[llvm-bugs] [Bug 50803] New: [SIMD] Compiler unable to vectorize trivial loops

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=50803

            Bug ID: 50803
           Summary: [SIMD] Compiler unable to vectorize trivial loops
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: normal
          Priority: P
         Component: Backend: WebAssembly
          Assignee: unassignedbugs at nondot.org
          Reporter: clang at evan.coeusgroup.com
                CC: llvm-bugs at lists.llvm.org

Here is an example (or on Compiler Explorer: https://godbolt.org/z/5Was18P7r):


#include <stdint.h>

typedef int32_t i32x4 __attribute__((__vector_size__(16)));

i32x4 add(i32x4 a, i32x4 b);

i32x4
add(i32x4 a, i32x4 b) {
    i32x4 r;

    #pragma clang loop vectorize(enable)
    for (int i = 0 ; i < 4 ; i++) {
        r[i] = a[i] + b[i];
    }

    return r;
}


Compile with -msimd128 -O3, no vectorization.  The scalar version seems pretty
bad, too.

Even with the pragma the compiler won't vectorize the loop.  I'm wondering if
this might have something to do with the -Wconditional-uninitialized diagnostic
emitted when returning r (it's not enabled by default, -Weverything or
-Wconditional-uninitialized to see it).

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20210622/71f5ba81/attachment.html>