00:01 <azonenberg> kristianpaul: well, the beginnings of it

00:02 <azonenberg> Tracking down bugs in the CPU one at a time

00:02 <azonenberg> i got it to run a matrix multiply compiled with gcc a few weeks ago

00:02 <azonenberg> earlier today i got it to work with caching enabled (sometimes a cache miss would result in an instruction getting skipped in a branch delay slot)

00:03 <azonenberg> but now it fails when i compile with -O

00:13 <kristianpaul> i guess eventually you're going release this work, or is part of university research for now and your paid work?

00:20 <azonenberg> The core CPU is already available on google code at http://code.google.com/p/utica-softcore/source/list

00:20 <azonenberg> i will be making a private fork for my thesis work once i have the baseline CPU finished

00:20 <azonenberg> and then most likely releasing the fork separately once i publish my thesis

00:21 <azonenberg> all commits up to r181 were imported from my internal repository, then i synced it to google code and started using that as the primary

00:21 <kristianpaul> ah, i missed that repo (utica)

00:22 <azonenberg> Btw i'm having some trouble in my elf-to-binary conversion

00:22 <azonenberg> when i turn on -O strings are shwoing up in wierd sections like .rodata.str1.4 instead of .rodata

00:23 <azonenberg> so my linker script and objdump that generates a flat binary will fail to find it

00:23 <azonenberg> and the executable gets truncated

00:23 <azonenberg> how can i force it to put them in .rodata?

14:32 <berndj> azonenberg, can't you tell the linker to lump it all together? something like .rodata (.rodata.*)?

14:32 <berndj> or add -fno-function-sections or whatever the equivalent is for data

14:32 <berndj> or is this not the GNU tools?

15:35 <azonenberg> berndj: i did add -fno-data-sections but that didnt work

15:35 <azonenberg> i found out how to tweak the linker script so it worked, though