forked from OSchip/llvm-project
40 lines
1.7 KiB
Plaintext
40 lines
1.7 KiB
Plaintext
Date: Tue, 13 Feb 2001 13:29:52 -0600 (CST)
|
|
From: Chris Lattner <sabre@nondot.org>
|
|
To: Vikram S. Adve <vadve@cs.uiuc.edu>
|
|
Subject: LLVM Concerns...
|
|
|
|
|
|
I've updated the documentation to include load store and allocation
|
|
instructions (please take a look and let me know if I'm on the right
|
|
track):
|
|
|
|
file:/home/vadve/lattner/llvm/docs/LangRef.html#memoryops
|
|
|
|
I have a couple of concerns I would like to bring up:
|
|
|
|
1. Reference types
|
|
Right now, I've spec'd out the language to have a pointer type, which
|
|
works fine for lots of stuff... except that Java really has
|
|
references: constrained pointers that cannot be manipulated: added and
|
|
subtracted, moved, etc... Do we want to have a type like this? It
|
|
could be very nice for analysis (pointer always points to the start of
|
|
an object, etc...) and more closely matches Java semantics. The
|
|
pointer type would be kept for C++ like semantics. Through analysis,
|
|
C++ pointers could be promoted to references in the LLVM
|
|
representation.
|
|
|
|
2. Our "implicit" memory references in assembly language:
|
|
After thinking about it, this model has two problems:
|
|
A. If you do pointer analysis and realize that two stores are
|
|
independent and can share the same memory source object, there is
|
|
no way to represent this in either the bytecode or assembly.
|
|
B. When parsing assembly/bytecode, we effectively have to do a full
|
|
SSA generation/PHI node insertion pass to build the dependencies
|
|
when we don't want the "pinned" representation. This is not
|
|
cool.
|
|
I'm tempted to make memory references explicit in both the assembly and
|
|
bytecode to get around this... what do you think?
|
|
|
|
-Chris
|
|
|