[LLVMbugs] [Bug 6209] New: Keep Results Into SIMD Registers

[bugzilla-daemon at cs.uiuc.edu]

http://llvm.org/bugs/show_bug.cgi?id=6209

           Summary: Keep Results Into SIMD Registers
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: All
            Status: NEW
          Keywords: code-quality
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: bearophile at mailas.com
                CC: llvmbugs at cs.uiuc.edu


An article by Gustavo Oliveira:
http://www.gamasutra.com/view/feature/4248/designing_fast_crossplatform_simd_.php?page=3

See the part: 2. Keep Results Into SIMD Registers

It suggests to don't write code like this (note the deltalength, that's a
float):

Vec4& x2 = m_x[i2];
Vec4 delta = x2-x1;
float deltalength = Sqrt(Dot(delta,delta));
float diff = (deltalength-restlength)/deltalength;
x1 += delta*half*diff;
x2 -= delta*half*diff;

He says that's expensive since the compiler needs to generate code that will
move data from and to the SIMD and FPU registers. He suggests to write it like
this, assuming the "Dot" function above replicates the result into the SIMD
4-quad words and the "w" component zeroed-out. Now the expensive casting
operations are no longer necessary:

Vec4& x2 = m_x[i2];
Vec4 delta = x2-x1;
Vec4 deltalength = Sqrt(Dot(delta,delta));
Vec4 diff = (deltalength-restlength)/deltalength;
x1 += delta*half*diff;
x2 -= delta*half*diff;

Maybe LLVM can sometimes perform a similar optimization (I think it's not
already done by LLVM, but I can be wrong).

--------------------

Below there's another interesting section:
3. Re-Arrange Data to Be Friendly to SIMD operations

But this looks like an optimization harder to do for a C compiler.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.