Skip to content

Optimize linear sums #904

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
kripken merged 1 commit into from
Feb 17, 2017
Merged

Optimize linear sums #904

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
110 changes: 110 additions & 0 deletions src/passes/OptimizeInstructions.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -357,6 +357,8 @@ struct OptimizeInstructions : public WalkerPass<PostWalker<OptimizeInstructions,
}
}
// note that both left and right may be consts, but then we let precompute compute the constant result
} else if (binary->op == AddInt32 || binary->op == SubInt32) {
return optimizeAddedConstants(binary);
} else if (binary->op == AndInt32) {
if (auto* right = binary->right->dynCast<Const>()) {
if (right->type == i32) {
Expand Down Expand Up @@ -518,6 +520,114 @@ struct OptimizeInstructions : public WalkerPass<PostWalker<OptimizeInstructions,
return boolean;
}

// find added constants in an expression tree, including multiplied/shifted, and combine them
// note that we ignore division/shift-right, as rounding makes this nonlinear, so not a valid opt
Expression* optimizeAddedConstants(Binary* binary) {
int32_t constant = 0;
std::vector<Const*> constants;
std::function<void (Expression*, int)> seek = [&](Expression* curr, int mul) {
if (auto* c = curr->dynCast<Const>()) {
auto value = c->value.geti32();
if (value != 0) {
constant += value * mul;
constants.push_back(c);
}
} else if (auto* binary = curr->dynCast<Binary>()) {
if (binary->op == AddInt32) {
seek(binary->left, mul);
seek(binary->right, mul);
return;
} else if (binary->op == SubInt32) {
// if the left is a zero, ignore it, it's how we negate ints
auto* left = binary->left->dynCast<Const>();
if (!left || left->value.geti32() != 0) {
seek(binary->left, mul);
}
seek(binary->right, -mul);
return;
} else if (binary->op == ShlInt32) {
if (auto* c = binary->right->dynCast<Const>()) {
seek(binary->left, mul * Pow2(c->value.geti32()));
return;
}
} else if (binary->op == MulInt32) {
if (auto* c = binary->left->dynCast<Const>()) {
seek(binary->right, mul * c->value.geti32());
return;
} else if (auto* c = binary->right->dynCast<Const>()) {
seek(binary->left, mul * c->value.geti32());
return;
}
}
}
};
// find all factors
seek(binary, 1);
if (constants.size() <= 1) return nullptr; // nothing to do
// wipe out all constants, we'll replace with a single added one
for (auto* c : constants) {
c->value = Literal(int32_t(0));
}
// remove added/subbed zeros
struct ZeroRemover : public PostWalker<ZeroRemover, Visitor<ZeroRemover>> {
// TODO: we could save the binarys and costs we drop, and reuse them later

PassOptions& passOptions;

ZeroRemover(PassOptions& passOptions) : passOptions(passOptions) {}

void visitBinary(Binary* curr) {
auto* left = curr->left->dynCast<Const>();
auto* right = curr->right->dynCast<Const>();
if (curr->op == AddInt32) {
if (left && left->value.geti32() == 0) {
replaceCurrent(curr->right);
return;
}
if (right && right->value.geti32() == 0) {
replaceCurrent(curr->left);
return;
}
} else if (curr->op == SubInt32) {
// we must leave a left zero, as it is how we negate ints
if (right && right->value.geti32() == 0) {
replaceCurrent(curr->left);
return;
}
} else if (curr->op == ShlInt32) {
// shifting a 0 is a 0, unless the shift has side effects
if (left && left->value.geti32() == 0 && !EffectAnalyzer(passOptions, curr->right).hasSideEffects()) {
replaceCurrent(left);
return;
}
} else if (curr->op == MulInt32) {
// multiplying by zero is a zero, unless the other side has side effects
if (left && left->value.geti32() == 0 && !EffectAnalyzer(passOptions, curr->right).hasSideEffects()) {
replaceCurrent(left);
return;
}
if (right && right->value.geti32() == 0 && !EffectAnalyzer(passOptions, curr->left).hasSideEffects()) {
replaceCurrent(right);
return;
}
}
}
};
Expression* walked = binary;
ZeroRemover(getPassOptions()).walk(walked);
if (constant == 0) return walked; // nothing more to do
if (auto* c = walked->dynCast<Const>()) {
assert(c->value.geti32() == 0);
c->value = Literal(constant);
return c;
}
Builder builder(*getModule());
return builder.makeBinary(AddInt32,
walked,
builder.makeConst(Literal(constant))
);
}

// expensive1 | expensive2 can be turned into expensive1 ? 1 : expensive2, and
// expensive | cheap can be turned into cheap ? 1 : expensive,
// so that we can avoid one expensive computation, if it has no side effects.
Expand Down
2 changes: 1 addition & 1 deletion src/support/bits.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,13 +114,13 @@ uint32_t Log2(uint32_t v) {

uint32_t Pow2(uint32_t v) {
switch (v) {
default: WASM_UNREACHABLE();
case 0: return 1;
case 1: return 2;
case 2: return 4;
case 3: return 8;
case 4: return 16;
case 5: return 32;
default: return 1 << v;
}
}

Expand Down
9 changes: 3 additions & 6 deletions test/emcc_hello_world.fromasm
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5206,15 +5206,12 @@
(i32.add
(i32.add
(get_local $8)
(i32.const 4)
)
(i32.shl
(i32.add
(i32.shl
(get_local $13)
(i32.const -1024)
(i32.const 2)
)
(i32.const 2)
)
(i32.const -4092)
)
)
)
Expand Down
9 changes: 3 additions & 6 deletions test/emcc_hello_world.fromasm.imprecise
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5143,18 +5143,15 @@
(i32.add
(i32.add
(get_local $8)
(i32.const 4)
)
(i32.shl
(i32.add
(i32.shl
(i32.div_s
(get_local $13)
(i32.const 9)
)
(i32.const -1024)
(i32.const 2)
)
(i32.const 2)
)
(i32.const -4092)
)
)
)
Expand Down
134 changes: 104 additions & 30 deletions test/passes/optimize-instructions.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -417,18 +417,12 @@
)
)
(func $load-off-2 (type $3) (param $0 i32) (result i32)
(i32.store offset=2
(i32.add
(i32.const 1)
(i32.const 3)
)
(i32.store
(i32.const 6)
(get_local $0)
)
(i32.store offset=2
(i32.add
(i32.const 3)
(i32.const 1)
)
(i32.store
(i32.const 6)
(get_local $0)
)
(i32.store offset=2
Expand Down Expand Up @@ -459,18 +453,12 @@
)
(get_local $0)
)
(i32.store offset=2
(i32.add
(i32.const -15)
(i32.const 17)
)
(i32.store
(i32.const 4)
(get_local $0)
)
(i32.store offset=2
(i32.add
(i32.const -21)
(i32.const 19)
)
(i32.store
(i32.const 0)
(get_local $0)
)
(i32.store
Expand All @@ -482,19 +470,13 @@
(get_local $0)
)
(drop
(i32.load offset=2
(i32.add
(i32.const 2)
(i32.const 4)
)
(i32.load
(i32.const 8)
)
)
(drop
(i32.load offset=2
(i32.add
(i32.const 4)
(i32.const 2)
)
(i32.load
(i32.const 8)
)
)
(drop
Expand Down Expand Up @@ -944,4 +926,96 @@
)
)
)
(func $linear-sums (type $4) (param $0 i32) (param $1 i32)
(drop
(i32.add
(get_local $1)
(i32.shl
(get_local $0)
(i32.const 4)
)
)
)
(drop
(i32.add
(i32.add
(get_local $1)
(i32.shl
(get_local $0)
(i32.const 3)
)
)
(i32.const 12)
)
)
(drop
(i32.const 4)
)
(drop
(i32.const 18)
)
(drop
(i32.const 6)
)
(drop
(i32.const -4)
)
(drop
(i32.const 2)
)
(drop
(i32.const 1)
)
(drop
(i32.const 26)
)
(drop
(i32.const -20)
)
(drop
(i32.const 22)
)
(drop
(i32.add
(i32.shl
(i32.const 1)
(get_local $0)
)
(i32.const 14)
)
)
(drop
(i32.add
(i32.shl
(get_local $1)
(i32.const 3)
)
(i32.const -66)
)
)
(drop
(i32.const 44)
)
(drop
(i32.add
(i32.mul
(get_local $0)
(i32.const 10)
)
(i32.const 14)
)
)
(drop
(i32.add
(i32.mul
(get_local $0)
(i32.const 2)
)
(i32.const 34)
)
)
(drop
(get_local $0)
)
)
)
Loading