Replay inline advisor that uses optimization remarks from inlining of previous build to guide current inlining. This is useful for inliner tuning.
Synopsis
Declared in <llvm/Analysis/ReplayInlineAdvisor.h>
class ReplayInlineAdvisor
: public InlineAdvisor
Base Classes
Name |
Description |
Interface for deciding whether to inline a call site or not. |
Member Functions
Name |
Description |
|
Construct a replay advisor from prior inlining remarks. |
Whether replay remarks were successfully loaded from the replay file. |
|
Get an InlineAdvice recommending whether to inline call site |
|
|
Compute advice for call site |
NOTE pass name is annotated only when inline advisor constructor provides InlineContext. |
|
|
Update advisor state when the Inliner pass is entered. |
|
Prepare advisor state when the Inliner pass is exited. |
|
Print advisor state to |
Protected Enums
Name |
Description |
Kind of mandatory inlining constraint on a call site. |
Protected Member Functions
Name |
Description |
|
Compute advice for call site |
Return the OptimizationRemarkEmitter for the caller of |
|
|
Compute mandatory (always/never) advice for call site |
Protected Static Member Functions
Name |
Description |
Classify the mandatory inlining kind of call site |
Protected Data Members
Name |
Description |
Annotated inline pass name for remarks, when |
|
Function analysis manager providing analyses for advice. |
|
Optional pipeline context used when constructing this advisor. |
|
Statistics for functions imported and then inlined. |
|
Module whose call sites are advised. |
Created with MrDocs