On Fri, Dec 02, 2016 at 03:59:26PM +0000, Alex Bennée wrote: > Hi Peter, > > I've been cleaning things up so I thought I should re-post my current > state. These all apply to the current master. > > I had to regenerate all the risu binaries as I'd used --no-fp for a > bunch of them originally which was causing failures. I'm not sure if > this is due to the FP registers not being cleared by the kernel if FP > isn't used - but we certainly don't do anything to them except when > set by the memory/context blocks (without --no-fp). This led me to > write the 3 noddy scripts included here. > > The record/playback is still aarch64 only. I'm open to ideas to do > this more cleanly otherwise if anyone has access to some PPC hardware > I can port the record/playback code to the other architectures.
Excellent job, Alex I can make the PPC port here. I still have work to do in Risu for PPC so I can include this in my task as well. I'm planning to start working on it this week. Thank you > > Feel free to cherry-pick any of the minor patches as you wish ;-) > > Alex Bennée (9): > risu: a bit more verbosity when running > aarch64: add hand-coded risu skeleton for directed testing > risu: add simple trace and replay support > risu: add support compressed tracefiles > risu_aarch64: it's -> it is > risugen: remove grocer's apostrophe from REs > new: generate_all.sh script > new: record_traces.sh helper script > new: run_risu.sh script > > Makefile | 10 +- > aarch64_simd_handcoded.risu.S | 208 > ++++++++++++++++++++++++++++++++++++++++++ > configure | 55 +++++++++++ > generate_all.sh | 55 +++++++++++ > record_traces.sh | 16 ++++ > risu.c | 116 ++++++++++++++++++----- > risu.h | 15 +++ > risu_aarch64.c | 89 +++++++++++++++++- > risu_reginfo_aarch64.h | 7 ++ > risugen | 2 +- > run_risu.sh | 51 +++++++++++ > 11 files changed, 598 insertions(+), 26 deletions(-) > create mode 100644 aarch64_simd_handcoded.risu.S > create mode 100755 generate_all.sh > create mode 100755 record_traces.sh > create mode 100755 run_risu.sh > > -- > 2.10.2 >
