Skip to content

A lightweight WebAssembly runtime that is fast, secure, and standards-compliant

24.6 KB
Raw Blame
//! Computation of basic block order in emitted code.
//!
//! This module handles the translation from CLIF BBs to VCode BBs.
//!
//! The basic idea is that we compute a sequence of "lowered blocks" that
//! correspond to one or more blocks in the graph: (CLIF CFG) `union` (implicit
//! block on *every* edge). Conceptually, the lowering pipeline wants to insert
//! moves for phi-nodes on every block-to-block transfer; these blocks always
//! conceptually exist, but may be merged with an "original" CLIF block (and
//! hence not actually exist; this is equivalent to inserting the blocks only on
//! critical edges).
//!
//! In other words, starting from a CFG like this (where each "CLIF block" and
//! "(edge N->M)" is a separate basic block):
//!
//! ```plain
//!
//!              CLIF block 0
//!               /           \
//!       (edge 0->1)         (edge 0->2)
//!              |                |
//!       CLIF block 1         CLIF block 2
//!              \                /
//!           (edge 1->3)   (edge 2->3)
//!                   \      /
//!                 CLIF block 3
//! ```
//!
//! We can produce a CFG of lowered blocks like so:
//!
//! ```plain
//!            +--------------+
//!            | CLIF block 0 |
//!            +--------------+
//!               /           \
//!     +--------------+     +--------------+
//!     | (edge 0->1)  |     | (edge 0->2)  |
//!     | CLIF block 1 |     | CLIF block 2 |
//!     | (edge 1->3)  |     | (edge 2->3)  |
//!     +--------------+     +--------------+
//!                \           /
//!                 \         /
//!                +------------+
//!                |CLIF block 3|
//!                +------------+
//! ```
//!
//! Each `LoweredBlock` names just an original CLIF block, or just an edge block.
//!
//! To compute this lowering, we do a DFS over the CLIF-plus-edge-block graph
//! (never actually materialized, just defined by a "successors" function), and
//! compute the reverse postorder.
//!
//! This algorithm isn't perfect w.r.t. generated code quality: we don't, for
//! example, consider any information about whether edge blocks will actually
//! have content, because this computation happens as part of lowering *before*
//! regalloc, and regalloc may or may not insert moves/spills/reloads on any
//! particular edge. But it works relatively well and is conceptually simple.
//! Furthermore, the [MachBuffer] machine-code sink performs final peephole-like
//! branch editing that in practice elides empty blocks and simplifies some of
//! the other redundancies that this scheme produces.

use crate::dominator_tree::DominatorTree;
use crate::entity::{EntityRef, SecondaryMap};
use crate::inst_predicates::visit_block_succs;
use crate::ir::{Block, Function, Inst, Opcode};
use crate::loop_analysis::{Loop, LoopAnalysis};
use crate::{FxHashMap, FxHashSet};
use crate::{machinst::*, trace};

/// Mapping from CLIF BBs to VCode BBs.
#[derive(Debug)]
pub struct BlockLoweringOrder {
    /// Lowered blocks, in BlockIndex order. Each block is some combination of
    /// (i) a CLIF block, and (ii) inserted crit-edge blocks before or after;
    /// see [LoweredBlock] for details.
    lowered_order: Vec<LoweredBlock>,
    /// BlockIndex values for successors for all lowered blocks, indexing `lowered_order`.
    lowered_succ_indices: Vec<BlockIndex>,
    /// Ranges in `lowered_succ_indices` giving the successor lists for each lowered
    /// block. Indexed by lowering-order index (`BlockIndex`).
    lowered_succ_ranges: Vec<(Option<Inst>, core::ops::Range<usize>)>,
    /// BlockIndex for each original Block.
    blockindex_by_block: SecondaryMap<Block, BlockIndex>,
    /// Cold blocks. These blocks are not reordered in the
    /// `lowered_order` above; the lowered order must respect RPO
    /// (uses after defs) in order for lowering to be
    /// correct. Instead, this set is used to provide `is_cold()`,
    /// which is used by VCode emission to sink the blocks at the last
    /// moment (when we actually emit bytes into the MachBuffer).
    cold_blocks: FxHashSet<BlockIndex>,
    /// Lowered blocks that are indirect branch targets.
    indirect_branch_targets: FxHashSet<BlockIndex>,
}

#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
pub enum LoweredBlock {
    /// Block in original CLIF.
    Orig {
        /// Original CLIF block.
        block: Block,
    },

    /// Critical edge between two CLIF blocks.
    CriticalEdge {
        /// The predecessor block.
        pred: Block,

        /// The successor block.
        succ: Block,

        /// The index of this branch in the successor edges from `pred`, following the same
        /// indexing order as `inst_predicates::visit_block_succs`. This is used to distinguish
        /// multiple edges between the same CLIF blocks.
        succ_idx: u32,
    },
}

impl LoweredBlock {
    /// Unwrap an `Orig` block.
    pub fn orig_block(&self) -> Option<Block> {
        match self {
            &LoweredBlock::Orig { block } => Some(block),
            &LoweredBlock::CriticalEdge { .. } => None,
        }
    }

    /// The associated in-edge predecessor, if this is a critical edge.
    #[cfg(test)]
    pub fn in_edge(&self) -> Option<Block> {
        match self {
            &LoweredBlock::CriticalEdge { pred, .. } => Some(pred),
            &LoweredBlock::Orig { .. } => None,
        }
    }

    /// The associated out-edge successor, if this is a critical edge.
    #[cfg(test)]
    pub fn out_edge(&self) -> Option<Block> {
        match self {
            &LoweredBlock::CriticalEdge { succ, .. } => Some(succ),
            &LoweredBlock::Orig { .. } => None,
        }
    }
}

impl BlockLoweringOrder {
    /// Compute and return a lowered block order for `f`.
    pub fn new(
        f: &Function,
        domtree: &DominatorTree,
        loop_analysis: &LoopAnalysis,
        ctrl_plane: &mut ControlPlane,
    ) -> BlockLoweringOrder {
        trace!("BlockLoweringOrder: function body {:?}", f);

        // Step 1: compute the in-edge and out-edge count of every block.
        let mut block_in_count = SecondaryMap::with_default(0);
        let mut block_out_count = SecondaryMap::with_default(0);

        // Block successors are stored as `LoweredBlocks` to simplify the construction of
        // `lowered_succs` in the final result. Initially, all entries are `Orig` values, and are
        // updated to be `CriticalEdge` when those cases are identified in step 2 below.
        let mut block_succs: SmallVec<[LoweredBlock; 128]> = SmallVec::new();
        let mut block_succ_range = SecondaryMap::with_default(0..0);

        let mut indirect_branch_target_clif_blocks = FxHashSet::default();

        for block in f.layout.blocks() {
            let start = block_succs.len();
            visit_block_succs(f, block, |_, succ, is_indirect| {
                block_out_count[block] += 1;
                block_in_count[succ] += 1;
                block_succs.push(LoweredBlock::Orig { block: succ });

                if is_indirect {
                    indirect_branch_target_clif_blocks.insert(succ);
                }
            });

            // Ensure that blocks terminated by br_table instructions
            // with an empty jump table are still treated like
            // conditional blocks from the point of view of critical
            // edge splitting. Also do the same for TryCall and
            // TryCallIndirect: we cannot have edge moves before the
            // branch, even if they have empty handler tables and thus
            // would otherwise have only one successor.
            if let Some(inst) = f.layout.last_inst(block) {
                match f.dfg.insts[inst].opcode() {
                    Opcode::BrTable | Opcode::TryCall | Opcode::TryCallIndirect => {
                        block_out_count[block] = block_out_count[block].max(2);
                    }
                    _ => {}
                }
            }

            let end = block_succs.len();
            block_succ_range[block] = start..end;
        }

        // Step 2: walk the postorder from the domtree in reverse to produce our desired node
        // lowering order, identifying critical edges to split along the way.

        let mut lowered_order = Vec::new();
        for &block in domtree.cfg_rpo() {
            lowered_order.push(LoweredBlock::Orig { block });

            if block_out_count[block] > 1 {
                let range = block_succ_range[block].clone();

                // If chaos-mode is enabled in the control plane, iterate over
                // the successors in an arbitrary order, which should have no
                // impact on correctness. The order of the blocks is generally
                // relevant: Uses must be seen before defs for dead-code
                // elimination.
                let succs = ctrl_plane.shuffled(block_succs[range].iter_mut().enumerate());

                for (succ_ix, lb) in succs {
                    let succ = lb.orig_block().unwrap();
                    if block_in_count[succ] > 1 {
                        // Mutate the successor to be a critical edge, as `block` has multiple
                        // edges leaving it, and `succ` has multiple edges entering it.
                        *lb = LoweredBlock::CriticalEdge {
                            pred: block,
                            succ,
                            succ_idx: succ_ix as u32,
                        };
                        lowered_order.push(*lb);
                    }
                }
            }
        }

        // Step 3: group blocks belonging to natural loops together. This
        // significantly helps register allocator that perform live range
        // splitting only using the linear instruction order.
        Self::group_loops(&mut lowered_order, loop_analysis);

        // Step 4: create a mapping between blocks and lowered blocks.
        let mut blockindex_by_block = SecondaryMap::with_default(BlockIndex::invalid());
        for (i, lb) in lowered_order.iter().enumerate() {
            if let LoweredBlock::Orig { block } = *lb {
                blockindex_by_block[block] = BlockIndex::new(i);
            }
        }
        let lb_to_bindex = FxHashMap::from_iter(
            lowered_order
                .iter()
                .enumerate()
                .map(|(i, &lb)| (lb, BlockIndex::new(i))),
        );

        // Step 5: build the successor tables given the lowering order. We can't perform this step
        // during the creation of `lowering_order`, as we need `lb_to_bindex` to be fully populated
        // first.
        let mut lowered_succ_indices = Vec::new();
        let mut cold_blocks = FxHashSet::default();
        let mut indirect_branch_targets = FxHashSet::default();
        let lowered_succ_ranges =
            Vec::from_iter(lowered_order.iter().enumerate().map(|(ix, lb)| {
                let bindex = BlockIndex::new(ix);
                let start = lowered_succ_indices.len();
                let opt_inst = match lb {
                    // Block successors are pulled directly over, as they'll have been mutated when
                    // determining the block order already.
                    &LoweredBlock::Orig { block } => {
                        let range = block_succ_range[block].clone();
                        lowered_succ_indices
                            .extend(block_succs[range].iter().map(|lb| lb_to_bindex[lb]));

                        if f.is_effectively_cold(block) {
                            cold_blocks.insert(bindex);
                        }

                        if indirect_branch_target_clif_blocks.contains(&block) {
                            indirect_branch_targets.insert(bindex);
                        }

                        let last = f.layout.last_inst(block).unwrap();
                        let opcode = f.dfg.insts[last].opcode();

                        assert!(opcode.is_terminator());

                        opcode.is_branch().then_some(last)
                    }

                    // Critical edges won't have successor information in block_succ_range, but
                    // they only have a single known successor to record anyway.
                    &LoweredBlock::CriticalEdge { succ, .. } => {
                        let succ_index = lb_to_bindex[&LoweredBlock::Orig { block: succ }];
                        lowered_succ_indices.push(succ_index);

                        // Edges inherit indirect branch and cold block metadata from their
                        // successor.

                        if f.is_effectively_cold(succ) {
                            cold_blocks.insert(bindex);
                        }

                        if indirect_branch_target_clif_blocks.contains(&succ) {
                            indirect_branch_targets.insert(bindex);
                        }

                        None
                    }
                };
                let end = lowered_succ_indices.len();
                (opt_inst, start..end)
            }));

        let result = BlockLoweringOrder {
            lowered_order,
            lowered_succ_indices,
            lowered_succ_ranges,
            blockindex_by_block,
            cold_blocks,
            indirect_branch_targets,
        };

        trace!("BlockLoweringOrder: {:#?}", result);
        result
    }

    /// Group basic blocks from natural loops contiguously, moving any exit
    /// blocks after the main loop body. Since natural loops are dominated by
    /// their loop header, this maintains the dominance order required by
    /// lowering, which visits blocks in reverse order to see uses before defs.
    fn group_loops(order: &mut Vec<LoweredBlock>, loops: &LoopAnalysis) {
        if !loops.is_valid() || loops.loops().len() == 0 {
            return;
        }

        #[derive(Clone, Copy)]
        enum Item {
            Block(LoweredBlock),
            Loop(usize),
        }

        // Get a loop index for each loop, with index 0 representing blocks
        // outside any loop.
        let loop_index = |lp: Option<Loop>| match lp {
            Some(lp) => lp.index() + 1,
            None => 0,
        };

        // Collect the set of blocks belonging to each loop, while otherwise
        // preserving the lowered block order.
        let mut loop_blocks = vec![SmallVec::<[Item; 8]>::new(); loops.loops().len() + 1];
        for node in order.drain(..) {
            let innermost_loop = match node {
                LoweredBlock::Orig { block } => {
                    let lp = loops.innermost_loop(block);
                    let loop_idx = loop_index(lp);

                    // The first block of a loop is always the loop header.
                    // Since loop analysis only records natural loops with a
                    // single entry point and not irreducible cycles, this is
                    // guaranteed to dominate all of the other blocks in the
                    // loop. When we encounter it, insert a placeholder for the
                    // loop in the parent loop.
                    if let Some(lp) = lp {
                        if loop_blocks[loop_idx].is_empty() {
                            debug_assert_eq!(loops.loop_header(lp), block);
                            loop_blocks[loop_index(loops.loop_parent(lp))]
                                .push(Item::Loop(loop_idx));
                        }
                    }

                    loop_idx
                }
                LoweredBlock::CriticalEdge {
                    pred,
                    succ,
                    succ_idx: _,
                } => {
                    // An edge belongs to the common loop region of its
                    // endpoints. Exit stubs move outside the loop they leave;
                    // incoming stubs stay outside and backedge stubs stay in.
                    //
                    // A critical edge can never be a loop header.
                    match (loops.innermost_loop(pred), loops.innermost_loop(succ)) {
                        (Some(a), Some(b)) => loop_index(loops.common_ancestor(a, b)),
                        _ => 0,
                    }
                }
            };
            loop_blocks[innermost_loop].push(Item::Block(node));
        }

        // Starting from the root loop, add all blocks in order, recursively
        // expanding inner loops where `Item::Loop` is found.
        let mut stack: SmallVec<[_; 8]> = smallvec![loop_blocks[0].iter()];
        while let Some(iter) = stack.last_mut() {
            match iter.next() {
                Some(&Item::Block(block)) => order.push(block),
                Some(&Item::Loop(child)) => stack.push(loop_blocks[child].iter()),
                None => {
                    stack.pop();
                }
            }
        }
    }

    /// Get the lowered order of blocks.
    pub fn lowered_order(&self) -> &[LoweredBlock] {
        &self.lowered_order[..]
    }

    /// Get the BlockIndex, if any, for a given Block.
    ///
    /// The result will be `None` if the given Block is unreachable
    /// (and thus does not appear in the lowered order).
    pub fn lowered_index_for_block(&self, block: Block) -> Option<BlockIndex> {
        let idx = self.blockindex_by_block[block];
        if idx.is_valid() { Some(idx) } else { None }
    }

    /// Get the successor indices for a lowered block.
    pub fn succ_indices(&self, block: BlockIndex) -> (Option<Inst>, &[BlockIndex]) {
        let (opt_inst, range) = &self.lowered_succ_ranges[block.index()];
        (*opt_inst, &self.lowered_succ_indices[range.clone()])
    }

    /// Determine whether the given lowered-block index is cold.
    pub fn is_cold(&self, block: BlockIndex) -> bool {
        self.cold_blocks.contains(&block)
    }

    /// Determine whether the given lowered block index is an indirect branch
    /// target.
    pub fn is_indirect_branch_target(&self, block: BlockIndex) -> bool {
        self.indirect_branch_targets.contains(&block)
    }
}

#[cfg(test)]
mod test {
    use super::*;
    use crate::cursor::{Cursor, FuncCursor};
    use crate::flowgraph::ControlFlowGraph;
    use crate::ir::UserFuncName;
    use crate::ir::types::*;
    use crate::ir::{AbiParam, InstBuilder, Signature};
    use crate::isa::CallConv;

    fn build_test_func(n_blocks: usize, edges: &[(usize, usize)]) -> BlockLoweringOrder {
        assert!(n_blocks > 0);

        let name = UserFuncName::testcase("test0");
        let mut sig = Signature::new(CallConv::SystemV);
        sig.params.push(AbiParam::new(I32));
        let mut func = Function::with_name_signature(name, sig);
        let blocks = (0..n_blocks)
            .map(|i| {
                let bb = func.dfg.make_block();
                assert!(bb.as_u32() == i as u32);
                bb
            })
            .collect::<Vec<_>>();

        let arg0 = func.dfg.append_block_param(blocks[0], I32);

        let mut pos = FuncCursor::new(&mut func);

        let mut edge = 0;
        for i in 0..n_blocks {
            pos.insert_block(blocks[i]);
            let mut succs = vec![];
            while edge < edges.len() && edges[edge].0 == i {
                succs.push(edges[edge].1);
                edge += 1;
            }
            if succs.len() == 0 {
                pos.ins().return_(&[arg0]);
            } else if succs.len() == 1 {
                pos.ins().jump(blocks[succs[0]], &[]);
            } else if succs.len() == 2 {
                pos.ins()
                    .brif(arg0, blocks[succs[0]], &[], blocks[succs[1]], &[]);
            } else {
                panic!("Too many successors");
            }
        }

        let mut cfg = ControlFlowGraph::new();
        cfg.compute(&func);
        let dom_tree = DominatorTree::with_function(&func, &cfg);
        let mut loops = LoopAnalysis::new();
        loops.compute(&func, &cfg, &dom_tree);

        BlockLoweringOrder::new(&func, &dom_tree, &loops, &mut Default::default())
    }

    #[test]
    fn multiple_latches_and_separate_exits() {
        //             0
        //             |
        //       +---->1<----+
        //       |    / \    |
        //       |   2   3   |
        //       +---|   |---+
        //           |   |
        //           4   5
        //            \ /
        //             6
        //
        // (2 -> 1 and 3 -> 1 are critical backedges and must be split.)
        //
        let order = build_test_func(
            7,
            &[
                (0, 1),
                (1, 2),
                (1, 3),
                (2, 1),
                (2, 4),
                (3, 1),
                (3, 5),
                (4, 6),
                (5, 6),
            ],
        );

        // Both latches and their split backedges stay with the loop; the
        // separate exits follow the complete loop body.
        use LoweredBlock::{CriticalEdge, Orig};
        assert_eq!(
            order.lowered_order,
            [
                Orig {
                    block: Block::new(0)
                },
                Orig {
                    block: Block::new(1)
                },
                Orig {
                    block: Block::new(3)
                },
                CriticalEdge {
                    pred: Block::new(3),
                    succ: Block::new(1),
                    succ_idx: 0,
                },
                Orig {
                    block: Block::new(2)
                },
                CriticalEdge {
                    pred: Block::new(2),
                    succ: Block::new(1),
                    succ_idx: 0,
                },
                Orig {
                    block: Block::new(5)
                },
                Orig {
                    block: Block::new(4)
                },
                Orig {
                    block: Block::new(6)
                },
            ]
        );
    }

    #[test]
    fn test_blockorder_diamond() {
        let order = build_test_func(4, &[(0, 1), (0, 2), (1, 3), (2, 3)]);

        // This test case doesn't need to introduce any critical edges, as all regalloc allocations
        // can sit on either the entry or exit of blocks 1 and 2.
        assert_eq!(order.lowered_order.len(), 4);
    }

    #[test]
    fn test_blockorder_critedge() {
        //            0
        //          /   \
        //         1     2
        //        /  \     \
        //       3    4    |
        //       |\  _|____|
        //       | \/ |
        //       | /\ |
        //       5    6
        //
        // (3 -> 5, and 3 -> 6 are critical edges and must be split)
        //
        let order = build_test_func(
            7,
            &[
                (0, 1),
                (0, 2),
                (1, 3),
                (1, 4),
                (2, 5),
                (3, 5),
                (3, 6),
                (4, 6),
            ],
        );

        assert_eq!(order.lowered_order.len(), 9);
        println!("ordered = {:?}", order.lowered_order);

        // block 0
        assert_eq!(order.lowered_order[0].orig_block().unwrap().as_u32(), 0);
        assert!(order.lowered_order[0].in_edge().is_none());
        assert!(order.lowered_order[0].out_edge().is_none());

        // block 2
        assert_eq!(order.lowered_order[1].orig_block().unwrap().as_u32(), 2);
        assert!(order.lowered_order[1].in_edge().is_none());
        assert!(order.lowered_order[1].out_edge().is_none());

        // block 1
        assert_eq!(order.lowered_order[2].orig_block().unwrap().as_u32(), 1);
        assert!(order.lowered_order[2].in_edge().is_none());
        assert!(order.lowered_order[2].out_edge().is_none());

        // block 4
        assert_eq!(order.lowered_order[3].orig_block().unwrap().as_u32(), 4);
        assert!(order.lowered_order[3].in_edge().is_none());
        assert!(order.lowered_order[3].out_edge().is_none());

        // block 3
        assert_eq!(order.lowered_order[4].orig_block().unwrap().as_u32(), 3);
        assert!(order.lowered_order[4].in_edge().is_none());
        assert!(order.lowered_order[4].out_edge().is_none());

        // critical edge 3 -> 5
        assert!(order.lowered_order[5].orig_block().is_none());
        assert_eq!(order.lowered_order[5].in_edge().unwrap().as_u32(), 3);
        assert_eq!(order.lowered_order[5].out_edge().unwrap().as_u32(), 5);

        // critical edge 3 -> 6
        assert!(order.lowered_order[6].orig_block().is_none());
        assert_eq!(order.lowered_order[6].in_edge().unwrap().as_u32(), 3);
        assert_eq!(order.lowered_order[6].out_edge().unwrap().as_u32(), 6);

        // block 6
        assert_eq!(order.lowered_order[7].orig_block().unwrap().as_u32(), 6);
        assert!(order.lowered_order[7].in_edge().is_none());
        assert!(order.lowered_order[7].out_edge().is_none());

        // block 5
        assert_eq!(order.lowered_order[8].orig_block().unwrap().as_u32(), 5);
        assert!(order.lowered_order[8].in_edge().is_none());
        assert!(order.lowered_order[8].out_edge().is_none());
    }
}