From a68b9598731345cd22f1c968c3a9814b592e37f0 Mon Sep 17 00:00:00 2001
From: George Steed <george.steed@arm.com>
Date: Wed, 22 May 2024 10:05:29 +0100
Subject: [PATCH] [AArch64] Add initial build system support for SME

Extend both the CMake and BUILD.gn configurations to support building a
library with the Arm Scalable Matrix Extension (SME). Add an initial
(empty) rotate_sme.cc source file to populate the library for now.

Change-Id: Icd4bd6a8ce72ba132299b00c99478a18a85d869a
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/5588664
Reviewed-by: Justin Green <greenjustin@google.com>
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
---
 BUILD.gn             | 28 ++++++++++++++++++++++++++++
 CMakeLists.txt       |  6 ++++++
 libyuv.gni           |  1 +
 source/rotate_sme.cc | 30 ++++++++++++++++++++++++++++++
 4 files changed, 65 insertions(+)
 create mode 100644 source/rotate_sme.cc

diff --git a/BUILD.gn b/BUILD.gn
index 5bbf77ee0..2f1473a26 100644
--- a/BUILD.gn
+++ b/BUILD.gn
@@ -38,6 +38,9 @@ config("libyuv_config") {
   if (!libyuv_use_sve) {
     defines += [ "LIBYUV_DISABLE_SVE" ]
   }
+  if (!libyuv_use_sme) {
+    defines += [ "LIBYUV_DISABLE_SME" ]
+  }
   if (libyuv_disable_rvv) {
     defines += [ "LIBYUV_DISABLE_RVV" ]
   }
@@ -273,6 +276,31 @@ if (libyuv_use_sve) {
   }
 }
 
+if (libyuv_use_sme) {
+  static_library("libyuv_sme") {
+    sources = [
+      "source/rotate_sme.cc",
+    ]
+
+    deps = [ ":libyuv_internal" ]
+
+    public_configs = [ ":libyuv_config" ]
+
+    # Always enable optimization for Release and NaCl builds (to workaround
+    # crbug.com/538243).
+    if (!is_debug) {
+      configs -= [ "//build/config/compiler:default_optimization" ]
+
+      # Enable optimize for speed (-O2) over size (-Os).
+      # TODO(fbarchard): Consider optimize_speed which is O3.
+      configs += [ "//build/config/compiler:optimize_max" ]
+    }
+
+    # SME is an Armv9-A feature.
+    cflags = [ "-march=armv9-a+sme" ]
+  }
+}
+
 if (libyuv_use_msa) {
   static_library("libyuv_msa") {
     sources = [
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 6dfc93cec..3f521e351 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -110,6 +110,12 @@ if(NOT MSVC)
       ${ly_src_dir}/row_sve.cc)
     TARGET_COMPILE_OPTIONS(${ly_lib_name}_sve PRIVATE -march=armv9-a+sve2)
     LIST(APPEND ly_lib_parts $<TARGET_OBJECTS:${ly_lib_name}_sve>)
+
+    # Enable AArch64 SME kernels.
+    ADD_LIBRARY(${ly_lib_name}_sme OBJECT
+      ${ly_src_dir}/rotate_sme.cc)
+    TARGET_COMPILE_OPTIONS(${ly_lib_name}_sme PRIVATE -march=armv9-a+sme)
+    LIST(APPEND ly_lib_parts $<TARGET_OBJECTS:${ly_lib_name}_sme>)
   endif()
 endif()
 
diff --git a/libyuv.gni b/libyuv.gni
index 6add96ad9..7a002c25f 100644
--- a/libyuv.gni
+++ b/libyuv.gni
@@ -19,6 +19,7 @@ declare_args() {
       current_cpu == "arm64" ||
       (current_cpu == "arm" && (arm_use_neon || arm_optionally_use_neon))
   libyuv_use_sve = current_cpu == "arm64"
+  libyuv_use_sme = current_cpu == "arm64"
   libyuv_use_msa =
       (current_cpu == "mips64el" || current_cpu == "mipsel") && mips_use_msa
   libyuv_use_mmi =
diff --git a/source/rotate_sme.cc b/source/rotate_sme.cc
new file mode 100644
index 000000000..3130cee8f
--- /dev/null
+++ b/source/rotate_sme.cc
@@ -0,0 +1,30 @@
+/*
+ *  Copyright 2024 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/rotate_row.h"
+#include "libyuv/row.h"
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#if !defined(LIBYUV_DISABLE_SME) && defined(__aarch64__)
+
+// TODO: Port rotate kernels to SME.
+
+#endif  // !defined(LIBYUV_DISABLE_SME) && defined(__aarch64__)
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif