Fix LowerSDIV24

[Matt Arsenault]

On Jul 23, 2014, at 11:17 PM, Matt Arsenault <arsenm2 at gmail.com> wrote:

> 
> On Jul 10, 2014, at 6:28 AM, Jan Vesely <jan.vesely at rutgers.edu> wrote:
>>> 
>> 
>> I think we need to move i24 handling to SDIVREM since i32 and i64 algs
>> for udivrem produce rem values as a side product and don't do X - Y *
>> (X / Y).
>> 
>> All llvm tests and piglits 'scalar-arithmetic-(u)int' still pass for me
>> with the attached patches applied on top of yours.
> 
> 
> 
> These look fine to me. I’ll commit my part. Here are also some more tests for sdiv24 and srem24 someone can apply to piglit.
> 
> <add_piglit_sdiv24_tests.diff>

r213843-r213844