Skip to content

Update DeepFlame2.0_28_1_2026.md #342

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
KashaDong merged 1 commit into from
Feb 4, 2026
Merged

Update DeepFlame2.0_28_1_2026.md #342

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 0 additions & 8 deletions source/_posts/DeepFlame2.0_28_1_2026.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,8 +29,6 @@ To address this challenge, we collaborated with Shanghai Shuqian Technology to d

The CoCo agent significantly lowers the barrier to high-performance computing development, allowing researchers to focus on high-level physical model design while delegating tedious code migration tasks to AI.

Related video:
<https://www.bilibili.com/video/BV1bfiCBbECK>

#### 2. FlamePilot: DeepFlame’s CFD Simulation Agent

Expand Down Expand Up @@ -77,12 +75,6 @@ Based on the Kunpeng ecosystem, DeepFlame 2.0 has carried out multiple adaptatio

In terms of usability, the DeepFlame 2.0 software stack can be natively compiled for Kunpeng platforms and supports one-click deployment and execution via the Jarvis tool. In terms of performance, DeepFlame 2.0 achieves deep optimization for ARM architectures and full-stack performance breakthroughs: at the hardware level, fine-grained core binding and memory allocation strategies are introduced for the Kunpeng 920 Professional Edition’s many-core, multi-NUMA, on-chip memory architecture; at the software level, the codebase is refactored based on the BiSheng compiler, integrating the Kunpeng Math Library (KML) to accelerate GEMM computations while ensuring accuracy and robustness; at the algorithmic level, ARM-native mixed-precision solvers (FP64 sparse solving + FP16 inference) are designed to balance accuracy and speed. For AI–CFD integrated inference acceleration, lightweight neural network models are developed at the model layer to achieve high-accuracy inference and are adapted to the Kunpeng SME instruction set.

Related report:
https://mp.weixin.qq.com/s/FHRABrwF0GjSMo1vLUlCtA

Related video:
https://www.bilibili.com/video/BV1VfKAzwEQU

## 04 Quick Access to DeepFlame

The DeepFlame repository in the DeepModeling community is available at:
Expand Down