mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-12-18 10:31:57 +00:00
8383b539ff
This commit adds the infrastructure for performing bottom-up SLP vectorization (and other optimizations) on parallel computations.
The infrastructure has three potential users:
1. The loop vectorizer needs to be able to vectorize AOS data structures such as (sum += A[i] + A[i+1]).
2. The BB-vectorizer needs this infrastructure for bottom-up SLP vectorization, because bottom-up vectorization is faster to compute.
3. A loop-roller needs to be able to analyze consecutive chains and roll them into a loop, in order to reduce code size. A loop roller does not need to create vector instructions, and this infrastructure separates the chain analysis from the vectorization.
This patch also includes a simple (100 LOC) bottom up SLP vectorizer that uses the infrastructure, and can vectorize this code:
void SAXPY(int *x, int *y, int a, int i) {
x[i] = a * x[i] + y[i];
x[i+1] = a * x[i+1] + y[i+1];
x[i+2] = a * x[i+2] + y[i+2];
x[i+3] = a * x[i+3] + y[i+3];
}
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179117 91177308-0d34-0410-b5e6-96231b3b80d8
141 lines
4.1 KiB
C++
141 lines
4.1 KiB
C++
//===-- Vectorize.h - Vectorization Transformations -------------*- C++ -*-===//
|
|
//
|
|
// The LLVM Compiler Infrastructure
|
|
//
|
|
// This file is distributed under the University of Illinois Open Source
|
|
// License. See LICENSE.TXT for details.
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
//
|
|
// This header file defines prototypes for accessor functions that expose passes
|
|
// in the Vectorize transformations library.
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#ifndef LLVM_TRANSFORMS_VECTORIZE_H
|
|
#define LLVM_TRANSFORMS_VECTORIZE_H
|
|
|
|
namespace llvm {
|
|
class BasicBlock;
|
|
class BasicBlockPass;
|
|
class Pass;
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
/// @brief Vectorize configuration.
|
|
struct VectorizeConfig {
|
|
//===--------------------------------------------------------------------===//
|
|
// Target architecture related parameters
|
|
|
|
/// @brief The size of the native vector registers.
|
|
unsigned VectorBits;
|
|
|
|
/// @brief Vectorize boolean values.
|
|
bool VectorizeBools;
|
|
|
|
/// @brief Vectorize integer values.
|
|
bool VectorizeInts;
|
|
|
|
/// @brief Vectorize floating-point values.
|
|
bool VectorizeFloats;
|
|
|
|
/// @brief Vectorize pointer values.
|
|
bool VectorizePointers;
|
|
|
|
/// @brief Vectorize casting (conversion) operations.
|
|
bool VectorizeCasts;
|
|
|
|
/// @brief Vectorize floating-point math intrinsics.
|
|
bool VectorizeMath;
|
|
|
|
/// @brief Vectorize the fused-multiply-add intrinsic.
|
|
bool VectorizeFMA;
|
|
|
|
/// @brief Vectorize select instructions.
|
|
bool VectorizeSelect;
|
|
|
|
/// @brief Vectorize comparison instructions.
|
|
bool VectorizeCmp;
|
|
|
|
/// @brief Vectorize getelementptr instructions.
|
|
bool VectorizeGEP;
|
|
|
|
/// @brief Vectorize loads and stores.
|
|
bool VectorizeMemOps;
|
|
|
|
/// @brief Only generate aligned loads and stores.
|
|
bool AlignedOnly;
|
|
|
|
//===--------------------------------------------------------------------===//
|
|
// Misc parameters
|
|
|
|
/// @brief The required chain depth for vectorization.
|
|
unsigned ReqChainDepth;
|
|
|
|
/// @brief The maximum search distance for instruction pairs.
|
|
unsigned SearchLimit;
|
|
|
|
/// @brief The maximum number of candidate pairs with which to use a full
|
|
/// cycle check.
|
|
unsigned MaxCandPairsForCycleCheck;
|
|
|
|
/// @brief Replicating one element to a pair breaks the chain.
|
|
bool SplatBreaksChain;
|
|
|
|
/// @brief The maximum number of pairable instructions per group.
|
|
unsigned MaxInsts;
|
|
|
|
/// @brief The maximum number of candidate instruction pairs per group.
|
|
unsigned MaxPairs;
|
|
|
|
/// @brief The maximum number of pairing iterations.
|
|
unsigned MaxIter;
|
|
|
|
/// @brief Don't try to form odd-length vectors.
|
|
bool Pow2LenOnly;
|
|
|
|
/// @brief Don't boost the chain-depth contribution of loads and stores.
|
|
bool NoMemOpBoost;
|
|
|
|
/// @brief Use a fast instruction dependency analysis.
|
|
bool FastDep;
|
|
|
|
/// @brief Initialize the VectorizeConfig from command line options.
|
|
VectorizeConfig();
|
|
};
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
//
|
|
// BBVectorize - A basic-block vectorization pass.
|
|
//
|
|
BasicBlockPass *
|
|
createBBVectorizePass(const VectorizeConfig &C = VectorizeConfig());
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
//
|
|
// LoopVectorize - Create a loop vectorization pass.
|
|
//
|
|
Pass *createLoopVectorizePass();
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
//
|
|
// SLPVectorizer - Create a bottom-up SLP vectorizer pass.
|
|
//
|
|
Pass *createSLPVectorizerPass();
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
/// @brief Vectorize the BasicBlock.
|
|
///
|
|
/// @param BB The BasicBlock to be vectorized
|
|
/// @param P The current running pass, should require AliasAnalysis and
|
|
/// ScalarEvolution. After the vectorization, AliasAnalysis,
|
|
/// ScalarEvolution and CFG are preserved.
|
|
///
|
|
/// @return True if the BB is changed, false otherwise.
|
|
///
|
|
bool vectorizeBasicBlock(Pass *P, BasicBlock &BB,
|
|
const VectorizeConfig &C = VectorizeConfig());
|
|
|
|
} // End llvm namespace
|
|
|
|
#endif
|