[
https://issues.apache.org/jira/browse/GROOVY-12065?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Daniel Sun updated GROOVY-12065:
--------------------------------
Description:
h2. Overview
This improvement introduces a dedicated, single-pass bytecode compaction layer
via {{PeepholeOptimizingMethodVisitor}}. This adapter wraps the underlying ASM
{{MethodVisitor}} during class generation ({{AsmClassGenerator}}), intercepting
instruction streams within local basic blocks to eliminate redundant
operations, rewrite conditional branches, and narrow constant instructions to
their most optimal forms.
Crucially, by decoupling peephole optimization from structural code generation,
this approach significantly simplifies the constant emission logic inside
{{OperandStack}}. Instead of maintaining verbose, duplicate routing logic for
specialized primitive opcodes (such as {{ICONST_x}}, {{BIPUSH}}, or
{{FCONST_x}}), {{OperandStack}} now delegates constant pushing uniformly via
{{visitLdcInsn}}. The stateful peephole layer then transparently condenses
these instructions into their tightest bytecode representations under the hood.
h2. Key Optimizations Implemented
h3. 1. Redundant Instruction & Dead Code Elimination
* *Discarded Assignment Values:* Eliminates wasteful {{DUP}} -> {{[X]STORE}} ->
{{POP}} patterns typically produced during assignments where the expression
result on the operand stack is unused. The optimizer flattens these directly
into a single {{[X]STORE}}.
* *Redundant Loads:* Detects situations where a local variable or constant is
loaded ({{[X]LOAD}} / {{LDC}}) but immediately discarded via {{POP}}/{{POP2}},
or followed immediately by a void {{RETURN}}. The optimizer safely drops both
operations while preserving interleaved local increments ({{IINC}}).
h3. 2. Conditional Jump & Zero-Comparison Rewriting
* Optimizes integer comparisons against zero by transforming expensive binary
comparison sequences (e.g., {{ILOAD}} -> {{ICONST_0}} -> {{IF_ICMPxx}}) into
compact unary zero-comparison instructions ({{IFxx}}). For instance, an
{{IF_ICMPEQ}} branch against a buffered {{0}} constant is cleanly rewritten
directly to {{IFEQ}}.
h3. 3. Instruction Narrowing & Constant Compaction
* Intercepts standard literal definitions and transparently narrows them down
to the smallest possible specialized opcodes ({{ICONST_M1}} through
{{ICONST_5}}, {{BIPUSH}}, {{SIPUSH}}, {{LCONST_0/1}}, {{FCONST_0/1/2}}, and
{{DCONST_0/1}}).
* Explicitly safeguards signed floating-point zeros ({{-0.0f}} and {{-0.0d}})
from accidental flattening to guarantee strict IEEE 754 runtime compliance.
h3. 4. Big Number Literal Lowering
* Centralizes string-constructed object instantiations for {{BigDecimal}} and
{{BigInteger}} literals by rewriting buffered constants into inline {{NEW}} ->
{{DUP}} -> {{LDC [string]}} -> {{INVOKESPECIAL <init>}} pipelines seamlessly
before stack emission.
h2. Implementation Strategy & Safety Guardrails
The {{PeepholeOptimizingMethodVisitor}} operates on a lightweight, stack-local
sliding window. To fully preserve runtime semantics, debugging capabilities,
and catch-block structures, the lookahead window is automatically flushed to
the delegate visitor when encountering non-local boundaries or frame-altering
instructions, including:
* Control flow jumps and basic block {{Label}} targets.
* Stack map frames ({{visitFrame}}).
* Line numbers and local variable debug maps.
* Method invocations and {{InvokeDynamic}} instructions.
was:
h3. Background
Groovy's bytecode generator relied on ad-hoc, per-type conditional branches
scattered across {{OperandStack}} and {{BytecodeHelper}} to select compact
constant-loading instructions ({{{}ICONST_x{}}}, {{{}LCONST_x{}}},
{{{}FCONST_x{}}}, {{{}DCONST_x{}}}, {{{}BIPUSH{}}}, {{{}SIPUSH{}}}) over the
generic {{LDC}} instruction. This logic was fragile, inconsistent, and
incomplete — for instance, {{BytecodeHelper.pushConstant()}} was missing the
{{ICONST_M1}} case for the integer constant {{{}-1{}}}.
h3. Change
A new {{PeepholeOptimizingMethodVisitor}} — a {{MethodVisitor}} decorator — is
introduced and inserted into the per-method visitor chain in
{{{}AsmClassGenerator{}}}. It intercepts every {{visitLdcInsn}} and
{{visitIntInsn(BIPUSH/SIPUSH, ...)}} call and rewrites each to the shortest
valid equivalent opcode:
||Value||Emitted instruction||
|{{int}} −1|{{ICONST_M1}}|
|{{int}} 0–5|{{ICONST_0}} – {{ICONST_5}}|
|{{int}} in [−128, 127]|{{BIPUSH}}|
|{{int}} in [−32768, 32767]|{{SIPUSH}}|
|{{long}} 0 / 1|{{LCONST_0}} / {{LCONST_1}}|
|{{float}} 0 / 1 / 2|{{FCONST_0}} / {{FCONST_1}} / {{FCONST_2}}|
|{{double}} 0 / 1|{{DCONST_0}} / {{DCONST_1}}|
Signed-zero {{float}} and {{double}} values are handled correctly by comparing
raw bit patterns (via {{Float.floatToRawIntBits}} /
{{{}Double.doubleToRawLongBits{}}}) rather than {{{}=={}}}, preserving the
{{-0.0}} / {{+0.0}} distinction that a straight equality check would collapse.
All per-type branches previously inlined in {{OperandStack}} are replaced with
plain {{visitLdcInsn(value)}} calls; the peephole visitor centralizes the
selection uniformly for every generated method.
{{BytecodeHelper.pushConstant()}} also gains the previously missing
{{ICONST_M1}} branch.
> Implement peephole optimization for bytecode generation
> -------------------------------------------------------
>
> Key: GROOVY-12065
> URL: https://issues.apache.org/jira/browse/GROOVY-12065
> Project: Groovy
> Issue Type: Improvement
> Reporter: Daniel Sun
> Priority: Major
>
> h2. Overview
> This improvement introduces a dedicated, single-pass bytecode compaction
> layer via {{PeepholeOptimizingMethodVisitor}}. This adapter wraps the
> underlying ASM {{MethodVisitor}} during class generation
> ({{AsmClassGenerator}}), intercepting instruction streams within local basic
> blocks to eliminate redundant operations, rewrite conditional branches, and
> narrow constant instructions to their most optimal forms.
> Crucially, by decoupling peephole optimization from structural code
> generation, this approach significantly simplifies the constant emission
> logic inside {{OperandStack}}. Instead of maintaining verbose, duplicate
> routing logic for specialized primitive opcodes (such as {{ICONST_x}},
> {{BIPUSH}}, or {{FCONST_x}}), {{OperandStack}} now delegates constant pushing
> uniformly via {{visitLdcInsn}}. The stateful peephole layer then
> transparently condenses these instructions into their tightest bytecode
> representations under the hood.
> h2. Key Optimizations Implemented
> h3. 1. Redundant Instruction & Dead Code Elimination
> * *Discarded Assignment Values:* Eliminates wasteful {{DUP}} -> {{[X]STORE}}
> -> {{POP}} patterns typically produced during assignments where the
> expression result on the operand stack is unused. The optimizer flattens
> these directly into a single {{[X]STORE}}.
> * *Redundant Loads:* Detects situations where a local variable or constant is
> loaded ({{[X]LOAD}} / {{LDC}}) but immediately discarded via
> {{POP}}/{{POP2}}, or followed immediately by a void {{RETURN}}. The optimizer
> safely drops both operations while preserving interleaved local increments
> ({{IINC}}).
> h3. 2. Conditional Jump & Zero-Comparison Rewriting
> * Optimizes integer comparisons against zero by transforming expensive binary
> comparison sequences (e.g., {{ILOAD}} -> {{ICONST_0}} -> {{IF_ICMPxx}}) into
> compact unary zero-comparison instructions ({{IFxx}}). For instance, an
> {{IF_ICMPEQ}} branch against a buffered {{0}} constant is cleanly rewritten
> directly to {{IFEQ}}.
> h3. 3. Instruction Narrowing & Constant Compaction
> * Intercepts standard literal definitions and transparently narrows them down
> to the smallest possible specialized opcodes ({{ICONST_M1}} through
> {{ICONST_5}}, {{BIPUSH}}, {{SIPUSH}}, {{LCONST_0/1}}, {{FCONST_0/1/2}}, and
> {{DCONST_0/1}}).
> * Explicitly safeguards signed floating-point zeros ({{-0.0f}} and {{-0.0d}})
> from accidental flattening to guarantee strict IEEE 754 runtime compliance.
> h3. 4. Big Number Literal Lowering
> * Centralizes string-constructed object instantiations for {{BigDecimal}} and
> {{BigInteger}} literals by rewriting buffered constants into inline {{NEW}}
> -> {{DUP}} -> {{LDC [string]}} -> {{INVOKESPECIAL <init>}} pipelines
> seamlessly before stack emission.
> h2. Implementation Strategy & Safety Guardrails
> The {{PeepholeOptimizingMethodVisitor}} operates on a lightweight,
> stack-local sliding window. To fully preserve runtime semantics, debugging
> capabilities, and catch-block structures, the lookahead window is
> automatically flushed to the delegate visitor when encountering non-local
> boundaries or frame-altering instructions, including:
> * Control flow jumps and basic block {{Label}} targets.
> * Stack map frames ({{visitFrame}}).
> * Line numbers and local variable debug maps.
> * Method invocations and {{InvokeDynamic}} instructions.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
