[Arm64] Enable SIMD (#14633)

* [Arm64] EXTERNAL_FeatureSIMD_Default = 1 * [Arm64] Enable SIMD compilation * [Arm64] Enable SIMD in crossgen * [Arm64] Allow SIMD in altjit Add flag SIMD16ByteOnly Set COMPlus_SIMD16ByteOnly=1 in x64_arm64_altjit.cmd If SIMD16ByteOnly, limit clear FLAG_USE_AVX2 to disable SIMD32 vector size Enable SIMD in protonjit * Fix #if per feedback
dotnet · Dec 5, 2017 · 2b9e70f · 2b9e70f
1 parent cced4d7
commit 2b9e70f
Show file tree

Hide file tree

Showing 6 changed files with 36 additions and 6 deletions.
diff --git a/src/inc/clrconfigvalues.h b/src/inc/clrconfigvalues.h
@@ -529,14 +529,18 @@ CONFIG_DWORD_INFO_EX(INTERNAL_JitLoopHoistStats, W("JitLoopHoistStats"), 0, "Dis
 CONFIG_DWORD_INFO_EX(INTERNAL_JitDebugLogLoopCloning, W("JitDebugLogLoopCloning"), 0, "In debug builds log places where loop cloning optimizations are performed on the fast path.", CLRConfig::REGUTIL_default);
 CONFIG_DWORD_INFO_EX(INTERNAL_JitVNMapSelLimit, W("JitVNMapSelLimit"), 0, "If non-zero, assert if # of VNF_MapSelect applications considered reaches this", CLRConfig::REGUTIL_default)
 RETAIL_CONFIG_DWORD_INFO(INTERNAL_JitVNMapSelBudget, W("JitVNMapSelBudget"), 100, "Max # of MapSelect's considered for a particular top-level invocation.")
-#if defined(_TARGET_AMD64_) || defined(_TARGET_X86_)
+#if defined(_TARGET_AMD64_) || defined(_TARGET_X86_) || defined(_TARGET_ARM64_)
 #define EXTERNAL_FeatureSIMD_Default 1
-#define EXTERNAL_JitEnableAVX_Default 1
-#else // !defined(_TARGET_AMD64_) && !defined(_TARGET_X86_)
+#else // !(defined(_TARGET_AMD64_) || defined(_TARGET_X86_) || defined(_TARGET_ARM64_))
 #define EXTERNAL_FeatureSIMD_Default 0
+#endif // !(defined(_TARGET_AMD64_) || defined(_TARGET_X86_) || defined(_TARGET_ARM64_))
+#if defined(_TARGET_AMD64_) || defined(_TARGET_X86_)
+#define EXTERNAL_JitEnableAVX_Default 1
+#else // !(defined(_TARGET_AMD64_) || defined(_TARGET_X86_)
 #define EXTERNAL_JitEnableAVX_Default 0
-#endif // !defined(_TARGET_AMD64_) && !defined(_TARGET_X86_)
+#endif // !(defined(_TARGET_AMD64_) || defined(_TARGET_X86_)
 RETAIL_CONFIG_DWORD_INFO_EX(EXTERNAL_FeatureSIMD, W("FeatureSIMD"), EXTERNAL_FeatureSIMD_Default, "Enable SIMD support with companion SIMDVector.dll", CLRConfig::REGUTIL_default)
+RETAIL_CONFIG_DWORD_INFO(INTERNAL_SIMD16ByteOnly, W("SIMD16ByteOnly"), 0, "Limit maximum SIMD vector length to 16 bytes (used by x64_arm64_altjit)")
 RETAIL_CONFIG_DWORD_INFO_EX(EXTERNAL_EnableAVX, W("EnableAVX"), EXTERNAL_JitEnableAVX_Default, "Enable AVX instruction set for wide operations as default", CLRConfig::REGUTIL_default)
 
 #if defined(_TARGET_X86_) || defined(_TARGET_AMD64_)

diff --git a/src/jit/CMakeLists.txt b/src/jit/CMakeLists.txt
@@ -9,6 +9,10 @@ if (CLR_CMAKE_TARGET_ARCH_AMD64 OR (CLR_CMAKE_TARGET_ARCH_I386 AND NOT CLR_CMAKE
   add_definitions(-DFEATURE_HW_INTRINSICS)
 endif ()
 
+if (CLR_CMAKE_TARGET_ARCH_ARM64)
+  add_definitions(-DFEATURE_SIMD)
+endif ()
+
 # JIT_BUILD disables certain PAL_TRY debugging features
 add_definitions(-DJIT_BUILD=1)
 
@@ -141,6 +145,7 @@ set( JIT_ARM64_SOURCES
   lowerarm64.cpp
   lsraarmarch.cpp
   lsraarm64.cpp
+  simd.cpp
   targetarm64.cpp
   unwindarm.cpp
   unwindarm64.cpp

diff --git a/src/jit/protononjit/CMakeLists.txt b/src/jit/protononjit/CMakeLists.txt
@@ -5,8 +5,6 @@ add_definitions(-DFEATURE_NO_HOST)
 add_definitions(-DSELF_NO_HOST)
 remove_definitions(-DFEATURE_MERGE_JIT_AND_ENGINE)
 
-remove_definitions(-DFEATURE_SIMD)
-
 remove_definitions(-DFEATURE_HW_INTRINSICS)
 
 if(FEATURE_READYTORUN)
@@ -15,6 +13,7 @@ endif(FEATURE_READYTORUN)
 
 if (CLR_CMAKE_PLATFORM_ARCH_I386)
     remove_definitions(-D_TARGET_X86_=1)
+    remove_definitions(-DFEATURE_SIMD)
     add_definitions(-D_TARGET_ARM_)
     set(JIT_ARCH_ALTJIT_SOURCES ${JIT_ARM_SOURCES})
     set(JIT_ARCH_LINK_LIBRARIES gcinfo_arm)

diff --git a/src/vm/codeman.cpp b/src/vm/codeman.cpp
@@ -1432,10 +1432,23 @@ void EEJitManager::SetCpuInfo()
             {
                 CPUCompileFlags.Set(CORJIT_FLAGS::CORJIT_FLAG_FEATURE_SIMD);
             }
+
+            if (CLRConfig::GetConfigValue(CLRConfig::INTERNAL_SIMD16ByteOnly) != 0)
+            {
+                CPUCompileFlags.Clear(CORJIT_FLAGS::CORJIT_FLAG_USE_AVX2);
+            }
         }
     }
 #endif // defined(_TARGET_X86_) || defined(_TARGET_AMD64_)
 
+#if defined(_TARGET_ARM64_)
+    static ConfigDWORD fFeatureSIMD;
+    if (fFeatureSIMD.val(CLRConfig::EXTERNAL_FeatureSIMD) != 0)
+    {
+        CPUCompileFlags.Set(CORJIT_FLAGS::CORJIT_FLAG_FEATURE_SIMD);
+    }
+#endif
+
     m_CPUCompileFlags = CPUCompileFlags;
 }
 

diff --git a/src/zap/zapper.cpp b/src/zap/zapper.cpp
@@ -1394,6 +1394,14 @@ void Zapper::InitializeCompilerFlags(CORCOMPILE_VERSION_INFO * pVersionInfo)
 
 #endif // _TARGET_X86_
 
+#if defined(_TARGET_ARM64_)
+    static ConfigDWORD fFeatureSIMD;
+    if (fFeatureSIMD.val(CLRConfig::EXTERNAL_FeatureSIMD) != 0)
+    {
+        m_pOpt->m_compilerFlags.Set(CORJIT_FLAGS::CORJIT_FLAG_FEATURE_SIMD);
+    }
+#endif
+
     if (   m_pOpt->m_compilerFlags.IsSet(CORJIT_FLAGS::CORJIT_FLAG_DEBUG_INFO)
         && m_pOpt->m_compilerFlags.IsSet(CORJIT_FLAGS::CORJIT_FLAG_DEBUG_CODE)
         && m_pOpt->m_compilerFlags.IsSet(CORJIT_FLAGS::CORJIT_FLAG_PROF_ENTERLEAVE))

diff --git a/tests/x64_arm64_altjit.cmd b/tests/x64_arm64_altjit.cmd
@@ -4,3 +4,4 @@ set COMPlus_AltJitName=protononjit.dll
 set COMPlus_NoGuiOnAssert=1
 set COMPlus_ContinueOnAssert=0
 set COMPlus_AltJitAssertOnNYI=1
+set COMPlus_SIMD16ByteOnly=1