[compiler-rt] Workaround for endless recursion in ctzdi2

[JDevlieghere]

Summary

Workaround for LLVM bug 11663. Prevent endless recursion in
c?zdi2(), where calls to builtin_c?z() are expanded to
c?zdi2() instead of c?zsi2().

Instead of placing this workaround in c?zdi2.c, put it in this global
header to prevent other C files from making the detour through
__c?zdi2() as well.

This problem has been observed on FreeBSD for sparc64 and mips64 with
GCC 4.2.1, and for riscv with GCC 5.2.0. Presumably it's any version of
GCC, and targeting an arch that does not have dedicated bit counting
instructions.

PR: https://bugs.llvm.org/show_bug.cgi?id=11663

This problem was brought to my attention by Adrian Glaubitz yesterday at
FOSDEM. He asked me if I could help him get this upstream. The workaround is
already present in Rust's version of compiler-rt, but obviously they'd like to have
this upstream: https://github.com/rust-lang/compiler-rt/pull/35/files

[joerg]

I really, really dislike this patch. It is using very blunt force to workaround a GCC bug. The comment is too verbose as well. Please try the following change from NetBSD instead:

https://github.com/NetBSD/src/commit/76ddb1c005bace0912c6091434e0e6560942bf7f#diff-13bc0724eb7139f6dec41626560c0ba7

[JDevlieghere]

Feedback Joerg: go with NetBSD's workaround.

[compnerd]

I believe that the commit message needs to be updated for the change. This is a pretty distasteful patch IMO, but, I don't see a better solution here :-(.

lib/builtins/clzdi2.c

  *
  * ===----------------------------------------------------------------------===
  */
 
 #include "int_lib.h"
 
 /* Returns: the number of leading 0-bits */
 
+#if !defined(__clang__) && (defined(__sparc64__) || defined(__mips64) || defined(__riscv__))
+/* gcc resolves __builtin_clz -> __clzdi2 leading to infinite recursion */
+#define __builtin_clz(a) __clzsi2(a)
+extern si_int __clzsi2(si_int);
+#endif
+
 /* Precondition: a != 0 */
 
 COMPILER_RT_ABI si_int
 __clzdi2(di_int a)
 {
     dwords x;
     x.all = a;
     const si_int f = -(x.s.high == 0);
     return __builtin_clz((x.s.high & ~f) | (x.s.low & f)) +
            (f & ((si_int)(sizeof(si_int) * CHAR_BIT)));
 }

lib/builtins/ctzdi2.c

  *
  * ===----------------------------------------------------------------------===
  */
 
 #include "int_lib.h"
 
 /* Returns: the number of trailing 0-bits  */
 
+#if !defined(__clang__) && (defined(__sparc64__) || defined(__mips64) || defined(__riscv__))
+/* gcc resolves __builtin_ctz -> __ctzdi2 leading to infinite recursion */
+#define __builtin_ctz(a) __ctzsi2(a)
+extern si_int __ctzsi2(si_int);
+#endif
+
 /* Precondition: a != 0 */
 
 COMPILER_RT_ABI si_int
 __ctzdi2(di_int a)
 {
     dwords x;
     x.all = a;
     const si_int f = -(x.s.low == 0);
     return __builtin_ctz((x.s.high & f) | (x.s.low & ~f)) +
               (f & ((si_int)(sizeof(si_int) * CHAR_BIT)));
 }