[compiler-rt] [builtin] Implement __atomic_fetch_nand_* builtins

[krytarowski]

Summary

Increase compatibility with GCC libatomic.

[jfb]

Please clarify what this is adding in the patch description.

compiler-rt/lib/builtins/atomic.c

334

Is there no lockfree version of this? If not, why? It should be there for relevant sizes.

[krytarowski]

In D81105#2072158, @jfb wrote:

Please clarify what this is adding in the patch description.

I'm adding builtins that exist in GCC libatomic. What should I document extra?

compiler-rt/lib/builtins/atomic.c

334

Not for NAND, at least not for my CPU & compiler I tested.

[theraven]

compiler-rt/lib/builtins/atomic.c

334

I don't understand this comment. A CPU that has load-linked / store-conditional or atomic compare-and-exchange for values that are this size has a lock-free version. That will be used by the compiler with some flags, so this implementation will not be atomic with respect to versions that use them. That's a big problem.

compiler-rt/lib/builtins/atomic.c

type __atomic_fetch_##opname##_##n(type *ptr, type val, int model) {         \
     Lock *l = lock_for_pointer(ptr);                                           \
     lock(l);                                                                   \
     type tmp = *ptr;                                                           \
     *ptr = tmp op val;                                                         \
     unlock(l);                                                                 \
     return tmp;                                                                \
   }
 
+// No lockfree atomic_fetch_*_explicit variation.
+#define ATOMIC_RMW_NEG(n, lockfree, type, opname, op)                          \
+  type __atomic_fetch_##opname##_##n(type *ptr, type val, int model) {         \
+    Lock *l = lock_for_pointer(ptr);                                           \
+    lock(l);                                                                   \
+    type tmp = *ptr;                                                           \
+    *ptr = ~(tmp op val);                                                      \
+    unlock(l);                                                                 \
+    return tmp;                                                                \
+  }
+
 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, add, +)
 OPTIMISED_CASES
 #undef OPTIMISED_CASE
 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, sub, -)
 OPTIMISED_CASES
 #undef OPTIMISED_CASE
 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, and, &)
 OPTIMISED_CASES
 #undef OPTIMISED_CASE
 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, or, |)
 OPTIMISED_CASES
 #undef OPTIMISED_CASE
 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, xor, ^)
 OPTIMISED_CASES
 #undef OPTIMISED_CASE
+#define OPTIMISED_CASE(n, lockfree, type)                                      \
+  ATOMIC_RMW_NEG(n, lockfree, type, nand, &)
+OPTIMISED_CASES
+#undef OPTIMISED_CASE