[AIX] Use posix_memalign instead of malloc to ensure 16 byte alignment

[ZarkoCA]

Summary

On AIX in 32bit mode, malloc may not be 16 byte aligned as required for vector
data types. This causes these test cases to generate incorrect output.
Using posix_memalign instead we ensure 16 byte alignment in 64 and
32 bit mode and correct output.

[bmahjour]

SingleSource/UnitTests/Vector/Altivec/alti.expandfft.c

38

The alignment requirements for the VMX instructions are OS-agnostic, so ideally we should just replace malloc with posix_memalign (without checking for AIX). I believe this test runs only on PPC platforms, so should be safe to change.

[ZarkoCA]

Remove AIX check for posix_memalign and use across platforms.

[ZarkoCA]

SingleSource/UnitTests/Vector/Altivec/alti.expandfft.c

38

I had read that Mac OS/X doesn't support posix_memalign and wanted to be conservative with the change. But, AFAIK, we don't even support that platform anymore on PPC, so yes, it should be safe to change for all PPC platforms.

[bmahjour]

LGTM

SingleSource/UnitTests/Vector/Altivec/alti.expandfft.c

38

Tried on my Mac (Darwin Kernel Version 20.4.0) and it worked. In any case, if some OS shows a failure, we could go back and guard those ones.

[ZarkoCA]

Landed in https://github.com/llvm/llvm-test-suite/commit/7d06bc9ba14517d7ea795c8ba1be6adc495801a9

SingleSource/UnitTests/Vector/Altivec/alti.expandfft.c

*/
 
    int first,i,icase,it,ln2,n;
    int nits=1000000;
    static float seed = 331.0;
    float error,fnm1,sign,z0,z1,ggl();
    float *x,*y,*z,*w;
    double t1,mflops;
 /* allocate storage for x,y,z,w on 4-word bndr. */
-   x = (float *) malloc(8*N);
-   y = (float *) malloc(8*N);
-   z = (float *) malloc(8*N);
-   w = (float *) malloc(4*N);
+   posix_memalign((void **)&x,16,8*N);
+   posix_memalign((void **)&y,16,8*N);
+   posix_memalign((void **)&z,16,8*N);
+   posix_memalign((void **)&w,16,4*N);
    n     = 2;
    for(ln2=1;ln2<21;ln2++){
       first = 1;
       for(icase=0;icase<2;icase++){
          if(first){
             for(i=0;i<2*n;i+=2){
                z0 = ggl(&seed);     /* real part of array */
                z1 = ggl(&seed);     /* imaginary part of array */

SingleSource/UnitTests/Vector/Altivec/alti.stepfft.c

/* SSE version of cfft2 - uses Apple intrinsics
    W. Petersen, SAM. Math. ETHZ 2 May, 2002 */
    int first,i,icase,it,n;
    int nits=1000;              /* number of iterations for timing test */
    float error,fnm1,sign,z0,z1;
    static float seed = 331.0;
    float *x,*y,*z,*w;
    float t1,ln2,mflops;
 /* allocate storage for x,y,z,w on 4-word bndr. */
-   x = (float *)malloc(8*N);
-   y = (float *)malloc(8*N);
-   z = (float *)malloc(8*N);
-   w = (float *)malloc(4*N);
+   posix_memalign((void **)&x,16,8*N);
+   posix_memalign((void **)&y,16,8*N);
+   posix_memalign((void **)&z,16,8*N);
+   posix_memalign((void **)&w,16,4*N);
    first = 1;
    for(icase=0;icase<2;icase++){
    if(first){
       for(i=0;i<2*N;i+=2){
          z0 = ggl(&seed);     /* real part of array */
          z1 = ggl(&seed);     /* imaginary part of array */
          x[i] = z0;
          z[i] = z0;           /* copy of initial real data */