Add support for delegate GDV and method-based vtable GDV (#68703)

Add support for instrumenting delegate calls and vtable calls into method handle histograms. Use these histograms to do GDV for delegate calls and also support method-based GDV for vtable calls.

For instrumentation we now support class probes at interface call sites, method probes at delegate call sites and both class probes and method probes at vtable call sites. For vtable calls, when turned on, instrumentation produces both histograms as PGO data so that the JIT can later make the choice about what is the best form of guard to use at that site.

For guarding, there are some things to take into account. Delegate calls currently (practically) always point to precode, so this PR is just guarding on getFunctionFixedEntryPoint which returns the precode address, and this is generally quite cheap (same cost as class-based GDV). That's the case for delegates pointing to instance methods anyway, this PR does not support static methods yet -- those will be more expensive.

For vtable calls the runtime will backpatch the slots when tiering, so the JIT guards the address retrieved from the vtable against an indirection of the slot, which is slightly more expensive than a class-based guard.

Currently the instrumentation is enabled conditionally with COMPlus_JitDelegateProfiling=1 (for delegates) and COMPlus_JitVTableProfiling=1 (for vtable calls). Currently delegate profiling is turned on by default while vtable profiling is off by default.

Author: jakobbotsch

eng/pipelines/common/templates/runtimes/run-test-job.yml

@@ -227,13 +227,13 @@ jobs:
         timeoutInMinutes: 300
       ${{ else }}:
         timeoutInMinutes: 200
-    ${{ if in(parameters.testGroup, 'outerloop', 'jit-experimental', 'pgo', 'jit-cfg') }}:
+    ${{ if in(parameters.testGroup, 'outerloop', 'jit-experimental', 'jit-cfg') }}:
       timeoutInMinutes: 270
     ${{ if in(parameters.testGroup, 'gc-longrunning', 'gc-simulator') }}:
       timeoutInMinutes: 480
     ${{ if in(parameters.testGroup, 'jitstress', 'jitstress-isas-arm', 'jitstressregs-x86', 'jitstressregs', 'jitstress2-jitstressregs', 'gcstress0x3-gcstress0xc', 'ilasm') }}:
       timeoutInMinutes: 390
-    ${{ if in(parameters.testGroup, 'gcstress-extra', 'r2r-extra', 'clrinterpreter') }}:
+    ${{ if in(parameters.testGroup, 'gcstress-extra', 'r2r-extra', 'clrinterpreter', 'pgo') }}:
       timeoutInMinutes: 510
     ${{ if eq(parameters.testGroup, 'jitstress-isas-x86') }}:
       timeoutInMinutes: 960
@@ -397,7 +397,7 @@ jobs:
           ${{ if eq(parameters.runtimeFlavor, 'mono') }}:
             # tiered compilation isn't done on mono yet
             scenarios:
-            - normal 
+            - normal
           ${{ elseif eq(variables['Build.Reason'], 'PullRequest') }}:
             scenarios:
             - no_tiered_compilation
@@ -545,7 +545,9 @@ jobs:
           - defaultpgo
           - dynamicpgo
           - fullpgo
+          - fullpgo_methodprofiling
           - fullpgo_random_gdv
+          - fullpgo_random_gdv_methodprofiling_only
           - fullpgo_random_edge
           - fullpgo_random_gdv_edge
         ${{ if in(parameters.testGroup, 'gc-longrunning') }}:
@@ -568,7 +570,6 @@ jobs:
           - jitelthookenabled_tiered
         ${{ if in(parameters.testGroup, 'jit-experimental') }}:
           scenarios:
-          - jitosr
           - jitosr_stress
           - jitosr_pgo
           - jitosr_stress_random

eng/pipelines/coreclr/libraries-pgo.yml

@@ -47,7 +47,7 @@ jobs:
     helixQueueGroup: libraries
     helixQueuesTemplate: /eng/pipelines/coreclr/templates/helix-queues-setup.yml
     jobParameters:
-      timeoutInMinutes: 150
+      timeoutInMinutes: 600
       testScope: innerloop
       liveRuntimeBuildConfig: checked
       dependsOnTestBuildConfiguration: Release

eng/pipelines/libraries/run-test-job.yml

@@ -173,10 +173,11 @@ jobs:
               - defaultpgo
               - dynamicpgo
               - fullpgo
+              - fullpgo_methodprofiling
               - fullpgo_random_gdv
+              - fullpgo_random_gdv_methodprofiling_only
               - fullpgo_random_edge
               - fullpgo_random_gdv_edge
-              - jitosr
               - jitosr_stress
               - jitosr_stress_random
               - jitosr_pgo

src/coreclr/inc/corinfo.h

@@ -640,9 +640,14 @@ enum CorInfoHelpFunc
     CORINFO_HELP_STACK_PROBE,               // Probes each page of the allocated stack frame
 
     CORINFO_HELP_PATCHPOINT,                // Notify runtime that code has reached a patchpoint
+    CORINFO_HELP_PARTIAL_COMPILATION_PATCHPOINT,  // Notify runtime that code has reached a part of the method that wasn't originally jitted.
+
     CORINFO_HELP_CLASSPROFILE32,            // Update 32-bit class profile for a call site
     CORINFO_HELP_CLASSPROFILE64,            // Update 64-bit class profile for a call site
-    CORINFO_HELP_PARTIAL_COMPILATION_PATCHPOINT,  // Notify runtime that code has reached a part of the method that wasn't originally jitted.
+    CORINFO_HELP_DELEGATEPROFILE32,         // Update 32-bit method profile for a delegate call site
+    CORINFO_HELP_DELEGATEPROFILE64,         // Update 64-bit method profile for a delegate call site
+    CORINFO_HELP_VTABLEPROFILE32,           // Update 32-bit method profile for a vtable call site
+    CORINFO_HELP_VTABLEPROFILE64,           // Update 64-bit method profile for a vtable call site
 
     CORINFO_HELP_VALIDATE_INDIRECT_CALL,    // CFG: Validate function pointer
     CORINFO_HELP_DISPATCH_INDIRECT_CALL,    // CFG: Validate and dispatch to pointer

src/coreclr/inc/corjit.h

@@ -330,34 +330,36 @@ class ICorJitInfo : public ICorDynamicInfo
 
     // Data structure for a single class probe using 32-bit count.
     //
-    // CLASS_FLAG and INTERFACE_FLAG are placed into the Other field in the schema
+    // CLASS_FLAG, INTERFACE_FLAG and DELEGATE_FLAG are placed into the Other field in the schema.
+    // If CLASS_FLAG is set the handle table consists of type handles, and otherwise method handles.
     //
     // Count is the number of times a call was made at that call site.
     //
     // SIZE is the number of entries in the table.
     //
     // SAMPLE_INTERVAL must be >= SIZE. SAMPLE_INTERVAL / SIZE
     // gives the average number of calls between table updates.
-    //
-    struct ClassProfile32
+    // 
+    struct HandleHistogram32
     {
         enum
         {
             SIZE = 8,
             SAMPLE_INTERVAL = 32,
             CLASS_FLAG     = 0x80000000,
             INTERFACE_FLAG = 0x40000000,
-            OFFSET_MASK    = 0x3FFFFFFF
+            DELEGATE_FLAG  = 0x20000000,
+            OFFSET_MASK    = 0x0FFFFFFF
         };
 
         uint32_t Count;
-        CORINFO_CLASS_HANDLE ClassTable[SIZE];
+        void* HandleTable[SIZE];
     };
 
-    struct ClassProfile64
+    struct HandleHistogram64
     {
         uint64_t Count;
-        CORINFO_CLASS_HANDLE ClassTable[ClassProfile32::SIZE];
+        void* HandleTable[HandleHistogram32::SIZE];
     };
 
     enum class PgoInstrumentationKind
@@ -387,7 +389,7 @@ class ICorJitInfo : public ICorDynamicInfo
         Done = None, // All instrumentation schemas must end with a record which is "Done"
         BasicBlockIntCount = (DescriptorMin * 1) | FourByte, // basic block counter using unsigned 4 byte int
         BasicBlockLongCount = (DescriptorMin * 1) | EightByte, // basic block counter using unsigned 8 byte int
-        HandleHistogramIntCount = (DescriptorMin * 2) | FourByte | AlignPointer, // 4 byte counter that is part of a type histogram. Aligned to match ClassProfile32's alignment.
+        HandleHistogramIntCount = (DescriptorMin * 2) | FourByte | AlignPointer, // 4 byte counter that is part of a type histogram. Aligned to match HandleHistogram32's alignment.
         HandleHistogramLongCount = (DescriptorMin * 2) | EightByte, // 8 byte counter that is part of a type histogram
         HandleHistogramTypes = (DescriptorMin * 3) | TypeHandle, // Histogram of type handles
         HandleHistogramMethods = (DescriptorMin * 3) | MethodHandle, // Histogram of method handles
@@ -396,6 +398,7 @@ class ICorJitInfo : public ICorDynamicInfo
         EdgeIntCount = (DescriptorMin * 6) | FourByte, // edge counter using unsigned 4 byte int
         EdgeLongCount = (DescriptorMin * 6) | EightByte, // edge counter using unsigned 8 byte int
         GetLikelyClass = (DescriptorMin * 7) | TypeHandle, // Compressed get likely class data
+        GetLikelyMethod = (DescriptorMin * 7) | MethodHandle, // Compressed get likely method data
     };
 
     struct PgoInstrumentationSchema
@@ -418,7 +421,7 @@ class ICorJitInfo : public ICorDynamicInfo
         Sampling= 6,    // PGO data derived from sampling
     };
 
-#define DEFAULT_UNKNOWN_TYPEHANDLE 1
+#define DEFAULT_UNKNOWN_HANDLE 1
 #define UNKNOWN_HANDLE_MIN 1
 #define UNKNOWN_HANDLE_MAX 33
 

src/coreclr/inc/jiteeversionguid.h

@@ -43,11 +43,11 @@ typedef const GUID *LPCGUID;
 #define GUID_DEFINED
 #endif // !GUID_DEFINED
 
-constexpr GUID JITEEVersionIdentifier = { /* af5b6632-6fbe-4a2e-82d6-24487a138e4a */
-    0xaf5b6632,
-    0x6fbe,
-    0x4a2e,
-    {0x82, 0xd6, 0x24, 0x48, 0x7a, 0x13, 0x8e, 0x4a}
+constexpr GUID JITEEVersionIdentifier = { /* f2faa5fc-a1ec-4244-aebb-5597bfd7153a */
+    0xf2faa5fc,
+    0xa1ec,
+    0x4244,
+    {0xae, 0xbb, 0x55, 0x97, 0xbf, 0xd7, 0x15, 0x3a}
   };
 
 //////////////////////////////////////////////////////////////////////////////////////////////////////////

src/coreclr/inc/jithelpers.h

@@ -328,9 +328,14 @@
 #endif
 
     JITHELPER(CORINFO_HELP_PATCHPOINT, JIT_Patchpoint, CORINFO_HELP_SIG_REG_ONLY)
+    JITHELPER(CORINFO_HELP_PARTIAL_COMPILATION_PATCHPOINT, JIT_PartialCompilationPatchpoint, CORINFO_HELP_SIG_REG_ONLY)
+
     JITHELPER(CORINFO_HELP_CLASSPROFILE32, JIT_ClassProfile32, CORINFO_HELP_SIG_REG_ONLY)
     JITHELPER(CORINFO_HELP_CLASSPROFILE64, JIT_ClassProfile64, CORINFO_HELP_SIG_REG_ONLY)
-    JITHELPER(CORINFO_HELP_PARTIAL_COMPILATION_PATCHPOINT, JIT_PartialCompilationPatchpoint, CORINFO_HELP_SIG_REG_ONLY)
+    JITHELPER(CORINFO_HELP_DELEGATEPROFILE32, JIT_DelegateProfile32, CORINFO_HELP_SIG_REG_ONLY)
+    JITHELPER(CORINFO_HELP_DELEGATEPROFILE64, JIT_DelegateProfile64, CORINFO_HELP_SIG_REG_ONLY)
+    JITHELPER(CORINFO_HELP_VTABLEPROFILE32, JIT_VTableProfile32, CORINFO_HELP_SIG_4_STACK)
+    JITHELPER(CORINFO_HELP_VTABLEPROFILE64, JIT_VTableProfile64, CORINFO_HELP_SIG_4_STACK)
 
 #if defined(TARGET_AMD64) || defined(TARGET_ARM64)
     JITHELPER(CORINFO_HELP_VALIDATE_INDIRECT_CALL, JIT_ValidateIndirectCall, CORINFO_HELP_SIG_REG_ONLY)

src/coreclr/inc/readytorun.h

@@ -16,7 +16,7 @@
 
 // Keep these in sync with src/coreclr/tools/Common/Internal/Runtime/ModuleHeaders.cs
 #define READYTORUN_MAJOR_VERSION 0x0006
-#define READYTORUN_MINOR_VERSION 0x0001
+#define READYTORUN_MINOR_VERSION 0x0002
 
 #define MINIMUM_READYTORUN_MAJOR_VERSION 0x006
 

src/coreclr/jit/ClrJit.PAL.exports

@@ -1,4 +1,5 @@
 getJit
 jitStartup
 getLikelyClasses
+getLikelyMethods
 jitBuildString

src/coreclr/jit/ClrJit.exports

@@ -5,4 +5,5 @@ EXPORTS
     getJit
     jitStartup
     getLikelyClasses
+    getLikelyMethods
     jitBuildString