commit a37e6bc81b52d39cdcfd0f1428f5d6c2b2bc9861
author George Steed <george.steed@arm.com> Mon Aug 19 10:10:21 2024 +0100
committer Frank Barchard <fbarchard@chromium.org> Mon Sep 23 09:29:53 2024 +0000
tree f54055cf9b353432172f2fab8b7db16f8279faf0
parent 8315fa1d3ac3bdfd35f71375a7e993c337d416d9

[AArch64] Re-enable SME only for Linux and new versions of Clang

This was previously disabled in
679e851f653866a49e21f69fe8380bd20123f0ee, so re-enable it but only for
Linux where SME is known to work correctly.

Change-Id: I2626b03f3854b27162df1b55fc6767e02ffe318d
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/5802958
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Justin Green <greenjustin@google.com>

include/libyuv/cpu_support.h

diff --git a/include/libyuv/cpu_support.h b/include/libyuv/cpu_support.h
index ac16575..0d67c34 100644
--- a/include/libyuv/cpu_support.h
+++ b/include/libyuv/cpu_support.h
@@ -72,9 +72,11 @@
 #define VISUALC_HAS_AVX2 1
 #endif  // VisualStudio >= 2012
 
-// Temporary disable SME.
-#if !defined(LIBYUV_DISABLE_SME)
-#define LIBYUV_DISABLE_SME
+// Clang 19 required for SME due to needing __arm_tpidr2_save from compiler-rt,
+// only enabled on Linux for now.
+#if !defined(LIBYUV_DISABLE_SME) && defined(__aarch64__) && \
+    defined(__gnu_linux__) && defined(__clang__) && (__clang_major__ >= 19)
+#define CLANG_HAS_SME 1
 #endif
 
 #ifdef __cplusplus

include/libyuv/rotate_row.h

diff --git a/include/libyuv/rotate_row.h b/include/libyuv/rotate_row.h
index 7672e2c..bbf217f 100644
--- a/include/libyuv/rotate_row.h
+++ b/include/libyuv/rotate_row.h
@@ -50,7 +50,8 @@
 #define HAS_TRANSPOSE4X4_32_NEON
 #endif
 
-#if !defined(LIBYUV_DISABLE_SME) && defined(__aarch64__)
+#if !defined(LIBYUV_DISABLE_SME) && defined(CLANG_HAS_SME) && \
+    defined(__aarch64__)
 #define HAS_TRANSPOSEWXH_SME
 #define HAS_TRANSPOSEUVWXH_SME
 #endif

source/rotate_sme.cc

diff --git a/source/rotate_sme.cc b/source/rotate_sme.cc
index 70e2a0d..dc9ab0c 100644
--- a/source/rotate_sme.cc
+++ b/source/rotate_sme.cc
@@ -18,7 +18,8 @@
 extern "C" {
 #endif
 
-#if !defined(LIBYUV_DISABLE_SME) && defined(__aarch64__)
+#if !defined(LIBYUV_DISABLE_SME) && defined(CLANG_HAS_SME) && \
+    defined(__aarch64__)
 
 __arm_locally_streaming __arm_new("za") void TransposeWxH_SME(
     const uint8_t* src,
@@ -164,7 +165,8 @@
   } while (height > 0);
 }
 
-#endif  // !defined(LIBYUV_DISABLE_SME) && defined(__aarch64__)
+#endif  // !defined(LIBYUV_DISABLE_SME) && defined(CLANG_HAS_SME) &&
+        // defined(__aarch64__)
 
 #ifdef __cplusplus
 }  // extern "C"