Skip to content

control flow graph to structured control flow #75

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

control flow graph to structured control flow #75

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
5 changes: 5 additions & 0 deletions language/evm/move-to-yul/src/experiments.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,4 +24,9 @@ impl Experiment {
/// during tests this is off.
/// Retention: permanent
pub const CAPTURE_SOURCE_INFO: &'static str = "capture-source-info";

/// Transform control flow graph to structured control flow.
/// This is off by default for now, we might want to make it default
/// if the performance improvement is significant.
pub const APPLY_CFG_TO_SCF: &'static str = "apply-cfg-to-scf";
}
217 changes: 156 additions & 61 deletions language/evm/move-to-yul/src/functions.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@ use move_stackless_bytecode::{
function_target_pipeline::FunctionVariant,
stackless_bytecode::{Bytecode, Constant, Label, Operation},
stackless_control_flow_graph::{BlockContent, BlockId, StacklessControlFlowGraph},
stackless_structured_control_flow::{compute_label_map, generate_scg_vec, StacklessSCG, StacklessSCGBlockKind},
};
use sha3::{Digest, Keccak256};
use std::collections::{btree_map::Entry, BTreeMap};
Expand Down Expand Up @@ -107,48 +108,153 @@ impl<'a> FunctionGenerator<'a> {
// Compute control flow graph, entry block, and label map
let code = target.data.code.as_slice();
let cfg = StacklessControlFlowGraph::new_forward(code);
let entry_bb = Self::get_actual_entry_block(&cfg);
let label_map = Self::compute_label_map(&cfg, code);

// Emit state machine to represent control flow.
// TODO: Eliminate the need for this, see also
// https://medium.com/leaningtech/solving-the-structured-control-flow-problem-once-and-for-all-5123117b1ee2
if cfg.successors(entry_bb).iter().all(|b| cfg.is_dummmy(*b)) {
// In this trivial case, we have only one block and can omit the state machine
if let BlockContent::Basic { lower, upper } = cfg.content(entry_bb) {
for offs in *lower..*upper + 1 {
self.bytecode(ctx, fun_id, target, &label_map, &code[offs as usize], false);
}
} else {
panic!("effective entry block is not basic")
if !ctx.options.apply_cfg_to_scf() {
self.emit_cfg(&cfg, code, ctx, fun_id, target);
} else {
self.emit_scf_from_cfg(&cfg, code, ctx, fun_id, target);
}
});
emitln!(ctx.writer)
}

fn emit_cfg(
&mut self,
cfg: &StacklessControlFlowGraph,
code: &[Bytecode],
ctx: &Context,
fun_id: &QualifiedInstId<FunId>,
target: &FunctionTarget,
) {
let entry_bb = Self::get_actual_entry_block(cfg);
let label_map = compute_label_map(cfg, code);
// Emit state machine to represent control flow.
// TODO: Eliminate the need for this, see also
// https://medium.com/leaningtech/solving-the-structured-control-flow-problem-once-and-for-all-5123117b1ee2
if cfg.successors(entry_bb).iter().all(|b| cfg.is_dummmy(*b)) {
// In this trivial case, we have only one block and can omit the state machine
if let BlockContent::Basic { lower, upper } = cfg.content(entry_bb) {
for offs in *lower..*upper + 1 {
self.bytecode(
ctx,
fun_id,
target,
&label_map,
&code[offs as usize],
false,
false,
);
}
} else {
emitln!(ctx.writer, "let $block := {}", entry_bb);
emit!(ctx.writer, "for {} true {} ");
panic!("effective entry block is not basic")
}
} else {
emitln!(ctx.writer, "let $block := {}", entry_bb);
emit!(ctx.writer, "for {} true {} ");
ctx.emit_block(|| {
emitln!(ctx.writer, "switch $block");
for blk_id in &cfg.blocks() {
if let BlockContent::Basic { lower, upper } = cfg.content(*blk_id) {
// Emit code for this basic block.
emit!(ctx.writer, "case {} ", blk_id);
ctx.emit_block(|| {
for offs in *lower..*upper + 1 {
self.bytecode(
ctx,
fun_id,
target,
&label_map,
&code[offs as usize],
true,
false,
);
}
})
}
}
})
}
}

fn emit_scf_from_cfg(
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

As we discussed offline, the construction of the SCG really doesn't belong in this module. It should fully go into the module which defines this data type.

&mut self,
cfg: &StacklessControlFlowGraph,
code: &[Bytecode],
ctx: &Context,
fun_id: &QualifiedInstId<FunId>,
target: &FunctionTarget,
) {
let scg_vec = generate_scg_vec(cfg, code);
for scg in scg_vec {
self.emit_scg(&scg, &cfg, code, ctx, fun_id, target);
}
}

pub fn emit_scg(
&mut self,
scg: &StacklessSCG,
cfg: &StacklessControlFlowGraph,
code: &[Bytecode],
ctx: &Context,
fun_id: &QualifiedInstId<FunId>,
target: &FunctionTarget,
) {
let label_map = compute_label_map(cfg, code);
match scg {
StacklessSCG::BasicBlock {
start_offset,
end_offset,
kind,
} => {
ctx.emit_block(|| {
emitln!(ctx.writer, "switch $block");
for blk_id in &cfg.blocks() {
if let BlockContent::Basic { lower, upper } = cfg.content(*blk_id) {
// Emit code for this basic block.
emit!(ctx.writer, "case {} ", blk_id);
ctx.emit_block(|| {
for offs in *lower..*upper + 1 {
self.bytecode(
ctx,
fun_id,
target,
&label_map,
&code[offs as usize],
true,
);
}
})
for offs in *start_offset..*end_offset + 1 {
self.bytecode(
ctx,
fun_id,
target,
&label_map,
&code[offs as usize],
false,
true,
);
}
match kind {
StacklessSCGBlockKind::Break => {
emitln!(ctx.writer, "break");
}
StacklessSCGBlockKind::Continue => {
emitln!(ctx.writer, "continue");
}
_ => {}
}
})
});
}
});
emitln!(ctx.writer)
StacklessSCG::IfBlock {
cond,
if_true,
if_false,
} => {
emitln!(ctx.writer, "switch $t{} ", cond);
emit!(ctx.writer, "case 0 ");
self.emit_scg(if_false, cfg, code, ctx, fun_id, target);
emit!(ctx.writer, "default ");
self.emit_scg(if_true, cfg, code, ctx, fun_id, target);
}
// TODO: need to emit codes of loops based on StacklessSCG::LoopBlock
// based on new bytecodes of Break and Continue
StacklessSCG::LoopBlock {
loop_header,
loop_body,
} => {
self.emit_scg(&*loop_header, &cfg, code, ctx, fun_id, target);
emit!(ctx.writer, "for {} true {} ");
ctx.emit_block(|| {
for loop_body_scg in loop_body {
self.emit_scg(loop_body_scg, &cfg, code, ctx, fun_id, target);
}
});
}
}
}

/// Compute the locals in the given function which are borrowed from and which are not
Expand Down Expand Up @@ -179,22 +285,6 @@ impl<'a> FunctionGenerator<'a> {
}
entry_bb
}

/// Compute a map from labels to block ids which those labels start.
fn compute_label_map(
cfg: &StacklessControlFlowGraph,
code: &[Bytecode],
) -> BTreeMap<Label, BlockId> {
let mut map = BTreeMap::new();
for id in cfg.blocks() {
if let BlockContent::Basic { lower, .. } = cfg.content(id) {
if let Bytecode::Label(_, l) = &code[*lower as usize] {
map.insert(*l, id);
}
}
}
map
}
}

// ================================================================================================
Expand All @@ -210,6 +300,7 @@ impl<'a> FunctionGenerator<'a> {
label_map: &BTreeMap<Label, BlockId>,
bc: &Bytecode,
has_flow: bool,
skip_block: bool,
) {
use Bytecode::*;
emitln!(
Expand Down Expand Up @@ -283,19 +374,23 @@ impl<'a> FunctionGenerator<'a> {
match bc {
Jump(_, l) => {
print_loc();
emitln!(ctx.writer, "$block := {}", get_block(l))
if !skip_block {
emitln!(ctx.writer, "$block := {}", get_block(l))
}
}
Branch(_, if_t, if_f, cond) => {
print_loc();
emitln!(
ctx.writer,
"switch {}\n\
case 0 {{ $block := {} }}\n\
default {{ $block := {} }}",
local(cond),
get_block(if_f),
get_block(if_t),
)
if !skip_block {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Since a branch is always at the end of a basic block, your algorithm can simply construct SCG::BasicBlock(start..end) which does not contain the branch anymore (let it 'end" just before the branch).

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

to make sure that I followed, "blocks" were assigned on "Branch" and "Jump", so to do it right, I want to only skip it on "Jump"? Thanks!

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Each block either ends with a branch, jump, or return. Those instructions cannot appear anywhere else than at the end of a block.

emitln!(
ctx.writer,
"switch {}\n\
case 0 {{ $block := {} }}\n\
default {{ $block := {} }}",
local(cond),
get_block(if_f),
get_block(if_t),
)
}
}
Assign(_, dest, src, _) => {
print_loc();
Expand Down
5 changes: 5 additions & 0 deletions language/evm/move-to-yul/src/options.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,4 +62,9 @@ impl Options {
pub fn generate_source_info(&self) -> bool {
!self.testing || self.experiment_on(Experiment::CAPTURE_SOURCE_INFO)
}

/// Returns true if control flow graph to structured control flow is applied.
pub fn apply_cfg_to_scf(&self) -> bool {
self.experiment_on(Experiment::APPLY_CFG_TO_SCF)
}
}
Loading