mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-20 06:49:15 +00:00
Ck tile/moe sorting (#1624)
* add moe_sorting & check ok * fix comments & typo * Run remod.py under include/ck_tile & example/ck_tile directories * format codes * fix output ci check bug * fix moe sorting readme and error commit file * use magiv div to accelerate compute * add an loop unroll for moe lds ops * add extblocksnel to set zeros for moebufs * [Ck_tile] moe set zero run ok, add size check and fix ref check * [Ck_tile]fix moe_sorting fuse set_zero remod * [Ck_tile] change name style, fix zero buffer size err, change folder * [Ck_tile] moe_sorting: fix name style * [Ck_tile] moe_sorting, remove useless params in traits * [Ck_tile] change outputtile cnt * unit_size; change output buf alloc --------- Co-authored-by: dummycoderfe <noplydummmycoder@163.com> Co-authored-by: Po Yen, Chen <PoYen.Chen@amd.com> Co-authored-by: carlushuang <carlus.huang@amd.com>
This commit is contained in:
dummycoderfe
232
include/ck_tile/ops/fused_moe/kernel/moe_sorting_kernel.hpp
Normal file
232
include/ck_tile/ops/fused_moe/kernel/moe_sorting_kernel.hpp
Normal file
@@ -0,0 +1,232 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#pragma once
|
||||
|
||||
#include "ck_tile/core.hpp"
|
||||
#include "ck_tile/ops/common.hpp"
|
||||
#include "ck_tile/ops/elementwise.hpp"
|
||||
#include "ck_tile/host/hip_check_error.hpp"
|
||||
#include <string>
|
||||
#include <type_traits>
|
||||
|
||||
namespace ck_tile {
|
||||
|
||||
struct MoeSortingHostArgs
|
||||
{
|
||||
const void* p_topk_ids;
|
||||
const void* p_weights;
|
||||
void* p_sorted_token_ids;
|
||||
void* p_sorted_weights;
|
||||
void* p_sorted_expert_ids;
|
||||
void* p_total_tokens_post_pad;
|
||||
void* p_moe_buf;
|
||||
index_t tokens;
|
||||
index_t unit_size;
|
||||
index_t num_experts;
|
||||
index_t topk;
|
||||
index_t moe_buf_bytes;
|
||||
};
|
||||
|
||||
template <typename Problem_>
|
||||
struct MoeSortingKernel
|
||||
{
|
||||
using Problem = remove_cvref_t<Problem_>;
|
||||
|
||||
using IndexType = typename Problem::IndexType;
|
||||
using WeightType = typename Problem::WeightType;
|
||||
|
||||
typedef MoeSortingHostArgs MoeSortingKargs;
|
||||
|
||||
using Hargs = MoeSortingHostArgs;
|
||||
|
||||
struct Kargs
|
||||
{
|
||||
const void* p_topk_ids;
|
||||
const void* p_weights;
|
||||
void* p_sorted_token_ids;
|
||||
void* p_sorted_weights;
|
||||
void* p_sorted_expert_ids;
|
||||
void* p_total_tokens_post_pad;
|
||||
void* p_moe_buf;
|
||||
index_t tokens;
|
||||
index_t num_experts;
|
||||
index_t moe_buf_bytes;
|
||||
|
||||
index_t tokens_per_thread;
|
||||
mdiv unit_size_mdiv;
|
||||
mdiv topk_mdiv;
|
||||
};
|
||||
|
||||
CK_TILE_HOST static constexpr auto GridSize(const Hargs& h)
|
||||
{
|
||||
// TODO: assume num-experts not too much
|
||||
return dim3(1 + ck_tile::integer_divide_ceil(h.moe_buf_bytes, BlockSize(h).x * 16));
|
||||
}
|
||||
|
||||
CK_TILE_HOST static constexpr auto BlockSize(const Hargs& h)
|
||||
{
|
||||
return dim3(ck_tile::integer_least_multiple(h.num_experts, ck_tile::get_warp_size()));
|
||||
}
|
||||
|
||||
// in byte
|
||||
CK_TILE_HOST static constexpr auto GetSmemSize(const Hargs& h)
|
||||
{
|
||||
const auto blocks = BlockSize(h);
|
||||
return ((blocks.x + 1) * h.num_experts + (h.num_experts + 1)) * sizeof(index_t);
|
||||
}
|
||||
|
||||
CK_TILE_HOST static constexpr auto MakeKargs(const Hargs& h)
|
||||
{
|
||||
Kargs k;
|
||||
k.p_topk_ids = h.p_topk_ids;
|
||||
k.p_weights = h.p_weights;
|
||||
k.p_sorted_token_ids = h.p_sorted_token_ids;
|
||||
k.p_sorted_weights = h.p_sorted_weights;
|
||||
k.p_sorted_expert_ids = h.p_sorted_expert_ids;
|
||||
k.p_moe_buf = h.p_moe_buf;
|
||||
k.p_total_tokens_post_pad = h.p_total_tokens_post_pad;
|
||||
k.tokens = h.tokens;
|
||||
k.num_experts = h.num_experts;
|
||||
k.moe_buf_bytes = h.moe_buf_bytes;
|
||||
|
||||
const auto blocks = BlockSize(h);
|
||||
k.tokens_per_thread = integer_divide_ceil(h.tokens * h.topk, blocks.x);
|
||||
k.unit_size_mdiv = mdiv{static_cast<uint32_t>(h.unit_size)};
|
||||
k.topk_mdiv = mdiv{static_cast<uint32_t>(h.topk)};
|
||||
return k;
|
||||
}
|
||||
|
||||
CK_TILE_DEVICE index_t calc_index(index_t total_col, index_t row, index_t col) const
|
||||
{
|
||||
return row * total_col + col;
|
||||
}
|
||||
|
||||
CK_TILE_DEVICE void moe_buf_set_zero_kernel(uint8x16_t* buf, index_t buf_bytes) const
|
||||
{
|
||||
const index_t offset = (blockIdx.x - 1) * blockDim.x + threadIdx.x;
|
||||
if(offset < buf_bytes / 16)
|
||||
{
|
||||
buf[offset] = uint8x16_t{0};
|
||||
}
|
||||
}
|
||||
|
||||
CK_TILE_DEVICE void moe_align_block_size_kernel(const IndexType* __restrict__ topk_id,
|
||||
const WeightType* __restrict__ weights,
|
||||
index_t* p_sorted_token_ids,
|
||||
WeightType* p_sorted_weights,
|
||||
index_t* p_sorted_expert_ids,
|
||||
index_t* p_total_tokens_post_pad,
|
||||
const index_t num_experts,
|
||||
const index_t tokens_per_thread,
|
||||
const index_t numel,
|
||||
const mdiv unit_size_mdiv,
|
||||
const mdiv topk_mdiv,
|
||||
void* smem) const
|
||||
{
|
||||
const index_t tid = static_cast<index_t>(threadIdx.x);
|
||||
const index_t start_idx = tid * tokens_per_thread;
|
||||
|
||||
index_t* shared_mem = reinterpret_cast<index_t*>(smem);
|
||||
|
||||
index_t* tokens_cnts = shared_mem; // 2d: (blockDim.x + 1, num_experts)
|
||||
index_t* cumsum = shared_mem + (blockDim.x + 1) * num_experts; // 1: (num_experts + 1)
|
||||
for(int i = 0; i < num_experts; ++i)
|
||||
{
|
||||
tokens_cnts[calc_index(num_experts, tid + 1, i)] = 0;
|
||||
}
|
||||
#pragma unroll Problem_::InternalLoadUnroll
|
||||
for(int i = start_idx; i < numel && i < start_idx + tokens_per_thread; ++i)
|
||||
{
|
||||
++tokens_cnts[calc_index(num_experts, tid + 1, topk_id[i])];
|
||||
}
|
||||
__syncthreads();
|
||||
|
||||
if(tid < num_experts)
|
||||
{
|
||||
tokens_cnts[calc_index(num_experts, 0, tid)] = 0;
|
||||
for(int i = 1; i <= static_cast<index_t>(blockDim.x); ++i)
|
||||
{
|
||||
tokens_cnts[calc_index(num_experts, i, tid)] +=
|
||||
tokens_cnts[calc_index(num_experts, i - 1, tid)];
|
||||
}
|
||||
}
|
||||
|
||||
// __syncthreads();
|
||||
if(tid == 0)
|
||||
{
|
||||
cumsum[0] = 0;
|
||||
for(int i = 1; i <= num_experts; ++i)
|
||||
{
|
||||
auto current_units = [&]() {
|
||||
index_t x_ = tokens_cnts[calc_index(num_experts, blockDim.x, i - 1)] +
|
||||
unit_size_mdiv.divisor - 1;
|
||||
index_t y_ = unit_size_mdiv.div(x_);
|
||||
return max(y_, 1) * unit_size_mdiv.divisor;
|
||||
}();
|
||||
cumsum[i] = cumsum[i - 1] + current_units;
|
||||
}
|
||||
*p_total_tokens_post_pad = cumsum[num_experts];
|
||||
}
|
||||
__syncthreads();
|
||||
if(tid < num_experts)
|
||||
{
|
||||
for(int i = cumsum[tid]; i < cumsum[tid + 1]; i += unit_size_mdiv.divisor)
|
||||
{
|
||||
p_sorted_expert_ids[unit_size_mdiv.div(i)] = tid;
|
||||
}
|
||||
}
|
||||
|
||||
#pragma unroll Problem_::InternalLoadUnroll
|
||||
for(int i = start_idx; i < numel && i < start_idx + tokens_per_thread; ++i)
|
||||
{
|
||||
index_t expert_id = topk_id[i];
|
||||
index_t rank_post_pad =
|
||||
tokens_cnts[calc_index(num_experts, tid, expert_id)] + cumsum[expert_id];
|
||||
p_sorted_token_ids[rank_post_pad] = topk_mdiv.div(i);
|
||||
p_sorted_weights[rank_post_pad] = weights[i];
|
||||
++tokens_cnts[calc_index(num_experts, tid, expert_id)];
|
||||
}
|
||||
|
||||
const index_t prefill_token = topk_mdiv.div(numel);
|
||||
if(tid < num_experts)
|
||||
{
|
||||
index_t expert_offset =
|
||||
cumsum[tid] + tokens_cnts[calc_index(num_experts, blockDim.x, tid)];
|
||||
while(expert_offset < cumsum[tid + 1])
|
||||
{
|
||||
p_sorted_token_ids[expert_offset] = prefill_token;
|
||||
p_sorted_weights[expert_offset] = static_cast<WeightType>(0.0);
|
||||
expert_offset++;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
CK_TILE_DEVICE void operator()(Kargs kargs) const
|
||||
{
|
||||
if(blockIdx.x > 0)
|
||||
{
|
||||
if(kargs.p_moe_buf)
|
||||
{
|
||||
moe_buf_set_zero_kernel(reinterpret_cast<uint8x16_t*>(kargs.p_moe_buf),
|
||||
kargs.moe_buf_bytes);
|
||||
}
|
||||
return;
|
||||
}
|
||||
const size_t numel = kargs.tokens * kargs.topk_mdiv.divisor;
|
||||
extern __shared__ char smem[];
|
||||
return moe_align_block_size_kernel(static_cast<const IndexType*>(kargs.p_topk_ids),
|
||||
static_cast<const WeightType*>(kargs.p_weights),
|
||||
static_cast<IndexType*>(kargs.p_sorted_token_ids),
|
||||
static_cast<WeightType*>(kargs.p_sorted_weights),
|
||||
static_cast<IndexType*>(kargs.p_sorted_expert_ids),
|
||||
static_cast<IndexType*>(kargs.p_total_tokens_post_pad),
|
||||
kargs.num_experts,
|
||||
kargs.tokens_per_thread,
|
||||
numel,
|
||||
kargs.unit_size_mdiv,
|
||||
kargs.topk_mdiv,
|
||||
smem);
|
||||
}
|
||||
};
|
||||
} // namespace ck_tile
|
||||
Reference in New Issue
Block a user