AMDGPUInline.cpp File Reference

This is AMDGPU specific replacement of the standard inliner. More...

#include "AMDGPU.h"
#include "llvm/Transforms/IPO.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/CallGraph.h"
#include "llvm/Analysis/InlineCost.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/IR/CallSite.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/Type.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Transforms/IPO/Inliner.h"

Include dependency graph for AMDGPUInline.cpp:

Go to the source code of this file.

Macros
#define	DEBUG_TYPE   "inline"

Functions
	INITIALIZE_PASS_BEGIN (AMDGPUInliner, "amdgpu-inline", "AMDGPU Function Integration/Inlining", false, false) INITIALIZE_PASS_END(AMDGPUInliner
static bool	isWrapperOnlyCall (CallSite CS)

Variables
static cl::opt< int >	ArgAllocaCost ("amdgpu-inline-arg-alloca-cost", cl::Hidden, cl::init(2200), cl::desc("Cost of alloca argument"))
static cl::opt< unsigned >	ArgAllocaCutoff ("amdgpu-inline-arg-alloca-cutoff", cl::Hidden, cl::init(256), cl::desc("Maximum alloca size to use for inline cost"))
amdgpu	inline
amdgpu AMDGPU Function Integration	Inlining
amdgpu AMDGPU Function Integration	false

Detailed Description

This is AMDGPU specific replacement of the standard inliner.

The main purpose is to account for the fact that calls not only expensive on the AMDGPU, but much more expensive if a private memory pointer is passed to a function as an argument. In this situation, we are unable to eliminate private memory in the caller unless inlined and end up with slow and expensive scratch access. Thus, we boost the inline threshold for such functions here.

Definition in file AMDGPUInline.cpp.

Macro Definition Documentation

DEBUG_TYPE

#define DEBUG_TYPE   "inline"

Definition at line 40 of file AMDGPUInline.cpp.

Referenced by isWrapperOnlyCall().

Function Documentation

INITIALIZE_PASS_BEGIN()

INITIALIZE_PASS_BEGIN	(	AMDGPUInliner	,
		"amdgpu-inline"	,
		"AMDGPU Function Integration/Inlining"	,
		false	,
		false
	)

isWrapperOnlyCall()

static bool isWrapperOnlyCall ( CallSite  CS )

static

Definition at line 152 of file AMDGPUInline.cpp.

References llvm::Attribute::AlwaysInline, llvm::TargetTransformInfo::areInlineCompatible(), Callee, llvm::dbgs(), DEBUG_TYPE, llvm::InlineParams::DefaultThreshold, F(), llvm::InlineCost::getAlways(), llvm::Function::getBasicBlockList(), llvm::CallSiteBase< FunTy, BBTy, ValTy, UserTy, UseTy, InstrTy, CallTy, InvokeTy, IterTy >::getCalledFunction(), llvm::CallSiteBase< FunTy, BBTy, ValTy, UserTy, UseTy, InstrTy, CallTy, InvokeTy, IterTy >::getCaller(), llvm::Function::getEntryBlock(), llvm::BasicBlock::getFirstNonPHI(), llvm::getInlineCost(), llvm::Value::getName(), llvm::InlineCost::getNever(), llvm::CallSiteBase< FunTy, BBTy, ValTy, UserTy, UseTy, InstrTy, CallTy, InvokeTy, IterTy >::hasFnAttr(), I, llvm::GlobalValue::isDeclaration(), llvm::isInlineViable(), llvm::CallSiteBase< FunTy, BBTy, ValTy, UserTy, UseTy, InstrTy, CallTy, InvokeTy, IterTy >::isNoInline(), LLVM_DEBUG, llvm::None, and llvm::Function::size().

Variable Documentation

ArgAllocaCost

cl::opt<int> ArgAllocaCost("amdgpu-inline-arg-alloca-cost", cl::Hidden, cl::init(2200), cl::desc("Cost of alloca argument"))

static

Referenced by llvm::createAMDGPUFunctionInliningPass().

ArgAllocaCutoff

cl::opt<unsigned> ArgAllocaCutoff("amdgpu-inline-arg-alloca-cutoff", cl::Hidden, cl::init(256), cl::desc("Maximum alloca size to use for inline cost"))

static

Referenced by llvm::createAMDGPUFunctionInliningPass().

false

amdgpu AMDGPU Function Integration false

Definition at line 89 of file AMDGPUInline.cpp.

inline

amdgpu inline

Definition at line 89 of file AMDGPUInline.cpp.

Inlining

amdgpu AMDGPU Function Integration Inlining

Definition at line 89 of file AMDGPUInline.cpp.