Implement __assume for non-windows builds

[AndyAyersMS]

MethodTable::GetManagedClassObjectIfExists() uses COMPILER_ASSUME to help the native compiler avoid a null check and branch in the common case.

COMPILER_ASSUME expands to __assume but the latter is defined in pal.h as (void)0 when not compiling with MSVC. Seems like we ought to be able try clang's __builtin_assume here and get similar benefits.

MethodTable::GetManagedClassObjectIfExists() is fairly hot since it is called in the fast path of Object.GetType().

If this works in Clang as it does in MSVC, it should give a small improvement in the performance of Object.GetType(), as in the attached program.

using System;

public class GetType
{
    public static volatile anObject = null;

    public static Type TypeReflectionObjectGetType(long count)
    {
        Type type = null;
        anObject = "aString";
        
        for (long i = 0; i < count; i++)
        {
            type = anObject.GetType();
        }
        
        return type;
    }
    
    public static int Main()
    {
        long count = 100000;
        Stopwatch sw = Stopwatch.StartNew();
        Type t = TypeReflectionObjectGetType(count);
        sw.Stop();
        Console.WriteLine("GetType: {0} iterations in {1:F2} ms", count, sw.Elapsed.TotalMilliseconds);
        return (t == typeof(string) ? 100 : 0);
    }
}

[wecing]

(cc @jkotas) It seems that clang warns about side effects in __builtin_assume by default, but VS (C4557) doesn't. Because of -Werror, implementing this feature would produce errors for perfectly fine use cases like this:

__assume(GetNativeHeader() != NULL);
return GetNativeHeader()->ImageBase;

Then should we manually rewrite all use cases of __assume to:

CORCOMPILE_HEADER *h = GetNativeHeader();
__assume(h != NULL);
return h->ImageBase;

, or use -Wno-assume globally?

[jkotas]

I think it should be fine to use -Wno-assume globally.

[ghost]

Due to lack of recent activity, this issue has been marked as a candidate for backlog cleanup. It will be closed if no further activity occurs within 14 more days. Any new comment (by anyone, not necessarily the author) will undo this process.

This process is part of our issue cleanup automation.