https://bugs.kde.org/show_bug.cgi?id=485148
--- Comment #6 from Petr <[email protected]> --- I'm not sure I follow your last comment. The best would be to just compile the code locally and to run it with and without valgrind - the results will be different, and this is what I'm trying to point out. The instrumentation is incorrect in this case and it happens only with FMA - for example vpaddss is instrumented correctly (the remaining 3 elements are not zeroed). I'm not sure following vgdb makes sense in this case if you cannot step over a single instruction in disassembly (if it's instrumented wrong you would have wrong results). You can step over in gdb without valgrind to have a comparison. -- You are receiving this mail because: You are watching all bug changes.
