TD-W8970 V1 has 34Kc core (with HW FPU) but need SOFT_FLOAT enabled?

For Developers
1

The TP-LINK TD-W8970 V1 has a 34Kc MIPS CPU, the wiki said 24Kc but this is actually wrong (read here).

EDIT
I was wrong the 34kc does not have an HW FPU.

2

To my knowledge, the 34Kf has floating point and the 34Kc does not. Could not find anything mentioning a floating point unit in the 34Kc datasheet: https://s3-eu-west-1.amazonaws.com/downloads-mips/documents/MD00418-2B-34Kc-DTS-01.21.pdf

1 Like
3

Ok sorry my error. I read the mips wiki and I think they forgot to add "f" when talking about the FPU.

4

gcc generates identical code for mips24kc and mips34kc, therefore it is not needed to duplicate packages (both in terms of buildbot usage and server storage) for both.

5

So why I get way less memory usage when I compile with -march=34kc (instead of the default -mips32f2 -mtune=24kc)? About 10% (64 MiB ram)

6


The 34Kc CPU implements the MIPS32 Release 2 Architecture. In addition to the base architecture, it features the following
application specific extensions (ASE):
• The MIPS MT ASE which defines multi-threaded operation.
• The MIPS DSP ASE which provides support for signal processing instructions.
• The MIPS16e™ ASE which reduces code size
vs

The 24Kc core implements the MIPS32 Release 2 Architecture in an 8-stage pipeline. It includes support for the MIPS16e™

application specific extension and the 32-bit privileged resource architecture. This standard architecture allows support by a

wide range of industry-standard tools and development systems

1 Like
7

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.