[llvm-bugs] [Bug 50805] New: [SIMD] min / max patterns not recognized

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=50805

            Bug ID: 50805
           Summary: [SIMD] min / max 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

Similar to #50793, but for min/max instead of pmin/pmax.  Probably less
important than pmin/pmax since most code doesn't bother checking for NaN, but
still nice to have.

If we add a quick check so the the code will return NaN if it is found, we end
up with the behavior of min/max instead of pmin/pmax.  Example
(https://godbolt.org/z/WEr9173h6):


#include <wasm_simd128.h>

#pragma clang diagnostic ignored "-Wmissing-prototypes"
#pragma clang diagnostic ignored "-Wvector-conversion"

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
min32(f32x4 a, f32x4 b) {
    i32x4 mask = (a != a) | (a < b);
    return (f32x4) (
        (((i32x4) a) &  mask) |
        (((i32x4) b) & ~mask));
}

f64x2
min64(f64x2 a, f64x2 b) {
    i64x2 mask = (a != a) | (a < b);
    return (f64x2) (
        (((i64x2) a) &  mask) |
        (((i64x2) b) & ~mask));
}

f32x4
max32(f32x4 a, f32x4 b) {
    i32x4 mask = (a != a) | (a > b);
    return (f32x4) (
        (((i32x4) a) &  mask) |
        (((i32x4) b) & ~mask));
}

f64x2
max64(f64x2 a, f64x2 b) {
    i64x2 mask = (a != a) | (a > b);
    return (f64x2) (
        (((i64x2) a) &  mask) |
        (((i64x2) b) & ~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/20210622/889617e0/attachment.html>