 2);
+  sum += __shfl_down_sync(0xffffffff, sum, 1);
+  */
+  return sum;
+}
+
+__device__ __forceinline__ int make_divisible(int c, int divisor){
+  return (c + divisor - 1) / divisor;
+}
+
+
+/*
+Computes GEMV (group_size = 64).
+
+Args:
+  inputs: vector of shape [batch_size, IC];
+  weight: matrix of shape [OC, IC / 8];
+  output: vector of shape [OC];
+  zeros: matrix of shape [OC, IC / group_size / 8];
+  scaling_factors: matrix of shape [OC, IC / group_size];
+
+Notes:
+  One cannot infer group_size from the shape of scaling factors.
+  the second dimension is rounded up to a multiple of PACK_FACTOR.
+*/
+__global__ void gemv_kernel_g64(
+  const float4* _inputs, const uint32_t* weight, const uint32_t* zeros, const half* scaling_factors, half* _outputs, 
+  const int IC, const int OC){
+    const int group_size = 64;
+    float psum = 0;
+    const int batch_idx = blockIdx.z;
+    const int oc_idx = blockIdx.y * blockDim.y + threadIdx.y; 
+    const float4* inputs = _inputs + batch_idx * IC / PACK_FACTOR;
+    half* outputs = _outputs + batch_idx * OC;
+    // This is essentially zeros_w.
+    const int num_groups_packed = make_divisible(make_divisible(IC / group_size, PACK_FACTOR), 2) * 2;
+    const int weight_w = IC / PACK_FACTOR;
+    // TODO (Haotian): zeros_w is incorrect, after fixing we got misaligned address
+    const int zeros_w = make_divisible(make_divisible(IC / group_size, PACK_FACTOR), 2) * 2;
+    // consistent with input shape
+    const int sf_w = make_divisible(make_divisible(IC / group_size, PACK_FACTOR), 2) * 2 * PACK_FACTOR;
+    // if(blockIdx.x == 0 && blockIdx.y == 0 && threadIdx.x == 0 && threadIdx.y == 0) printf("%d %d %d %d %d\n", IC, group_size, PACK_FACTOR, zeros_w, sf_w);
+    // tile size: 4 OC x 1024 IC per iter
+    for(int packed_group_idx = 0; packed_group_idx < num_groups_packed / 2; packed_group_idx++){
+      // 1024 numbers in one iteration across warp. Need 1024 / group_size zeros.
+      uint64_t packed_zeros = *reinterpret_cast<const uint64_t*>(zeros + oc_idx * zeros_w + packed_group_idx * 2);
+      uint32_t packed_weights[4];
+      // use float4 to load weights, each thread load 32 int4 numbers (1 x float4)
+      *((float4*)(packed_weights)) = *((float4*)(weight + oc_idx * weight_w + packed_group_idx * (WARP_SIZE * 4) + threadIdx.x * 4));
+      // load scaling factors
+      // g64: two threads -> 64 numbers -> 1 group; 1 warp = 16 groups.
+      float scaling_factor = __half2float(scaling_factors[oc_idx * sf_w + packed_group_idx * 16 + (threadIdx.x / 2)]);
+      float current_zeros = (float)((packed_zeros >> (threadIdx.x / 2 * 4)) & 0xF);
+      int inputs_ptr_delta = packed_group_idx * WARP_SIZE * 4 + threadIdx.x * 4; 
+      const float4* inputs_ptr = inputs + inputs_ptr_delta;
+      // multiply 32 weights with 32 inputs
+      #pragma unroll
+      for (int ic_0 = 0; ic_0 < 4; ic_0++){
+        // iterate over different uint32_t packed_weights in this loop
+        uint32_t current_packed_weight = packed_weights[ic_0];
+        half packed_inputs[PACK_FACTOR];
+        // each thread load 8 inputs, starting index is packed_group_idx * 128 * 8 (because each iter loads 128*8)
+        if (inputs_ptr_delta + ic_0 < IC / PACK_FACTOR) {
+          *((float4*)packed_inputs) = *(inputs_ptr + ic_0);
+          #pragma unroll
+          for (int ic_1 = 0; ic_1 < PACK_FACTOR; ic_1++){
+            // iterate over 8 numbers packed within each uint32_t number
+            float current_single_weight_fp = (float)(current_packed_weight & 0xF);
+            float dequantized_weight = scaling_factor * (current_single_weight_fp - current_zeros);
+            //if(blockIdx.x == 0 && blockIdx.y == 0 && threadIdx.x == 0 && threadIdx.y == 0 && ic_0 == 0 && ic_1 == 0 && packed_group_idx == 0) printf("%f %f %f %f %X %X\n", dequantized_weight, current_single_weight_fp, scaling_factor, current_zeros, current_packed_weight, packed_zeros);
+            psum += dequantized_weight * __half2float(packed_inputs[ic_1]);
+            current_packed_weight = current_packed_weight >> 4;
+          }
+        }
+      }
+    }
+    psum = warp_reduce_sum(psum);
+    if (threadIdx.x == 0) {
+     outputs[oc_idx] = __float2half(psum); 
+    }
+}
+
+
+/*
+Computes GEMV (group_size = 128).
+
+Args:
+  inputs: vector of shape [batch_size, IC];
+  weight: matrix of shape [OC, IC / 8];
+  output: vector of shape [OC];
+  zeros: matrix of shape [OC, IC / group_size / 8];
+  scaling_factors: matrix of shape [OC, IC / group_size];
+
+Notes:
+  One cannot infer group_size from the shape of scaling factors.
+  the second dimension is rounded up to a multiple of PACK_FACTOR.
+*/
+__global__ void gemv_kernel_g128(
+  const float4* _inputs, const uint32_t* weight, const uint32_t* zeros, const half* scaling_factors, half* _outputs, 
+  const int IC, const int OC){
+    const int group_size = 128;
+    float psum = 0;
+    const int batch_idx = blockIdx.z;
+    const int oc_idx = blockIdx.y * blockDim.y + threadIdx.y; 
+    const float4* inputs = _inputs + batch_idx * IC / PACK_FACTOR;
+    half* outputs = _outputs + batch_idx * OC;
+    const int num_groups_packed = make_divisible(IC / group_size, PACK_FACTOR);
+    const int weight_w = IC / PACK_FACTOR;
+    // TODO (Haotian): zeros_w is incorrect, after fixing we got misaligned address
+    const int zeros_w = make_divisible(IC / group_size, PACK_FACTOR);
+    // consistent with input shape
+    const int sf_w = make_divisible(IC / group_size, PACK_FACTOR) * PACK_FACTOR;
+    //if(blockIdx.x == 0 && blockIdx.y == 0 && threadIdx.x == 0 && threadIdx.y == 0) printf("%d %d %d %d\n", IC, group_size, PACK_FACTOR, zeros_w);
+    // tile size: 4 OC x 1024 IC per iter
+    for(int packed_group_idx = 0; packed_group_idx < num_groups_packed; packed_group_idx++){
+      // 1024 numbers in one iteration across warp. Need 1024 / group_size zeros.
+      uint32_t packed_zeros = *(zeros + oc_idx * zeros_w + packed_group_idx);
+      uint32_t packed_weights[4];
+      // use float4 to load weights, each thread load 32 int4 numbers (1 x float4)
+      *((float4*)(packed_weights)) = *((float4*)(weight + oc_idx * weight_w + packed_group_idx * (WARP_SIZE * 4) + threadIdx.x * 4));
+      // load scaling factors
+      // g128: four threads -> 128 numbers -> 1 group; 1 warp = 8 groups.
+      float scaling_factor = __half2float(scaling_factors[oc_idx * sf_w + packed_group_idx * 8 + (threadIdx.x / 4)]);
+      float current_zeros = (float)((packed_zeros >> (threadIdx.x / 4 * 4)) & 0xF);
+      int inputs_ptr_delta = packed_group_idx * WARP_SIZE * 4 + threadIdx.x * 4; 
+      const float4* inputs_ptr = inputs + inputs_ptr_delta;
+      // multiply 32 weights with 32 inputs
+      #pragma unroll
+      for (int ic_0 = 0; ic_0 < 4; ic_0++){
+        // iterate over different uint32_t packed_weights in this loop
+        uint32_t current_packed_weight = packed_weights[ic_0];
+        half packed_inputs[PACK_FACTOR];
+        // each thread load 8 inputs, starting index is packed_group_idx * 128 * 8 (because each iter loads 128*8)
+        if (inputs_ptr_delta + ic_0 < IC / PACK_FACTOR) {
+          *((float4*)packed_inputs) = *(inputs_ptr + ic_0);
+          #pragma unroll
+          for (int ic_1 = 0; ic_1 < PACK_FACTOR; ic_1++){
+            // iterate over 8 numbers packed within each uint32_t number
+            float current_single_weight_fp = (float)(current_packed_weight & 0xF);
+            float dequantized_weight = scaling_factor * (current_single_weight_fp - current_zeros);
+            //if(blockIdx.x == 0 && blockIdx.y == 0 && threadIdx.x == 0 && threadIdx.y == 0 && ic_0 == 0 && ic_1 == 0 && packed_group_idx == 0) printf("%f %f %f %f %X %X\n", dequantized_weight, current_single_weight_fp, scaling_factor, current_zeros, current_packed_weight, packed_zeros);
+            psum += dequantized_weight * __half2float(packed_inputs[ic_1]);
+            current_packed_weight = current_packed_weight >> 4;
+          }
+        }
+      }
+    }
+    psum = warp_reduce_sum(psum);
+    if (threadIdx.x == 0) {
+     outputs[oc_idx] = __float2half(psum); 
+    }
+}
+
+
+/*
+Computes GEMV (PyTorch interface).
+
+Args:
+  _in_feats: tensor of shape [B, IC];
+  _kernel: int tensor of shape [OC, IC // 8];
+  _zeros: int tensor of shape [OC, IC // G // 8];
+  _scaling_factors: tensor of shape [OC, IC // G];
+  blockDim_x: size of thread block, dimension x, where blockDim_x * workload_per_thread = IC;
+  blockDim_y: size of thread block, dimension y, where blockDim_y * gridDim_y = OC;
+
+Returns:
+  out_feats: tensor of shape [B, OC];
+*/
+torch::Tensor gemv_forward_cuda(
+    torch::Tensor _in_feats,
+    torch::Tensor _kernel,
+    torch::Tensor _scaling_factors,
+    torch::Tensor _zeros,
+    int group_size)
+{
+    int num_in_feats = _in_feats.size(0);
+    int num_in_channels = _in_feats.size(1);
+    // int kernel_volume = _out_in_map.size(1);
+    auto in_feats = reinterpret_cast<float4*>(_in_feats.data_ptr<at::Half>());
+    auto kernel = reinterpret_cast<uint32_t*>(_kernel.data_ptr<int>());
+    auto zeros = reinterpret_cast<uint32_t*>(_zeros.data_ptr<int>());
+    auto scaling_factors = reinterpret_cast<half*>(_scaling_factors.data_ptr<at::Half>());
+    // auto out_in_map = _out_in_map.data_ptr<int>();
+    auto options =
+    torch::TensorOptions().dtype(_in_feats.dtype()).device(_in_feats.device());
+    // kernel is [OC, IC]
+    at::Tensor _out_feats = torch::empty({num_in_feats, _kernel.size(0)}, options);
+    int num_out_feats = _out_feats.size(-2);
+    int num_out_channels = _out_feats.size(-1);
+    auto out_feats = reinterpret_cast<half*>(_out_feats.data_ptr<at::Half>());
+    int blockDim_z = num_out_feats;
+    dim3 num_blocks(1, num_out_channels / 4, num_out_feats);
+    dim3 num_threads(32, 4);
+    const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
+    if (group_size == 64)
+    {
+      gemv_kernel_g64<<<num_blocks, num_threads, 0, stream>>>(
+        // pointers
+        in_feats, kernel, zeros, scaling_factors, out_feats,
+        // constants
+        num_in_channels, num_out_channels
+      );
+    }
+    else if (group_size == 128)
+    {
+      gemv_kernel_g128<<<num_blocks, num_threads, 0, stream>>>(
+        // pointers
+        in_feats, kernel, zeros, scaling_factors, out_feats,
+        // constants
+        num_in_channels, num_out_channels
+      );
+    }
+    return _out_feats;
+;}
+

awq_cuda/quantization/gemv_cuda.h

+#pragma once
+#include <torch/extension.h>
+
+torch::Tensor gemv_forward_cuda(
+    torch::Tensor _in_feats,
+    torch::Tensor _kernel,
+    torch::Tensor _scaling_factors,
+    torch::Tensor _zeros,
+    int group_size);

examples/basic_quant.py

@@ -3,7 +3,7 @@ from transformers import AutoTokenizer
 
 model_path = 'lmsys/vicuna-7b-v1.5'
 quant_path = 'vicuna-7b-v1.5-awq'
-quant_config = { "zero_point": True, "q_group_size": 128, "w_bit": 4 }
+quant_config = { "zero_point": True, "q_group_size": 128, "w_bit": 4, "version": "GEMM" }
 
 # Load model
 model = AutoAWQForCausalLM.from_pretrained(model_path)

examples/benchmark.py

+import time
+import torch
+import argparse
+import numpy as np
+import pandas as pd
+from awq import AutoAWQForCausalLM
+from transformers import AutoTokenizer
+from torch.cuda import OutOfMemoryError
+
+def warmup(model):
+    warm_up = torch.randn((4096,4096)).to(next(model.parameters()).device)
+    torch.mm(warm_up,warm_up)
+
+def generate(model, input_ids, n_generate):
+    context_time = 0
+    generate_time = []
+
+    with torch.inference_mode():
+        for i in range(n_generate):
+            torch.cuda.synchronize()
+            start = time.time()
+
+            if i == 0:
+                # prefill context
+                inputs = torch.as_tensor(input_ids, device=next(model.parameters()).device)
+            else:
+                # decode tokens
+                inputs = torch.as_tensor(token, device=next(model.parameters()).device)
+            
+            out = model(inputs, use_cache=True)
+
+            torch.cuda.synchronize()
+            token = out[0][:, -1].max(1)[1].unsqueeze(1)
+
+            if i == 0:
+                context_time += time.time() - start
+            else:
+                generate_time.append(time.time() - start)
+    
+    return context_time, generate_time
+
+def run_round(model_path, quant_file, n_generate, input_ids, batch_size):
+    print(f" -- Loading model...")
+    model = AutoAWQForCausalLM.from_quantized(
+        model_path, quant_file, fuse_layers=True,
+        max_new_tokens=n_generate, batch_size=batch_size
+    )
+
+    print(f" -- Warming up...")
+    warmup(model)
+
+    print(f" -- Generating {n_generate} tokens, {input_ids.shape[1]} in context...")
+    
+    try:
+        context_time, generate_time = generate(model, input_ids, n_generate)
+        successful_generate = True
+    except RuntimeError as ex:
+        if 'cuda out of memory' in str(ex).lower():
+            successful_generate = False
+        else:
+            raise RuntimeError(ex)
+    
+    device = next(model.parameters()).device
+    memory_used = torch.cuda.max_memory_allocated(device) / (1024 ** 3)
+    memory_pct = memory_used / (torch.cuda.get_device_properties(device).total_memory / (1024 ** 3)) * 100
+
+    if successful_generate:
+        # number of tokens in context / time for processing context * batch size
+        prefill_tokens_per_second = input_ids.shape[1] / context_time * batch_size
+        # 1 second / median time per token in seconds * batch size
+        decode_tokens_per_second = 1 / np.median(generate_time) * batch_size
+
+        print(f" ** Speed (Prefill): {prefill_tokens_per_second:.2f} tokens/second")
+        print(f" ** Speed (Decode): {decode_tokens_per_second:.2f} tokens/second")
+        print(f" ** Max Memory (VRAM): {memory_used:.2f} GB ({memory_pct:.2f}%)")
+    else:
+        prefill_tokens_per_second = 'OOM'
+        decode_tokens_per_second = 'OOM'
+
+    return {
+        "Batch Size": batch_size,
+        "Prefill Length": input_ids.shape[1],
+        "Decode Length": n_generate,
+        "Prefill tokens/s": prefill_tokens_per_second,
+        "Decode tokens/s": decode_tokens_per_second,
+        "Memory (VRAM)": f"{memory_used:.2f} GB ({memory_pct:.2f}%)"
+    }, model.quant_config["version"]
+
+def main(args):
+    rounds = [
+        {"context": 32, "n_generate": 32},
+        {"context": 64, "n_generate": 64},
+        {"context": 128, "n_generate": 128},
+        {"context": 256, "n_generate": 256},
+        {"context": 512, "n_generate": 512},
+        {"context": 1024, "n_generate": 1024},
+        {"context": 2048, "n_generate": 2048},
+    ]
+
+    all_stats = []
+    tokenizer = AutoTokenizer.from_pretrained(args.model_path, trust_remote_code=True)
+
+    for settings in rounds:
+        input_ids = torch.randint(0, tokenizer.vocab_size, (args.batch_size, settings["context"])).cuda()
+
+        stats, model_version = run_round(
+            args.model_path,
+            args.quant_file,
+            settings["n_generate"],
+            input_ids,
+            args.batch_size
+        )
+        
+        all_stats.append(stats)
+
+        if stats["Prefill tokens/s"] == 'OOM':
+            break
+    
+    df = pd.DataFrame(all_stats)
+    print('GPU:', torch.cuda.get_device_name())
+    print('Model:', args.model_path)
+    print('Version:', model_version)
+    print(df.to_markdown(index=False))
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--model_path", type=str, default="casperhansen/vicuna-7b-v1.5-awq", help="path to the model")
+    parser.add_argument("--quant_file", type=str, default="awq_model_w4_g128.pt", help="weights filename")
+    parser.add_argument("--batch_size", type=int, default=1, help="weights filename")
+    args = parser.parse_args()
+
+    main(args)
\ No newline at end of file

setup.py

@@ -44,14 +44,28 @@ requirements = [
     "toml",
     "attributedict",
     "protobuf",
-    "torchvision"
+    "torchvision",
+    "tabulate"
 ]
 
-include_dirs = []
+def get_include_dirs():
+    include_dirs = []
 
-conda_cuda_include_dir = os.path.join(get_python_lib(), "nvidia/cuda_runtime/include")
-if os.path.isdir(conda_cuda_include_dir):
-    include_dirs.append(conda_cuda_include_dir)
+    conda_cuda_include_dir = os.path.join(get_python_lib(), "nvidia/cuda_runtime/include")
+    if os.path.isdir(conda_cuda_include_dir):
+        include_dirs.append(conda_cuda_include_dir)
+    this_dir = os.path.dirname(os.path.abspath(__file__))
+    include_dirs.append(this_dir)
+
+    return include_dirs
+
+def get_generator_flag():
+    generator_flag = []
+    torch_dir = torch.__path__[0]
+    if os.path.exists(os.path.join(torch_dir, "include", "ATen", "CUDAGeneratorImpl.h")):
+        generator_flag = ["-DOLD_GENERATOR_PATH"]
+    
+    return generator_flag
 
 def check_dependencies():
     if CUDA_HOME is None:
@@ -77,6 +91,8 @@ def get_compute_capabilities():
     return capability_flags
 
 check_dependencies()
+include_dirs = get_include_dirs()
+generator_flags = get_generator_flag()
 arch_flags = get_compute_capabilities()
 
 if os.name == "nt":
@@ -86,8 +102,21 @@ if os.name == "nt":
     }
 else:
     extra_compile_args={
-        "cxx": ["-g", "-O3", "-fopenmp", "-lgomp", "-std=c++17"],
-        "nvcc": ["-O3", "-std=c++17"] + arch_flags
+        "cxx": ["-g", "-O3", "-fopenmp", "-lgomp", "-std=c++17", "-DENABLE_BF16"],
+        "nvcc": [
+            "-O3", 
+            "-std=c++17",
+            "-DENABLE_BF16",
+            "-U__CUDA_NO_HALF_OPERATORS__",
+            "-U__CUDA_NO_HALF_CONVERSIONS__",
+            "-U__CUDA_NO_BFLOAT16_OPERATORS__",
+            "-U__CUDA_NO_BFLOAT16_CONVERSIONS__",
+            "-U__CUDA_NO_BFLOAT162_OPERATORS__",
+            "-U__CUDA_NO_BFLOAT162_CONVERSIONS__",
+            "--expt-relaxed-constexpr",
+            "--expt-extended-lambda",
+            "--use_fast_math",
+        ] + arch_flags + generator_flags
     }
 
 extensions = [
@@ -97,7 +126,10 @@ extensions = [
             "awq_cuda/pybind.cpp",
             "awq_cuda/quantization/gemm_cuda_gen.cu",
             "awq_cuda/layernorm/layernorm.cu",
-            "awq_cuda/position_embedding/pos_encoding_kernels.cu"
+            "awq_cuda/position_embedding/pos_encoding_kernels.cu",
+            "awq_cuda/quantization/gemv_cuda.cu",
+            "awq_cuda/attention/ft_attention.cpp",
+            "awq_cuda/attention/decoder_masked_multihead_attention.cu"
         ], extra_compile_args=extra_compile_args
     )
 ]