You would need to:
1) know what type of file-system your files reside on.
2) Open the device (note: not opening the file!)
3) interpret the file system structure to find the blocks corresponding to your file on disk
4) retrieve the file blocks corresponding to the file and perform your search on the data in memory.
---
Of course you can search multiple files on the same disk, using the same device handle, so of course you save yourself the overhead of multiple "open" calls.

Note: this likely would not work if the file was a remote (networked) file.
Note 2: This would likely be quite painful, but if you're into that sort of thing...:-)

As others have said, within the definition of your problem is an "inherent" open to be able to look into the contents of the file. What are you trying to do that you want to avoid opening the file in perl?

FWIW, I'd change Ted's answer, above, to use "pcregrep" instead of "grep" to maintain use of perl-compatible regular expressions :-).

p.s. -- sometimes you just have to put on your hat of "literalness" :-)