Various AArch64 optimizations (https://github.com/Shopify/ruby/pull/433)

* When we're storing an immediate 0 value at a memory address, we
  can use STUR XZR, Xd instead of loading 0 into a register and
  then storing that register.
* When we're moving 0 into an argument register, we can use
  MOV Xd, XZR instead of loading the value into a register first.
* In the newarray instruction, we can skip looking at the stack at
  all if the number of values we're using is 0.

1 files changed, 8 insertions, 3 deletions

diff --git a/yjit/src/codegen.rs b/yjit/src/codegen.rs
index e23171d2a0..1336fe3c57 100644
--- a/yjit/src/codegen.rs
+++ b/yjit/src/codegen.rs
@@ -1168,9 +1168,14 @@ fn gen_newarray(
     // Save the PC and SP because we are allocating
     jit_prepare_routine_call(jit, ctx, asm);
 
-    let offset_magnitude = SIZEOF_VALUE as u32 * n;
-    let values_opnd = ctx.sp_opnd(-(offset_magnitude as isize));
-    let values_ptr = asm.lea(values_opnd);
+    // If n is 0, then elts is never going to be read, so we can just pass null
+    let values_ptr = if n == 0 {
+        Opnd::UImm(0)
+    } else {
+        let offset_magnitude = SIZEOF_VALUE as u32 * n;
+        let values_opnd = ctx.sp_opnd(-(offset_magnitude as isize));
+        asm.lea(values_opnd)
+    };
 
     // call rb_ec_ary_new_from_values(struct rb_execution_context_struct *ec, long n, const VALUE *elts);
     let new_ary = asm.ccall(