https://bugs.kde.org/show_bug.cgi?id=404843
--- Comment #3 from Andreas Arnez <ar...@linux.ibm.com> --- To summarize some of the discussion on IRC about this topic today (please correct if I misunderstood anything): * Valgrind's unwinder needs to be extra fast. For instance, Memcheck takes a stack trace for each block allocated, and then again when it's freed. * In order to be fast, the unwinder limits the number of registers to unwind. Currently no platform unwinds more than 5 registers. * Initial register values are copied into an UnwindStartRegs in VG_(get_UnwindStartRegs) from the registers of the specified simulated threads. * In order to use FPRs for unwinding, they have to be added to the UnwindStartRegs and D3UnwindRegs structures, and summary rule fields for the s390 variant of DiCfSI must be added as well. Preferably we only add a few FPRs, to limit the performance impact. I'll probably look into that soon. -- You are receiving this mail because: You are watching all bug changes.
