You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
; Misaligned parts of the string are read from the nearest 16-bytes boundary
; and the irrelevant part masked out. It may read both before the begin of
; the string and after the end, but will never load any unnecessary cache
; line and never trigger a page fault for reading from non-existing memory
; pages because it never reads past the nearest following 16-bytes boundary.
; It may, though, trigger any debug watch within the same 16-bytes boundary.
So it should be possible to modify RapidJSON's SkipWhitespace_SIMD() to make it work safely. Currently it read 16 bytes unaligned from start of string, which will be possible to read bytes past the end of page boundary and cause segmentation fault.
The text was updated successfully, but these errors were encountered:
https://code.google.com/p/rapidjson/issues/detail?id=104
According to Agner's SSE
A_strlen()
implementation (http://www.agner.org/optimize/asmlib.zip),So it should be possible to modify RapidJSON's
SkipWhitespace_SIMD()
to make it work safely. Currently it read 16 bytes unaligned from start of string, which will be possible to read bytes past the end of page boundary and cause segmentation fault.The text was updated successfully, but these errors were encountered: