This optimization identifies DIV instructions in a BB that can be profitably bypassed and carried out with a shorter, faster divide.
Synopsis
Declared in <llvm/Transforms/Utils/BypassSlowDivision.h>
bool
bypassSlowDivision(
BasicBlock* BB,
DenseMap<unsigned int, unsigned int> const& BypassWidth,
DomTreeUpdater* DTU = nullptr,
LoopInfo* LI = nullptr,
BranchProbabilityInfo* BPI = nullptr);
Description
This optimization may add basic blocks immediately after BB; for obvious reasons, you shouldn't pass those blocks to bypassSlowDivision.
Return Value
True if any division or remainder in BB was rewritten.
Parameters
Name |
Description |
BB |
Basic block whose divisions may be bypassed. |
BypassWidth |
Map from operand bit width to profitable bypass width. |
DTU |
Optional dominator‐tree updater for CFG changes. |
LI |
Optional loop info to keep in sync with CFG changes. |
BPI |
Optional branch probabilities for inserted bypass branches. |
Created with MrDocs