[clang][driver] Allow unaligned access on ARMv7 and higher by default (#82400)

ARM's Clang and GCC embedded compilers default to allowing unaligned
access for ARMv7+. This patch changes the Clang driver default to match.
Users can opt out with `-mno-unaligned-access`.

Fixes #59560

Author: ilovepi

clang/docs/ReleaseNotes.rst

@@ -302,6 +302,17 @@ X86 Support
 Arm and AArch64 Support
 ^^^^^^^^^^^^^^^^^^^^^^^
 
+- ARMv7+ targets now default to allowing unaligned access, except Armv6-M, and
+  Armv8-M without the Main Extension. Baremetal targets should check that the
+  new default will work with their system configurations, since it requires
+  that SCTLR.A is 0, SCTLR.U is 1, and that the memory in question is
+  configured as "normal" memory. This brings Clang in-line with the default
+  settings for GCC and Arm Compiler. Aside from making Clang align with other
+  compilers, changing the default brings major performance and code size
+  improvements for most targets. We have not changed the default behavior for
+  ARMv6, but may revisit that decision in the future. Users can restore the old
+  behavior with -m[no-]unaligned-access.
+
 Android Support
 ^^^^^^^^^^^^^^^
 

clang/lib/Driver/ToolChains/Arch/ARM.cpp

@@ -890,25 +890,25 @@ llvm::ARM::FPUKind arm::getARMTargetFeatures(const Driver &D,
     // SCTLR.U bit, which is architecture-specific. We assume ARMv6
     // Darwin and NetBSD targets support unaligned accesses, and others don't.
     //
-    // ARMv7 always has SCTLR.U set to 1, but it has a new SCTLR.A bit
-    // which raises an alignment fault on unaligned accesses. Linux
-    // defaults this bit to 0 and handles it as a system-wide (not
-    // per-process) setting. It is therefore safe to assume that ARMv7+
-    // Linux targets support unaligned accesses. The same goes for NaCl
-    // and Windows.
+    // ARMv7 always has SCTLR.U set to 1, but it has a new SCTLR.A bit which
+    // raises an alignment fault on unaligned accesses. Assume ARMv7+ supports
+    // unaligned accesses, except ARMv6-M, and ARMv8-M without the Main
+    // Extension. This aligns with the default behavior of ARM's downstream
+    // versions of GCC and Clang.
     //
-    // The above behavior is consistent with GCC.
+    // Users can change the default behavior via -m[no-]unaliged-access.
     int VersionNum = getARMSubArchVersionNumber(Triple);
     if (Triple.isOSDarwin() || Triple.isOSNetBSD()) {
       if (VersionNum < 6 ||
           Triple.getSubArch() == llvm::Triple::SubArchType::ARMSubArch_v6m)
         Features.push_back("+strict-align");
-    } else if (Triple.isOSLinux() || Triple.isOSNaCl() ||
-               Triple.isOSWindows()) {
-      if (VersionNum < 7)
-        Features.push_back("+strict-align");
-    } else
+    } else if (VersionNum < 7 ||
+               Triple.getSubArch() ==
+                   llvm::Triple::SubArchType::ARMSubArch_v6m ||
+               Triple.getSubArch() ==
+                   llvm::Triple::SubArchType::ARMSubArch_v8m_baseline) {
       Features.push_back("+strict-align");
+    }
   }
 
   // llvm does not support reserving registers in general. There is support

clang/test/Driver/arm-alignment.c

@@ -22,6 +22,21 @@
 // RUN: %clang -target armv7-windows -### %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-UNALIGNED-ARM < %t %s
 
+// RUN: %clang --target=armv6 -### %s 2> %t
+// RUN: FileCheck --check-prefix=CHECK-ALIGNED-ARM < %t %s
+
+// RUN: %clang --target=armv7 -### %s 2> %t
+// RUN: FileCheck --check-prefix=CHECK-UNALIGNED-ARM < %t %s
+
+// RUN: %clang -target thumbv6m-none-gnueabi -mcpu=cortex-m0 -### %s 2> %t
+// RUN: FileCheck --check-prefix CHECK-ALIGNED-ARM <%t %s
+
+// RUN: %clang -target thumb-none-gnueabi -mcpu=cortex-m0 -### %s 2> %t
+// RUN: FileCheck --check-prefix CHECK-ALIGNED-ARM <%t %s
+
+// RUN: %clang -target thumbv8m.base-none-gnueabi -### %s 2> %t
+// RUN: FileCheck --check-prefix CHECK-ALIGNED-ARM <%t %s
+
 // RUN: %clang --target=aarch64 -munaligned-access -### %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-UNALIGNED-AARCH64 < %t %s