Daniel Wagner
Daniel Wagner
Currently, our precise memory model is quite slow. It's possible that one reason for this is that it needs to check quite a lot of sanity conditions on every read...
The new memory model is slow slow slow. We think one possible reason is its use of symbolic structs -- we always know concretely what field we want, so symbolic...
When reading from memory, the new memory model currently reassembles bytes by shifting and masking. If we use bitvector concatenation instead, that will more clearly indicate our intent -- and...
Our precise memory model makes up 8-ish uninterpreted functions each time it reads a pointer from memory, one for each byte read. We should instead have just one uninterpreted function,...