Custom Query (1030 matches)

Attempting to use the ARM version of CCL 1.9 on a recent model Raspberry Pi (Linux 3.10.25+), armcl fails to start with the message: "CPU doesn't support required features"

Investigation shows that the check_arm_cpu function in pmcl_kernel.c is not correctly interpreting the output of /proc/cpuinfo. My Raspberry Pi shows the following:

$ cat /proc/cpuinfo
processor	: 0
model name	: ARMv6-compatible processor rev 7 (v6l)
BogoMIPS	: 2.00
Features	: swp half thumb fastmult vfp edsp java tls 
CPU implementer	: 0x41
CPU architecture: 7
CPU variant	: 0x0
CPU part	: 0xb76
CPU revision	: 7

Hardware	: BCM2708
Revision	: 000d
Serial		: ******************

check_arm_cpu looks for the string "Processor" (with an upper case P) at the start of each line, using strncmp. However, because the Raspberry Pi is returning "processor" (with a lower case p) the line is not recognised.

The fix is simply to change strncmp to strncasecmp.

There are other issues with the function as well, one being that calls are made to malloc() but the return value is not checked. These can be avoided by allocating procline and cpuline on the stack, with the same size as the input line buffer. This trades off wasting a maximum of 258 bytes of memory on the stack to avoid the overhead of calling malloc() and free() and having to check the return values from malloc().

A modified version of pmcl_kernel.c is included - the changes are to check_arm_cpu only. However, it has not been tested with anything except the Raspberry Pi in my possession.

Start a trunk ccl and do (require 'cocoa). Choose "Preferences..." from the application menu.

> Error: Invalid memory operation.
> While executing: GUI::|-[CCLFontToNameTransformer transformedValue:]|, in process Initial(0).

The function in question is in ccl:cocoa-ide;preferences.lisp.

It works on 1.9.

I suspect a compiler bug, but I have not looked at it very hard, so I could be wrong about that. The trunk and 1.9 versions of the function are identical.

Are specialized arrays for complex numbers supposed to work, yet?

I tried loading bordeaux-fft on an Odroid U3 (linuxarm) and got this:

Welcome to Clozure Common Lisp Version 1.10-dev-r16104-trunk (LinuxARM32)!

? (ql:quickload "bordeaux-fft") To load "bordeaux-fft":

Load 1 ASDF system:

bordeaux-fft

; Loading "bordeaux-fft" [package bordeaux-fft]

Error: bad regspec: :COMPLEX-DOUBLE-FLOAT-VECTOR While executing: CCL::%HARD-REGSPEC-CLASS, in process listener(1). Type :POP to abort, :R for a list of available restarts. Type :? for other options.

The backtrace is attached.