| 1 |
Date: Tue, 13 Feb 2001 13:29:52 -0600 (CST) |
| 2 |
From: Chris Lattner <[email protected]> |
| 3 |
To: Vikram S. Adve <[email protected]> |
| 4 |
Subject: LLVM Concerns... |
| 5 |
|
| 6 |
|
| 7 |
I've updated the documentation to include load store and allocation |
| 8 |
instructions (please take a look and let me know if I'm on the right |
| 9 |
track): |
| 10 |
|
| 11 |
file:/home/vadve/lattner/llvm/docs/LangRef.html#memoryops |
| 12 |
|
| 13 |
I have a couple of concerns I would like to bring up: |
| 14 |
|
| 15 |
1. Reference types |
| 16 |
Right now, I've spec'd out the language to have a pointer type, which |
| 17 |
works fine for lots of stuff... except that Java really has |
| 18 |
references: constrained pointers that cannot be manipulated: added and |
| 19 |
subtracted, moved, etc... Do we want to have a type like this? It |
| 20 |
could be very nice for analysis (pointer always points to the start of |
| 21 |
an object, etc...) and more closely matches Java semantics. The |
| 22 |
pointer type would be kept for C++ like semantics. Through analysis, |
| 23 |
C++ pointers could be promoted to references in the LLVM |
| 24 |
representation. |
| 25 |
|
| 26 |
2. Our "implicit" memory references in assembly language: |
| 27 |
After thinking about it, this model has two problems: |
| 28 |
A. If you do pointer analysis and realize that two stores are |
| 29 |
independent and can share the same memory source object, there is |
| 30 |
no way to represent this in either the bytecode or assembly. |
| 31 |
B. When parsing assembly/bytecode, we effectively have to do a full |
| 32 |
SSA generation/PHI node insertion pass to build the dependencies |
| 33 |
when we don't want the "pinned" representation. This is not |
| 34 |
cool. |
| 35 |
I'm tempted to make memory references explicit in both the assembly and |
| 36 |
bytecode to get around this... what do you think? |
| 37 |
|
| 38 |
-Chris |
| 39 |
|