Bug 1883866: Remove unnecessary memory barriers in generated atomic operations. r=jandem

Remove unnecessary memory barriers to match JIT generated code.

x86:
- Load: No barriers are needed before and after loads.
- Store: No barrier needed before stores.

arm32 and arm64:
- Load: No barrier needed before loads.

Differential Revision: https://phabricator.services.mozilla.com/D203737

js/src/jit/GenerateAtomicOperations.py

@@ -50,8 +50,6 @@ def gen_load(fun_name, cpp_type, size, barrier):
     # - MacroAssembler::wasmLoad
     if cpu_arch in ("x86", "x86_64"):
         insns = ""
-        if barrier:
-            insns += fmt_insn("mfence")
         if size == 8:
             insns += fmt_insn("movb (%[arg]), %[res]")
         elif size == 16:
@@ -61,8 +59,6 @@ def gen_load(fun_name, cpp_type, size, barrier):
         else:
             assert size == 64
             insns += fmt_insn("movq (%[arg]), %[res]")
-        if barrier:
-            insns += fmt_insn("mfence")
         return """
             INLINE_ATTR %(cpp_type)s %(fun_name)s(const %(cpp_type)s* arg) {
                 %(cpp_type)s res;
@@ -78,8 +74,6 @@ def gen_load(fun_name, cpp_type, size, barrier):
         }
     if cpu_arch == "aarch64":
         insns = ""
-        if barrier:
-            insns += fmt_insn("dmb ish")
         if size == 8:
             insns += fmt_insn("ldrb %w[res], [%x[arg]]")
         elif size == 16:
@@ -106,8 +100,6 @@ def gen_load(fun_name, cpp_type, size, barrier):
         }
     if cpu_arch == "arm":
         insns = ""
-        if barrier:
-            insns += fmt_insn("dmb sy")
         if size == 8:
             insns += fmt_insn("ldrb %[res], [%[arg]]")
         elif size == 16:
@@ -141,8 +133,6 @@ def gen_store(fun_name, cpp_type, size, barrier):
     # - MacroAssembler::wasmStore
     if cpu_arch in ("x86", "x86_64"):
         insns = ""
-        if barrier:
-            insns += fmt_insn("mfence")
         if size == 8:
             insns += fmt_insn("movb %[val], (%[addr])")
         elif size == 16: