公开(公告)号：US20240289527A1

公开(公告)日：2024-08-29

申请号：US18042431

申请日：2022-10-12

Applicant: MediaTek Inc.

Inventor： Da-Shan SHIU ,  Alexandru CIOBA ,  Fu-Chieh CHANG

IPC: G06F30/392 ,  G06F30/398

CPC classification number: G06F30/392 ,  G06F30/398

Abstract: A neural network (NN) performs macro placement on a chip. A mask is updated to mark invalid regions occupied by already-placed macros on a chip canvas. A policy network of the NN generates summary statistics of a two-dimensional (2D) continuous probability distribution over a continuous action space for a given state of the chip canvas. The NN selects an action based on the continuous probability distribution. The selected action corresponds to a coordinate in an unmasked region. The NN generates a trajectory including (state, action) pairs. The final state in the trajectory corresponds to a completed placement of macros.

公开(公告)号：US20250013813A1

公开(公告)日：2025-01-09

申请号：US18264303

申请日：2023-03-17

Applicant: MediaTek Inc.

Inventor： Da-Shan SHIU ,  Alexandru CIOBA ,  Fu-Chieh CHANG

IPC: G06F30/392 ,  G06F30/398

Abstract: Macros are placed on a canvas based on density map calculations. First, a grid representing the canvas is initialized. The grid is formed by grid cells. To place a macro on the grid, a coordinate on the grid is chosen, and multiple density maps are calculated using average-pooling filters of multiple resolutions or orientations. The lowest-level density map is described by the grid with each grid cell having a corresponding density value. The density value in a higher-level density map is calculated by performing an average-pooling operation on density values in a lower- level density map. The placement of the macro at the coordinate is validated when no density value in the density maps exceeds a corresponding density threshold.

公开(公告)号：US20240289603A1

公开(公告)日：2024-08-29

申请号：US18042439

申请日：2022-10-12

Applicant: MediaTek Inc.

Inventor： Da-Shan SHIU ,  Alexandru CIOBA ,  Fu-Chieh CHANG

IPC: G06N3/08

CPC classification number: G06N3/08

Abstract: A system trains a neural network (NN) for macro placement. The system constructs a set of positive samples of trajectories by sequentially removing the same set of macros in different orders from an at least partially-placed canvas of a chip. The system also constructs a set of negative samples of trajectories by placing not-yet-placed macros at random positions on an at least partially-empty canvas of the chip. The system then trains the NN and a graph NN (GNN) in the NN using the positive samples and the negative samples.

公开(公告)号：US20240289602A1

公开(公告)日：2024-08-29

申请号：US18042423

申请日：2022-10-12

Applicant: MediaTek Inc.

Inventor： Da-Shan SHIU ,  Alexandru CIOBA ,  Fu-Chieh CHANG

IPC: G06N3/08

CPC classification number: G06N3/08

Abstract: A system uses a neural network (NN) for macro placement. The system receives an input including objectives and a subspace of preferences. Each preference is a vector of weights assigned to corresponding objectives, and each objective is a measurement of a placement characteristic. The system trains the NN to place macros on a training set of chips to optimize a reward, where the reward is calculated from the objectives and the preferences. The NN generates a probability distribution of an action under a current state of a chip, where the action indicates a coordinate on the chip to place a macro. The NN further generates a sequence of (state, action) pairs to form a trajectory. The final state in the trajectory corresponds to a completed macro placement.