ops.h

#pragma once

#include <optional>
#include <torch/library.h>

#include "core/scalar_type.hpp"

void paged_attention_v1(
    torch::Tensor& out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step);

void paged_attention_v2(
    torch::Tensor& out, torch::Tensor& exp_sums, torch::Tensor& max_logits,
    torch::Tensor& tmp_out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step);

void paged_attention_v1_opt(
    torch::Tensor& out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step);

void paged_attention_v2_opt(
    torch::Tensor& out, torch::Tensor& exp_sums, torch::Tensor& max_logits,
    torch::Tensor& tmp_out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step);

void paged_attention_v1_opt_tc(
    torch::Tensor& out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step);

void paged_attention_v2_opt_tc(
    torch::Tensor& out, torch::Tensor& exp_sums, torch::Tensor& max_logits,
    torch::Tensor& tmp_out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step);


// paged_attention with attn_masks
void paged_attention_v1_with_mask(
    torch::Tensor& out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step,
    const c10::optional<torch::Tensor>& attn_masks,
    const int64_t attn_masks_stride=0);

void paged_attention_v2_with_mask(
    torch::Tensor& out, torch::Tensor& exp_sums, torch::Tensor& max_logits,
    torch::Tensor& tmp_out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step,
    const c10::optional<torch::Tensor>& attn_masks,
    const int64_t attn_masks_stride=0);

void paged_attention_v1_opt_with_mask(
    torch::Tensor& out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step,
    const c10::optional<torch::Tensor>& attn_masks,
    const int64_t attn_masks_stride=0);

void paged_attention_v2_opt_with_mask(
    torch::Tensor& out, torch::Tensor& exp_sums, torch::Tensor& max_logits,
    torch::Tensor& tmp_out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step,
    const c10::optional<torch::Tensor>& attn_masks,
    const int64_t attn_masks_stride=0);

void paged_attention_v1_opt_tc_with_mask(
    torch::Tensor& out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step,
    const c10::optional<torch::Tensor>& attn_masks,
    const int64_t attn_masks_stride=0);

void paged_attention_v2_opt_tc_with_mask(
    torch::Tensor& out, torch::Tensor& exp_sums, torch::Tensor& max_logits,
    torch::Tensor& tmp_out, torch::Tensor& query, torch::Tensor& key_cache,
    torch::Tensor& value_cache, int64_t num_kv_heads, double scale,
    torch::Tensor& block_tables, torch::Tensor& seq_lens, int64_t block_size,
    int64_t max_seq_len, const c10::optional<torch::Tensor>& alibi_slopes,
    const std::string& kv_cache_dtype, double k_scale, double v_scale,
    const int64_t tp_rank, const int64_t blocksparse_local_blocks,
    const int64_t blocksparse_vert_stride, const int64_t blocksparse_block_size,
    const int64_t blocksparse_head_sliding_step,
    const c10::optional<torch::Tensor>& attn_masks,
    const int64_t attn_masks_stride=0);


void rms_norm(torch::Tensor& out, torch::Tensor& input, torch::Tensor& weight,
              double epsilon);

void fused_add_rms_norm(torch::Tensor& input, torch::Tensor& residual,
                        torch::Tensor& weight, double epsilon);

void rms_norm_opt(torch::Tensor& out, torch::Tensor& input, torch::Tensor& weight,
              double epsilon);

void fused_add_rms_norm_opt(torch::Tensor& input, torch::Tensor& residual,
                        torch::Tensor& weight, double epsilon);

void rotary_embedding(torch::Tensor& positions, torch::Tensor& query,
                      torch::Tensor& key, int64_t head_size,
                      torch::Tensor& cos_sin_cache, bool is_neox);

void batched_rotary_embedding(torch::Tensor& positions, torch::Tensor& query,
                              torch::Tensor& key, int64_t head_size,
                              torch::Tensor& cos_sin_cache, bool is_neox,
                              int64_t rot_dim,
                              torch::Tensor& cos_sin_cache_offsets);
void rotary_embedding_tgi(
  torch::Tensor& query,
  torch::Tensor& key,
  int64_t head_size,
  torch::Tensor& cos_cache,
  torch::Tensor& sin_cache,
  bool is_neox);

void silu_and_mul(torch::Tensor& out, torch::Tensor& input);

void gelu_and_mul(torch::Tensor& out, torch::Tensor& input);

void gelu_tanh_and_mul(torch::Tensor& out, torch::Tensor& input);

void silu_and_mul_opt(torch::Tensor& out, torch::Tensor& input);

void gelu_and_mul_opt(torch::Tensor& out, torch::Tensor& input);

void gelu_tanh_and_mul_opt(torch::Tensor& out, torch::Tensor& input);

void gelu_new(torch::Tensor& out, torch::Tensor& input);

void gelu_fast(torch::Tensor& out, torch::Tensor& input);

void gelu_quick(torch::Tensor& out, torch::Tensor& input);

void trans_w16_gemm(torch::Tensor dst, torch::Tensor src, int64_t row, int64_t col);

void advance_step_flashattn(int64_t num_seqs, int64_t num_queries,
                            int64_t block_size, torch::Tensor& input_tokens,
                            torch::Tensor& sampled_token_ids,
                            torch::Tensor& input_positions,
                            torch::Tensor& seq_lens,
                            torch::Tensor& slot_mapping,
                            torch::Tensor& block_tables);

void advance_step_flashinfer(
    int64_t num_seqs, int64_t num_queries, int64_t block_size,
    torch::Tensor& input_tokens, torch::Tensor& sampled_token_ids,
    torch::Tensor& input_positions, torch::Tensor& seq_lens,
    torch::Tensor& slot_mapping, torch::Tensor& block_tables,
    torch::Tensor& paged_kv_indices, torch::Tensor& paged_kv_indptr,
    torch::Tensor& paged_kv_last_page_len, torch::Tensor& block_table_bounds);

#ifndef USE_ROCM
torch::Tensor aqlm_gemm(const torch::Tensor& input, const torch::Tensor& codes,
                        const torch::Tensor& codebooks,
                        const torch::Tensor& scales,
                        const std::vector<int64_t>& codebook_partition_sizes,
                        const std::optional<torch::Tensor>& bias);

torch::Tensor aqlm_dequant(
    const torch::Tensor& codes, const torch::Tensor& codebooks,
    const std::vector<int64_t>& codebook_partition_sizes);

torch::Tensor awq_gemm(torch::Tensor _in_feats, torch::Tensor _kernel,
                       torch::Tensor _scaling_factors, torch::Tensor _zeros,
                       int64_t split_k_iters);

torch::Tensor awq_dequantize(torch::Tensor _kernel,
                             torch::Tensor _scaling_factors,
                             torch::Tensor _zeros, int64_t split_k_iters,
                             int64_t thx, int64_t thy);

torch::Tensor marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
                          torch::Tensor& b_scales, torch::Tensor& workspace,
                          int64_t size_m, int64_t size_n, int64_t size_k);

namespace machete {

std::vector<std::string> supported_schedules(
    vllm::ScalarTypeTorchPtr const& btype);

torch::Tensor gemm(torch::Tensor const& A, torch::Tensor const& B,
                   vllm::ScalarTypeTorchPtr const& btype,
                   c10::optional<torch::Tensor> const& scales,
                   c10::optional<torch::Tensor> const& zeros,
                   c10::optional<int64_t> group_size,
                   c10::optional<torch::Tensor> const& C,
                   c10::optional<double> alpha, c10::optional<double> beta,
                   c10::optional<std::string> schedule);

torch::Tensor prepack_B(torch::Tensor const& B,
                        vllm::ScalarTypeTorchPtr const& btype);

};  // namespace machete

torch::Tensor permute_cols(torch::Tensor const& A, torch::Tensor const& perm);

torch::Tensor gptq_marlin_24_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
                                  torch::Tensor& b_meta,
                                  torch::Tensor& b_scales,
                                  torch::Tensor& workspace,
                                  vllm::ScalarTypeTorchPtr const& b_q_type,
                                  int64_t size_m, int64_t size_n,
                                  int64_t size_k);

torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
                               torch::Tensor& b_scales, torch::Tensor& b_zeros,
                               torch::Tensor& g_idx, torch::Tensor& perm,
                               torch::Tensor& workspace,
                               vllm::ScalarTypeTorchPtr const& b_q_type,
                               int64_t size_m, int64_t size_n, int64_t size_k,
                               bool is_k_full, bool has_zp,
                               bool use_fp32_reduce);

torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
                                 int64_t size_k, int64_t size_n,
                                 int64_t num_bits);

torch::Tensor gptq_marlin_repack_meta(torch::Tensor& b_q_weight,
                                      torch::Tensor& perm, c10::SymInt size_k,
                                      c10::SymInt size_n, int64_t num_bits);

torch::Tensor awq_marlin_repack(torch::Tensor& b_q_weight, int64_t size_k,
                                int64_t size_n, int64_t num_bits);

torch::Tensor awq_marlin_repack_meta(torch::Tensor& b_q_weight,
                                     c10::SymInt size_k, c10::SymInt size_n,
                                     int64_t num_bits);

torch::Tensor ggml_dequantize(torch::Tensor W, int64_t type, int64_t m,
                              int64_t n);

torch::Tensor ggml_mul_mat_vec_a8(torch::Tensor W, torch::Tensor X,
                                  int64_t type, int64_t row);

torch::Tensor ggml_mul_mat_a8(torch::Tensor W, torch::Tensor X, int64_t type,
                              int64_t row);

torch::Tensor fp8_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
                              torch::Tensor& b_scales, torch::Tensor& workspace,
                              int64_t num_bits, int64_t size_m, int64_t size_n,
                              int64_t size_k);

bool cutlass_scaled_mm_supports_fp8(int64_t cuda_device_capability);

void cutlass_scaled_mm(torch::Tensor& out, torch::Tensor const& a,
                       torch::Tensor const& b, torch::Tensor const& a_scales,
                       torch::Tensor const& b_scales,
                       c10::optional<torch::Tensor> const& bias);

void cutlass_scaled_mm_azp(torch::Tensor& out, torch::Tensor const& a,
                           torch::Tensor const& b,
                           torch::Tensor const& a_scales,
                           torch::Tensor const& b_scales,
                           torch::Tensor const& azp_adj,
                           c10::optional<torch::Tensor> const& azp,
                           c10::optional<torch::Tensor> const& bias);

torch::Tensor marlin_qqq_gemm(torch::Tensor const& a,
                              torch::Tensor const& b_q_weight,
                              torch::Tensor const& s_tok,
                              torch::Tensor const& s_ch,
                              torch::Tensor const& s_group,
                              torch::Tensor& workspace, int64_t size_m,
                              int64_t size_n, int64_t size_k);
#endif

void static_scaled_int8_quant(torch::Tensor& out, torch::Tensor const& input,
                              torch::Tensor const& scale,
                              c10::optional<torch::Tensor> const& azp);

void dynamic_scaled_int8_quant(torch::Tensor& out, torch::Tensor const& input,
                               torch::Tensor& scales,
                               c10::optional<torch::Tensor> const& azp);

// torch::Tensor gptq_gemm(torch::Tensor a, torch::Tensor b_q_weight,
//                         torch::Tensor b_gptq_qzeros,
//                         torch::Tensor b_gptq_scales, torch::Tensor b_g_idx,
//                         bool use_exllama, int64_t bit);

// void gptq_shuffle(torch::Tensor q_weight, torch::Tensor q_perm, int64_t bit);

// void static_scaled_fp8_quant(torch::Tensor& out, torch::Tensor const& input,
//                              torch::Tensor const& scale);

// void dynamic_scaled_fp8_quant(torch::Tensor& out, torch::Tensor const& input,
//                               torch::Tensor& scale);

// void dynamic_per_token_scaled_fp8_quant(
//     torch::Tensor& out, torch::Tensor const& input, torch::Tensor& scale,
//     c10::optional<torch::Tensor> const& scale_ub);

void moe_align_block_size(torch::Tensor topk_ids, int64_t num_experts,
                          int64_t block_size, torch::Tensor sorted_token_ids,
                          torch::Tensor experts_ids,
                          torch::Tensor num_tokens_post_pad);

std::vector<torch::Tensor> selective_scan_fwd(
    const torch::Tensor& u, const torch::Tensor& delta, const torch::Tensor& A,
    const torch::Tensor& B, const torch::Tensor& C,
    const c10::optional<torch::Tensor>& D_,
    const c10::optional<torch::Tensor>& z_,
    const c10::optional<torch::Tensor>& delta_bias_, bool delta_softplus,
    const c10::optional<torch::Tensor>& index_,
    const c10::optional<torch::Tensor>& x);

at::Tensor causal_conv1d_update(
    const at::Tensor& x, const at::Tensor& conv_state, const at::Tensor& weight,
    const c10::optional<at::Tensor>& bias, bool silu_activation,
    const c10::optional<at::Tensor>& conv_state_indices);

at::Tensor causal_conv1d_fwd(const at::Tensor& x, const at::Tensor& weight,
                             const c10::optional<at::Tensor>& bias_,
                             const c10::optional<at::Tensor>& seq_idx_,
                             const c10::optional<at::Tensor>& initial_states_,
                             const c10::optional<at::Tensor>& final_states_out_,
                             bool silu_activation);

#ifndef USE_ROCM
using fptr_t = int64_t;
fptr_t init_custom_ar(torch::Tensor& meta, torch::Tensor& rank_data,
                      const std::vector<std::string>& handles,
                      const std::vector<int64_t>& offsets, int64_t rank,
                      bool full_nvlink);
void all_reduce_reg(fptr_t _fa, torch::Tensor& inp, torch::Tensor& out);
void all_reduce_unreg(fptr_t _fa, torch::Tensor& inp, torch::Tensor& reg_buffer,
                      torch::Tensor& out);
void dispose(fptr_t _fa);
int64_t meta_size();
void register_buffer(fptr_t _fa, torch::Tensor& t,
                     const std::vector<std::string>& handles,
                     const std::vector<int64_t>& offsets);
std::tuple<torch::Tensor, std::vector<int64_t>> get_graph_buffer_ipc_meta(
    fptr_t _fa);
void register_graph_buffers(fptr_t _fa, const std::vector<std::string>& handles,
                            const std::vector<std::vector<int64_t>>& offsets);
#endif