Merge #176

176: Fix `validate_unicode_scalar_sequence` build errors on aarch64 r=Bromeon a=therealbnut

Hey, I'm not sure if this is as performant as it was intended, but this at least allows me to compile on `aarch64`.

Before these changes I had these errors:
```
error[E0599]: no function or associated item named `load_unaligned` found for struct `std::arch::aarch64::uint32x4_t` in the current scope
  --> godot-core/src/builtin/string_chars.rs:58:37
   |
58 |             let block = uint32x4_t::load_unaligned(ptr as *const u32);
   |                                     ^^^^^^^^^^^^^^ function or associated item not found in `uint32x4_t`

error[E0425]: cannot find function `vqmovltq_u32` in this scope
    --> godot-core/src/builtin/string_chars.rs:61:17
     |
61   |             if (vqmovltq_u32(block, vdupq_n_u32(char::MAX as u32))).any() {
     |                 ^^^^^^^^^^^^ help: a function with a similar name exists: `vmovl_u32`
     |
    ::: stdarch/crates/core_arch/src/arm_shared/neon/mod.rs:3335:1
     |
3335 | pub unsafe fn vmovl_u32(a: uint32x2_t) -> uint64x2_t {
     | ---------------------------------------------------- similarly named function `vmovl_u32` defined here

error[E0599]: no method named `is_zero` found for struct `std::arch::aarch64::uint32x4_t` in the current scope
  --> godot-core/src/builtin/string_chars.rs:70:14
   |
66 |               if !vandq_u32(
   |  _________________-
67 | |                 vcgtq_u32(block, vdupq_n_u32(0xD7FF)),
68 | |                 vcltq_u32(block, vdupq_n_u32(0xE000)),
69 | |             )
70 | |             .is_zero()
   | |             -^^^^^^^ method not found in `uint32x4_t`
   | |_____________|
```

It now passes `./check.sh test`.

Co-authored-by: therealbnut <[email protected]>

Author: bors[bot]

godot-core/src/builtin/string_chars.rs

@@ -47,28 +47,23 @@ pub fn validate_unicode_scalar_sequence(seq: &[u32]) -> Option<&[char]> {
             ptr = ptr_next;
         }
 
-        // still untested but it should work
         #[cfg(target_arch = "aarch64")]
         loop {
             let ptr_next = ptr.add(4);
             if ptr_next > ptr_end {
                 break;
             }
 
-            let block = uint32x4_t::load_unaligned(ptr as *const u32);
+            let block = vld1q_u32(ptr as *const u32);
 
             // check if has any character bigger than `char::MAX`
-            if (vqmovltq_u32(block, vdupq_n_u32(char::MAX as u32))).any() {
+            if vmaxvq_u32(block) >= char::MAX as u32 {
                 return None;
             }
 
             // check if has any high-surrogate and low-surrogate code points
-            if !vandq_u32(
-                vcgtq_u32(block, vdupq_n_u32(0xD7FF)),
-                vcltq_u32(block, vdupq_n_u32(0xE000)),
-            )
-            .is_zero()
-            {
+            // This is in the range `0xD800..0xE000`.
+            if vminvq_u32(vsubq_u32(block, vdupq_n_u32(0xD800))) < (0xE000 - 0xD800) {
                 return None;
             }