[llvm-bugs] [Bug 50793] New: [SIMD] pmin / pmax patterns not recognized

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=50793

            Bug ID: 50793
           Summary: [SIMD] pmin / pmax patterns not recognized
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: WebAssembly
          Assignee: unassignedbugs at nondot.org
          Reporter: clang at evan.coeusgroup.com
                CC: llvm-bugs at lists.llvm.org

The compiler doesn't seem to recognize the obvious pattern for pmin and pmax,
and generates a comparison and a bitselect instead of pmin or pmax.  Quick
example below (or Compiler Explorer at https://godbolt.org/z/Y6K6KssT3)



#include <wasm_simd128.h>

typedef float f32x4 __attribute__((__vector_size__(16)));
typedef int32_t i32x4 __attribute__((__vector_size__(16)));
typedef double f64x2 __attribute__((__vector_size__(16)));
typedef int64_t i64x2 __attribute__((__vector_size__(16)));

f32x4
pmin32(f32x4 a, f32x4 b) {
    i32x4 mask = (b < a);
    return (f32x4) (
        (((i32x4) b) & mask) |
        (((i32x4) a) & ~mask));
}

f32x4
pmax32(f32x4 a, f32x4 b) {
    i32x4 mask = (a < b);
    return (f32x4) (
        (((i32x4) b) & mask) |
        (((i32x4) a) & ~mask));
}

f64x2
pmin64(f64x2 a, f64x2 b) {
    i64x2 mask = (b < a);
    return (f64x2) (
        (((i64x2) b) & mask) |
        (((i64x2) a) & ~mask));
}

f64x2
pmax64(f64x2 a, f64x2 b) {
    i64x2 mask = (a < b);
    return (f64x2) (
        (((i64x2) b) & mask) |
        (((i64x2) a) & ~mask));
}

-- 
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/20210622/42647121/attachment.html>