Want to use the powerpc e200z6 SPE APU

[Peter Dufault]

On Jan 27, 2010, at 1:40 , Mike Arthur wrote:

> Have you had any more success with hacking SPE support?


I'm building the mpc55xxevb with these flags:

CPU_CFLAGS = -mcpu=8540 -meabi -msdata -fno-common -mfloat-gprs=single -mspe   -D__ppc_generic -mstrict-align

This is working fine for single precision floating point.  The repetitive signal processing that I'm doing is taking 7% of the CPU and so I haven't felt the need to use the SPE instructions.

I have enabled the SPE and I'm saving and restoring the context during context switch but I don't know that that even matters if I only use single-precision FP.

I plan to clean up my diffs and send them before 4.10 is released.  If someone needs them now I can document them and send them out, but I haven't reapplied them in a month or so and can't promise they are in good shape.

The patches are to:

1. Save and restore the SPE;
2. Support the Phytec MPC5554 PhyCORE module, including the SMC91111 ethernet;
3. Compile-time conditionalize some of the chip selects, initial TLB setup, etc. when building MPC55xxevb to make it more generic.

That last part needs to be cleaned up which is what's been holding me up.  I've been pretty busy lately.

Peter
-----------------
Peter Dufault
HD Associates, Inc.      Software and System Engineering