SSA optimizer enhancements #242
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jserv
requested review from
ChAoSUnItY,
DrXiao,
fennecJ,
nosba0957 and
vacantron
DrXiao
reviewed
Aug 13, 2025
jserv
force-pushed
the
improve-ssa
branch
3 times, most recently
from
a8695c2 to
9ecf518
Compare
jserv
force-pushed
the
improve-ssa
branch
2 times, most recently
from
cdd443a to
98643c5
Compare
|
In practice, would constant sign extension / truncation also be handled by SCCP or other optimization units? If so, maybe it's good to implement them as well. |
- Add comprehensive constant folding for arithmetic, bitwise, logical, and comparison operations - Include zero-division protection for division and modulo operations - Add unary operation constant folding (negate, bit_not, log_not)
- Implement SCCP (Sparse Conditional Constant Propagation) optimization pass - Add constant propagation through assignments - Add constant folding for arithmetic and comparison operations - Add branch folding when conditions are compile-time constants - Include dead code elimination through unreachable branch removal
- Replace magic numbers with configuration constants (PHI_WORKLIST_SIZE, DCE_WORKLIST_SIZE) - Add overflow protection for PHI worklist to prevent buffer overruns - Improve safety checks in dead code elimination
- Add comprehensive CSE support for arithmetic, bitwise, logical, and comparison operations - Support commutative operation detection (add, mul, and, or, xor, eq, neq) - Maintain existing array access pattern optimization (add + read) - Add safety checks to prevent CSE with global variables
- Add unreachable block detection using dominator analysis - Implement conservative dead store elimination for immediate overwrites - Add redundant assignment elimination (x = x patterns) - Enhance DCE worklist algorithm with better bounds checking - Simplify escape analysis to be maximally conservative - Add instruction elimination metrics tracking - Fix phi operand marking in DCE worklist processing
This enhances the SCCP optimizer with support for constant truncation operations, providing multiple optimization strategies to handle compile-time constant cast operations efficiently. The optimization correctly handles patterns like: const %.t977, 300 %.t978 = trunc %.t977, 1 And performs compile-time evaluation (e.g., 300 & 0xFF = 44 for char).
This extends the SCCP optimizer to handle compile-time constant sign extension operations (OP_sign_ext), complementing the existing constant truncation optimization. - 8-bit to 32-bit sign extension (char to int) - 16-bit to 32-bit sign extension (short to int) - Proper sign bit propagation for negative values When sign extension operations have constant operands, they are evaluated at compile-time and replaced with direct constant loads, eliminating runtime overhead. Example transformations: - char c = -1; int i = c; → i directly loads 0xFFFFFFFF (-1) - char c = 127; int i = c; → i directly loads 0x0000007F (127)
DrXiao
reviewed
Aug 17, 2025
Comment on lines
+271
to
+278
| /* Remove the truncation instruction by converting it to | ||
| * NOP-like | ||
| */ | ||
| next_insn->opcode = OP_load_constant; | ||
| next_insn->rd->is_const = true; | ||
| next_insn->rd->init_val = result; | ||
| next_insn->rs1 = NULL; | ||
| next_insn->sz = 0; |
There was a problem hiding this comment.
How about processing it as following?
insn->next = next_insn->next;
if (next_insn->next)
next_insn->next->prev = insn;Instead of converting it to NOP-like instruction, we can directly link the previous and next instructions to skip this unused instruction.
There was a problem hiding this comment.
Instead of converting it to NOP-like instruction, we can directly link the previous and next instructions to skip this unused instruction.
Direct instruction linking was investigated, but it causes bootstrap failures.
This unified the constant folding logic for binary operations (add, sub, mul) and comparison operations (eq, neq, lt, leq, gt, geq) into a single case statement.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This extends the SSA optimization framework with multiple optimization techniques that operate at the basic block level:
Summary by Bito
This pull request enhances the SSA optimization framework by introducing Sparse Conditional Constant Propagation (SCCP), improved constant folding techniques, and Common Subexpression Elimination (CSE). These optimizations operate at the basic block level, significantly improving performance, efficiency, and robustness in handling constant values and reducing redundant computations.