[llvm-commits] [PATCH 08/11] (patch) Mips specific inline asm: constraint 'O':

[Eric Christopher]

On Apr 26, 2012, at 2:49 PM, Jim Grosbach <grosbach at apple.com> wrote:

> 
> On Apr 26, 2012, at 2:43 PM, Eric Christopher <echristo at apple.com> wrote:
> 
>> 
>> On Apr 12, 2012, at 1:51 PM, Jack Carter <jcarter at mips.com> wrote:
>> 
>>> A signed 15 bit constant.
>>> 
>>> Example:
>>> 
>>> short s_input = 7;short s_result = 0;
>>> 
>>> // Good: value is within range.
>>> short s_val = 16383;
>>> __asm__ __volatile__(
>>>  "addi %0,%1,%2" : "=r" (s_result)  : "r" (s_input), "O" (s_val));
>>> printf ("mips_addi(%d,%d) = %d\n", s_input, s_val, s_result);
>>> 
>>> // Good: value is still within range.
>>> short s_val = -16383;
>>> __asm__ __volatile__(
>>>  "addi %0,%1,%2" : "=r" (s_result)  : "r" (s_input), "O" (s_val));
>>> printf ("mips_addi(%d,%d) = %d\n", s_input, s_val, s_result);
>>> 
>>> // Bad: value is out of range. llc should produce an error
>>> short s_val = 16384;
>>> __asm__ __volatile__(
>>>  "addi %0,%1,%2" : "=r" (s_result)  : "r" (s_input), "O" (s_val));
>>> printf ("mips_addi(%d,%d) = %d\n", s_input, s_val, s_result);
>>> ---
>>> lib/Target/Mips/MipsISelLowering.cpp         |   11 +++++++++++
>>> test/CodeGen/Mips/inlineasm-cnstrnt-bad-O.ll |   13 +++++++++++++
>>> test/CodeGen/Mips/inlineasm_constraint.ll    |    6 ++++++
>>> 3 files changed, 30 insertions(+), 0 deletions(-)
>> 
>> Slight problem here. I don't like using XFAIL for "this should fail" type of tests. You'll need to come up with another way to test this one.
> 
> Expected failures where "correct" is to get a diagnostic?
> 
> Check test/MC/ARM/thumb-diagnostics.s for examples of that.

Yep. :) Sorry for not putting the link in. Thanks Jim :)

-eric