mirror of
https://github.com/deepseek-ai/FlashMLA
synced 2025-06-26 18:15:54 +00:00
Performance Update (2025.04.22) (#71)
* Fix benchmark script * Performance optimization for compute-bound cases * Add new testcase (s_k = 16384) * Update README.md * Update comment * Update README.md * Add the deep-dive blog * Add background color for MLA Kernel Sched.drawio.svg * Use relative path for the schedule image * Move flash_mla.h to kernels/params.h
This commit is contained in:
13
csrc/kernels/config.h
Normal file
13
csrc/kernels/config.h
Normal file
@@ -0,0 +1,13 @@
|
||||
#pragma once
|
||||
|
||||
namespace Config {
|
||||
|
||||
static constexpr int BLOCK_SIZE_M = 64;
|
||||
static constexpr int PAGE_BLOCK_SIZE = 64;
|
||||
|
||||
static constexpr int HEAD_DIM_K = 576;
|
||||
static constexpr int HEAD_DIM_V = 512;
|
||||
|
||||
static constexpr int FIXED_OVERHEAD_NUM_BLOCKS = 5;
|
||||
|
||||
}
|
||||
Reference in New Issue
Block a user