increase max atomic size to 16 on 64-bit platforms (JuliaLang#42268)

RelationalAI · Mar 9, 2024 · 39f141d · 39f141d
1 parent e9c84c8
commit 39f141d
Show file tree

Hide file tree

Showing 4 changed files with 13 additions and 7 deletions.
diff --git a/src/Makefile b/src/Makefile
@@ -26,7 +26,7 @@ endif
 JCFLAGS += -Wold-style-definition -Wstrict-prototypes -Wc++-compat
 
 ifeq ($(USECLANG),1)
-FLAGS += -Wno-return-type-c-linkage
+FLAGS += -Wno-return-type-c-linkage -Wno-atomic-alignment
 endif
 
 FLAGS += -DJL_BUILD_ARCH='"$(ARCH)"'

diff --git a/src/datatype.c b/src/datatype.c
@@ -8,6 +8,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include <stdarg.h>
+#include <stdalign.h>
 #include "julia.h"
 #include "julia_internal.h"
 #include "julia_assert.h"
@@ -1012,7 +1013,7 @@ JL_DLLEXPORT int jl_is_foreign_type(jl_datatype_t *dt)
 
 #if MAX_POINTERATOMIC_SIZE >= 16
 typedef struct _jl_uint128_t {
-    uint64_t a;
+    alignas(16) uint64_t a;
     uint64_t b;
 } jl_uint128_t;
 #endif
@@ -1356,7 +1357,7 @@ JL_DLLEXPORT int jl_atomic_storeonce_bits(jl_datatype_t *dt, char *dst, const jl
 #endif
 #if MAX_POINTERATOMIC_SIZE >= 16
     else if (nb <= 16) {
-        jl_uint128_t y128 = 0;
+        jl_uint128_t y128 = {0};
         jl_uint128_t z128 = zext_read128(src, nb);
         while (1) {
             success = jl_atomic_cmpswap((_Atomic(jl_uint128_t)*)dst, &y128, z128);

diff --git a/src/julia.h b/src/julia.h
@@ -46,11 +46,16 @@
 #endif
 
 // Define the largest size (bytes) of a properly aligned object that the
-// processor family and compiler typically supports without a lock
-// (assumed to be at least a pointer size). Since C is bad at handling 16-byte
-// types, we currently use 8 here as the default.
+// processor family (MAX_ATOMIC_SIZE) and compiler (MAX_POINTERATOMIC_SIZE)
+// typically supports without a lock (assumed to be at least a pointer size)
+// with MAX_POINTERATOMIC_SIZE >= MAX_ATOMIC_SIZE.
+#ifdef _P64
+#define MAX_ATOMIC_SIZE 16
+#define MAX_POINTERATOMIC_SIZE 16
+#else
 #define MAX_ATOMIC_SIZE 8
 #define MAX_POINTERATOMIC_SIZE 8
+#endif
 
 #ifdef _P64
 #define NWORDS(sz) (((sz)+7)>>3)

diff --git a/test/intrinsics.jl b/test/intrinsics.jl
@@ -220,7 +220,7 @@ for TT in (Int8, Int16, Int32, Int64, Int128, Int256, Int512, Complex{Int32}, Co
                 @test_throws TypeError Core.Intrinsics.atomic_pointerreplace(p, T(10), S(3), :sequentially_consistent, :sequentially_consistent)
             end
             @test Core.Intrinsics.pointerref(p, 1, 1) === T(10) === r[]
-            if sizeof(r) > 8
+            if sizeof(r) > 2*sizeof(Int)
                 @test_throws ErrorException("atomic_pointerref: invalid pointer for atomic operation") unsafe_load(p, :sequentially_consistent)
                 @test_throws ErrorException("atomic_pointerset: invalid pointer for atomic operation") unsafe_store!(p, T(1), :sequentially_consistent)
                 @test_throws ErrorException("atomic_pointerswap: invalid pointer for atomic operation") unsafe_swap!(p, T(100), :sequentially_consistent)