Merge #636

636: CachePadded: update alignment based on golang's CacheLinePadSize r=jeehoonkang a=taiki-e

This updates alignment of `CachePadded` on the following architectures based on golang's [`CacheLinePadSize`](https://github.com/golang/go/blob/3510a1e32cbc86b73db143aefcc00aadc44c27bd/src/internal/cpu/cpu.go#L20).

- On [powerpc64](https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_ppc64x.go#L9), use 128-byte alignment.
- On [arm](https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_arm.go#L7), [mips](https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_mips.go#L7), [mips64](https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_mips64x.go#L9), and [riscv64](https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_riscv64.go#L7), use 32-byte alignment.
- On [s390x](https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_s390x.go#L7), use 256-byte alignment.

On x86_64 and aarch64, use 128-byte alignment as before. See #331 and #499 for reasons.
On all others, use 64-byte alignment as before.

Closes #427

cc @stjepang (when this merged, I'll send a same patch to smol-rs/cache-padded)
FYI @yoshuawuyts @cuviper (because you all were [discussing about adding `CachePadded` to the standard library](https://rust-lang.zulipchat.com/#narrow/stream/219381-t-libs/topic/.60std.3A.3Amem.3A.3ACachePadded.60))

Co-authored-by: Taiki Endo <[email protected]>

Author: bors[bot]

crossbeam-utils/src/cache_padded.rs

@@ -13,7 +13,9 @@ use core::ops::{Deref, DerefMut};
 ///
 /// Cache lines are assumed to be N bytes long, depending on the architecture:
 ///
-/// * On x86-64 and aarch64, N = 128.
+/// * On x86-64, aarch64, and powerpc64, N = 128.
+/// * On arm, mips, mips64, and riscv64, N = 32.
+/// * On s390x, N = 256.
 /// * On all others, N = 64.
 ///
 /// Note that N is just a reasonable guess and is not guaranteed to match the actual cache line
@@ -64,13 +66,63 @@ use core::ops::{Deref, DerefMut};
 // - https://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-optimization-manual.pdf
 // - https://github.com/facebook/folly/blob/1b5288e6eea6df074758f877c849b6e73bbb9fbb/folly/lang/Align.h#L107
 //
-// ARM's big.LITTLE architecture has asymmetric cores and "big" cores have 128 byte cache line size
+// ARM's big.LITTLE architecture has asymmetric cores and "big" cores have 128-byte cache line size.
+//
 // Sources:
 // - https://www.mono-project.com/news/2016/09/12/arm64-icache/
 //
-#[cfg_attr(any(target_arch = "x86_64", target_arch = "aarch64"), repr(align(128)))]
+// powerpc64 has 128-byte cache line size.
+//
+// Sources:
+// - https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_ppc64x.go#L9
+#[cfg_attr(
+    any(
+        target_arch = "x86_64",
+        target_arch = "aarch64",
+        target_arch = "powerpc64",
+    ),
+    repr(align(128))
+)]
+// arm, mips, mips64, and riscv64 have 32-byte cache line size.
+//
+// Sources:
+// - https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_arm.go#L7
+// - https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_mips.go#L7
+// - https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_mipsle.go#L7
+// - https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_mips64x.go#L9
+// - https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_riscv64.go#L7
+#[cfg_attr(
+    any(
+        target_arch = "arm",
+        target_arch = "mips",
+        target_arch = "mips64",
+        target_arch = "riscv64",
+    ),
+    repr(align(32))
+)]
+// s390x has 256-byte cache line size.
+//
+// Sources:
+// - https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_s390x.go#L7
+#[cfg_attr(target_arch = "s390x", repr(align(256)))]
+// x86 and wasm have 64-byte cache line size.
+//
+// Sources:
+// - https://github.com/golang/go/blob/dda2991c2ea0c5914714469c4defc2562a907230/src/internal/cpu/cpu_x86.go#L9
+// - https://github.com/golang/go/blob/3dd58676054223962cd915bb0934d1f9f489d4d2/src/internal/cpu/cpu_wasm.go#L7
+//
+// All others are assumed to have 64-byte cache line size.
 #[cfg_attr(
-    not(any(target_arch = "x86_64", target_arch = "aarch64")),
+    not(any(
+        target_arch = "x86_64",
+        target_arch = "aarch64",
+        target_arch = "powerpc64",
+        target_arch = "arm",
+        target_arch = "mips",
+        target_arch = "mips64",
+        target_arch = "riscv64",
+        target_arch = "s390x",
+    )),
     repr(align(64))
 )]
 pub struct CachePadded<T> {