Term
What execution units are associated with Execution Port 0 |
|
Definition
ALU 0 (ADD, LOGIC, BRANCH)
FP Mov (FP MOVES, FXCH) |
|
|
Term
What execution units are associated with Execution Port 1 |
|
Definition
ALU 1(ADD)
INT(SHIFT, ROTATE)
FP EXE |
|
|
Term
What execution units are associated with Execution Port 2 |
|
Definition
|
|
Term
What execution units are associated with Execution Port 4 |
|
Definition
|
|
Term
What are some micro optimizations for the execution stage? |
|
Definition
Well-Blended Execution Types
Low Cache Misses
Hyper-Threading
Low Data Dependencies |
|
|
Term
What is the different between parallel and serial execution? |
|
Definition
Serial execution uses a single variable, while parallel execution uses multiple variables (Loop Unrolling) |
|
|
Term
|
Definition
Updates Machine State
Notifies Missed Branch
Stores Memory
Issues Exceptions
Removes Micro-Ops |
|
|
Term
What does the Application Stage do? |
|
Definition
Creates World/Object Space Triangles
Stores Tris/Indicies in non-local video memory
Drivers
Sends commands to GPU
|
|
|
Term
What does the Vertex Processing Stage do? |
|
Definition
Fetches verts from Video Memory
Runs vertex Shader
Vertex Shaders support instruction level parallelism |
|
|
Term
What does the Rasterization Stage do? |
|
Definition
Interpolates Vertex Attributes
Shades Pixels
Resolves Pixels |
|
|
Term
What is the optimization process? |
|
Definition
1) Benchmark
2) Detection(measure/analyze)
3) Fix Problem
4) Repeat |
|
|