1 =========================== 2 LLVM Branch Weight Metadata 3 =========================== 4 5 .. contents:: 6 :local: 7 8 Introduction 9 ============ 10 11 Branch Weight Metadata represents branch weights as its likeliness to be taken 12 (see :doc:`BlockFrequencyTerminology`). Metadata is assigned to the 13 ``TerminatorInst`` as a ``MDNode`` of the ``MD_prof`` kind. The first operator 14 is always a ``MDString`` node with the string "branch_weights". Number of 15 operators depends on the terminator type. 16 17 Branch weights might be fetch from the profiling file, or generated based on 18 `__builtin_expect`_ instruction. 19 20 All weights are represented as an unsigned 32-bit values, where higher value 21 indicates greater chance to be taken. 22 23 Supported Instructions 24 ====================== 25 26 ``BranchInst`` 27 ^^^^^^^^^^^^^^ 28 29 Metadata is only assigned to the conditional branches. There are two extra 30 operands for the true and the false branch. 31 32 .. code-block:: none 33 34 !0 = metadata !{ 35 metadata !"branch_weights", 36 i32 <TRUE_BRANCH_WEIGHT>, 37 i32 <FALSE_BRANCH_WEIGHT> 38 } 39 40 ``SwitchInst`` 41 ^^^^^^^^^^^^^^ 42 43 Branch weights are assigned to every case (including the ``default`` case which 44 is always case #0). 45 46 .. code-block:: none 47 48 !0 = metadata !{ 49 metadata !"branch_weights", 50 i32 <DEFAULT_BRANCH_WEIGHT> 51 [ , i32 <CASE_BRANCH_WEIGHT> ... ] 52 } 53 54 ``IndirectBrInst`` 55 ^^^^^^^^^^^^^^^^^^ 56 57 Branch weights are assigned to every destination. 58 59 .. code-block:: none 60 61 !0 = metadata !{ 62 metadata !"branch_weights", 63 i32 <LABEL_BRANCH_WEIGHT> 64 [ , i32 <LABEL_BRANCH_WEIGHT> ... ] 65 } 66 67 ``CallInst`` 68 ^^^^^^^^^^^^^^^^^^ 69 70 Calls may have branch weight metadata, containing the execution count of 71 the call. It is currently used in SamplePGO mode only, to augment the 72 block and entry counts which may not be accurate with sampling. 73 74 .. code-block:: none 75 76 !0 = metadata !{ 77 metadata !"branch_weights", 78 i32 <CALL_BRANCH_WEIGHT> 79 } 80 81 Other 82 ^^^^^ 83 84 Other terminator instructions are not allowed to contain Branch Weight Metadata. 85 86 .. _\__builtin_expect: 87 88 Built-in ``expect`` Instructions 89 ================================ 90 91 ``__builtin_expect(long exp, long c)`` instruction provides branch prediction 92 information. The return value is the value of ``exp``. 93 94 It is especially useful in conditional statements. Currently Clang supports two 95 conditional statements: 96 97 ``if`` statement 98 ^^^^^^^^^^^^^^^^ 99 100 The ``exp`` parameter is the condition. The ``c`` parameter is the expected 101 comparison value. If it is equal to 1 (true), the condition is likely to be 102 true, in other case condition is likely to be false. For example: 103 104 .. code-block:: c++ 105 106 if (__builtin_expect(x > 0, 1)) { 107 // This block is likely to be taken. 108 } 109 110 ``switch`` statement 111 ^^^^^^^^^^^^^^^^^^^^ 112 113 The ``exp`` parameter is the value. The ``c`` parameter is the expected 114 value. If the expected value doesn't show on the cases list, the ``default`` 115 case is assumed to be likely taken. 116 117 .. code-block:: c++ 118 119 switch (__builtin_expect(x, 5)) { 120 default: break; 121 case 0: // ... 122 case 3: // ... 123 case 5: // This case is likely to be taken. 124 } 125 126 CFG Modifications 127 ================= 128 129 Branch Weight Metatada is not proof against CFG changes. If terminator operands' 130 are changed some action should be taken. In other case some misoptimizations may 131 occur due to incorrect branch prediction information. 132 133 Function Entry Counts 134 ===================== 135 136 To allow comparing different functions during inter-procedural analysis and 137 optimization, ``MD_prof`` nodes can also be assigned to a function definition. 138 The first operand is a string indicating the name of the associated counter. 139 140 Currently, one counter is supported: "function_entry_count". The second operand 141 is a 64-bit counter that indicates the number of times that this function was 142 invoked (in the case of instrumentation-based profiles). In the case of 143 sampling-based profiles, this operand is an approximation of how many times 144 the function was invoked. 145 146 For example, in the code below, the instrumentation for function foo() 147 indicates that it was called 2,590 times at runtime. 148 149 .. code-block:: llvm 150 151 define i32 @foo() !prof !1 { 152 ret i32 0 153 } 154 !1 = !{!"function_entry_count", i64 2590} 155 156 If "function_entry_count" has more than 2 operands, the later operands are 157 the GUID of the functions that needs to be imported by ThinLTO. This is only 158 set by sampling based profile. It is needed because the sampling based profile 159 was collected on a binary that had already imported and inlined these functions, 160 and we need to ensure the IR matches in the ThinLTO backends for profile 161 annotation. The reason why we cannot annotate this on the callsite is that it 162 can only goes down 1 level in the call chain. For the cases where 163 foo_in_a_cc()->bar_in_b_cc()->baz_in_c_cc(), we will need to go down 2 levels 164 in the call chain to import both bar_in_b_cc and baz_in_c_cc. 165
