ARM NEON Bench Marking OpenMAx Numbers

Question

ARM NEON Bench Marking OpenMAx Numbers

I am currently working on a scanner that denotes the performance of the Cortex-A8 processor using various compilers (tested by DS-5 and IAR) and various libraries (Openmax, NE10). I am specifically working to see what the NEON block can do when comparing 1. ARM only 2.ARM + VFPv3

Openmax is available in two versions 1. Custumized for NEON 2. Sample Code for Arm Processors I used the following table using OpenMax. As expected, the performance of the NEON code is unmatched.

https://docs.google.com/file/d/0B318pI7x2QVceWRDTE5FQk8zcVk/edit

NE10 is the only other signal processing library currently available for Cortex-A. I am still working on the NE10 library.

Can these numbers be compared with other processors? or are there any software benchmarks for comparing processors?

Below is the link to the Openmax api http://www.khronos.org/files/openmax_dl_spec_1_0_2.pdf

+3

benchmarking arm neon fpu openmax

Aurum 06 Feb At 5:57 am

source to share

No one has answered this question yet

Check out similar questions:

sixteen

Cortex A9 NEON vs VFP error

eleven

How to check if NEON exists on a hand?

8

No return line from SIGABRT signal on ARM platform?

4

NEON instruction throws exception in Windows CE 7

3

Will my code improve the use of the restrict keyword in ARM?

2

How to Build and Decode Using ARM's OpenMAX Development Layer (DL)

1

Compile an ARM program in blocks of code on GNU / Linux

1

ARM CM3 programming in ROM and RAM

1

NE10 NEON library not working

0

Why is GCC generating Fxxx instead of Vxxx floating point assembly instructions for Cortex-A9?

ARM NEON Bench Marking OpenMAx Numbers

More articles: