LLVM Bugzilla is read-only and represents the historical archive of all LLVM issues filled before November 26, 2021. Use github to submit LLVM bugs

Bug 1356 - [inlineasm] integer mem input uses int operand as address instead of value
Summary: [inlineasm] integer mem input uses int operand as address instead of value
Status: RESOLVED FIXED
Alias: None
Product: libraries
Classification: Unclassified
Component: Common Code Generator Code (show other bugs)
Version: 1.0
Hardware: Macintosh MacOS X
: P normal
Assignee: Chris Lattner
URL:
Keywords: miscompilation
Depends on:
Blocks:
 
Reported: 2007-04-25 01:45 PDT by Oliver Hunt
Modified: 2010-03-06 13:59 PST (History)
2 users (show)

See Also:
Fixed By Commit(s):

Attachments
Simple test case (491 bytes, text/plain)
2007-04-25 01:46 PDT, Oliver Hunt 		Details
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Oliver Hunt 2007-04-25 01:45:10 PDT 
llvm-gcc generates invalid code for a function consisting only of a single inline asm statement
Comment 1 Oliver Hunt 2007-04-25 01:46:25 PDT 
Created attachment 800 [details]
Simple test case

This simple test case compiles and runs fine in gcc at any optimisation level,
but crashes when compiled with llvm-gcc
Comment 2 Chris Lattner 2007-04-27 12:23:05 PDT 
Taking a look, we probably don't handle the 'o' constraint.
Comment 3 Chris Lattner 2007-04-27 13:19:33 PDT 
hrm, this is nasty.  The code generator is seeing o as a memory constraint (correct) then seeing the integer 
constant as a valid offsetable address (wrong).  Tracking.
Comment 4 Chris Lattner 2007-04-27 16:45:56 PDT 
I am working on fixing this, but it may take several days.  The workaround is to remove 'o' from the inline 
asm.
Comment 5 Chris Lattner 2007-04-28 01:44:21 PDT 
Fixed.  Many patches made this possible, but the final patch is here: 
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20070423/048601.html

Testcase here: CodeGen/X86/2007-04-27-InlineAsm-IntMemInput.ll

Many thanks for the reduced testcase!

-Chris
Comment 6 Oliver Hunt 2007-04-28 01:47:12 PDT 
Awesome, it will probably be a wee while before i can verify alas :(