M2compiler-V3/docs/comparison-PIM4-gm2-V3.md

PIM4 vs GNU Modula-2 vs V3 — feature comparison

Date: 2026-10-22 · V3 state: v3-proc-types (suite 255/255, self-hosting fixpoint OK 3,935,235 bytes, corpus 503/612, 0 crashes).

Scope. PIM4 is Wirth's Programming in Modula-2, 4th edition (the language specification, not a compiler). GNU Modula-2 (gm2) is the GCC front end: its PIM and ISO modes plus the extensions common in its testsuite. V3 is this compiler: a PIM/ISO subset plus the gm2 compatibility it deliberately chose.

Legend: ✅ supported / standard · 🔄 partial or caveated · ❌ absent · ⚠️ non-standard extension · n/a not applicable.

The tables describe the common/default behaviour; the exact flags and modes of gm2 can change individual cells. V3's deliberate deviations are listed at the end.

Standards & dialect

Feature	PIM4	GNU Modula-2	V3
PIM base language	✅ (definition)	✅ (PIM 2/3/4 modes)	✅ (PIM/ISO subset)
ISO 10514-1 base	❌ (predates ISO)	✅	🔄 mostly; see missing ISO rows below
ISO 10514-2 generics	❌	✅	✅ (refinement form)
ISO 10514-3 OO	❌	❌	❌ (CLASS is Clarion-form, not ISO OO)
Declaration before use	✅ strict	🔄 multi-pass, accepts forward refs	✅ strict for data; 🔄 forward procedure calls allowed
FORWARD declarations	✅	✅	✅
EXPORT QUALIFIED / UNQUALIFIED	❌ (PIM export list only)	✅	✅
Foreign modules (DEFINITION MODULE FOR "C")	❌	✅	✅

Lexical

Feature	PIM4	GNU Modula-2	V3
Identifiers with _	❌ (classic names)	✅	✅
Numeric bases	✅ octal B, hex H	✅ B/H (and more)	❌ (0x/0X only; 012H/666B rejected)
Nested (* *) comments	✅	✅	✅
'…' and "…" strings	✅	✅	✅
Unicode UCHAR/UString	❌	❌	✅ (extension: U'a', U"…", UCHR, UORD)

Types

Feature	PIM4	GNU Modula-2	V3
INTEGER/CARDINAL/SHORTINT/LONGINT, REAL/LONGREAL, BOOLEAN, CHAR	✅	✅	✅
Subranges	✅	✅	✅ (typed CHAR/enum subranges since 2026-10-22)
Enumerations	✅	✅	✅
Sets (SET OF)	✅ any finite ordinal (impl. limit)	✅	🔄 bool/char/enum/small-int subrange; span ≤ 65536; no SET OF CARDINAL
PACKEDSET	❌ (ISO)	✅	❌
Fixed arrays, multi-dimensional	✅ ARRAY [a], [b] OF	✅	✅ both [a], [b] and [a, b]
Open arrays (1-D)	✅	✅	✅
Multi-dimensional open arrays	❌ (ISO addition)	✅	❌ (V3 descriptor-per-level representation; unbounded* gaps)
Records, variant records	✅	✅	✅ (plus empty arms/ELSE extensions)
Pointers, NIL, opaque types	✅	✅	✅
Procedure types	✅ (PROCEDURE)	✅ (PROCEDURE, PROC)	✅ (PROCEDURE, PROC; comma/unnamed forms)
Complex types (COMPLEX, LONGCOMPLEX)	❌	✅ (ISO)	❌ (COMPLEX64 is a REAL placeholder, no arithmetic)
SYSTEM sized types (WORD32, CARDINAL8, BITSET32, …)	❌	✅	✅ names, but mapped to V3's 32-bit/64-bit model
BYTE granularity	✅ 8-bit	✅ 8-bit	🔄 BYTE = 32-bit CARDINAL; ARRAY OF BYTE is not byte-granular
SHORTREAL (32-bit)	❌	✅	❌ (all reals are 8 bytes)
ADDRESS arithmetic (ADDADR/SUBADR/DIFADR)	❌	✅	❌
Absolute variables VAR [addr]: T	❌	✅ (extension)	❌
Set ordering (<, <=, >, >=)	❌ (=/#/IN only)	✅ (extension)	❌

Statements & procedures

Feature	PIM4	GNU Modula-2	V3
IF/CASE/WHILE/REPEAT/LOOP/EXIT/WITH, RETURN, HALT	✅	✅	✅
FOR over any ordinal (CHAR/enum)	✅	✅	✅ (since 2026-10-22)
GOTO / labels	✅	✅	✅
INC/DEC on enumerations	❌ (integer)	✅	❌ (integer only)
Empty string ""/'' as CHAR	❌/unclear (ISO rule)	✅	✅ (ISO rule, since 2026-10-22)
LENGTH	❌ (ISO adds it)	✅	✅
HIGH	✅	✅	✅ / 🔄 (HIGH of a string CONST still missing)
TSIZE/SIZE/ADR	✅	✅	✅
SHIFT/ROTATE	⚠️ implementation SYSTEM extension	✅	✅
TBITSIZE	❌	✅	❌
ISO exceptions (EXCEPT, RETRY, EXCEPTIONS)	❌	✅	❌
Module finalization (FINALLY)	❌	✅	❌
__FILE__ / __LINE__ / __FUNCTION__	❌	✅	✅ / ✅ / ❌
__BUILTIN__ / __ATTRIBUTE__ pragmas	❌	✅	✅ (parsed, ignored)
Inline ASM	❌	✅	❌

Modules, compilation, libraries

Feature	PIM4	GNU Modula-2	V3
Definition / implementation / program modules	✅	✅	✅
Local modules	✅	✅	✅
Separate compilation	✅	✅ (object files/archives)	🔄 one invocation compiles all units into one image; no incremental objects
Generics	❌	✅	✅
Object orientation	❌	❌	🔄 Clarion-form CLASS (single inheritance, virtual dispatch), not ISO 10514-3
PIM coroutines (PROCESS, NEWPROCESS, TRANSFER, IOTRANSFER)	✅	✅	✅ (IOTRANSFER degrades to a transfer)
ISO COROUTINES module	❌	✅	❌ (PIM interface only)
Standard library	❌ (none defined)	✅ PIM + ISO libraries	🔄 subset: ISO/PIM I/O, strings, storage, SysClock; no ISO exceptions/termination libs
Self-hosting	n/a	✅ (part of GCC)	✅ (byte-identical 3-stage fixpoint)
Backend / target	n/a	GCC (many targets)	QBE → C (amd64 SysV today)

The practical differences

• PIM4 is the clean specification: strict declaration order, no ISO exceptions/complex numbers/PACKEDSET/generics, no standard library, no sized SYSTEM types or address arithmetic.
• gm2 is the superset in practice: PIM and ISO modes, generics, exceptions, complex numbers, PACKEDSET, multi-dimensional open arrays, sized SYSTEM types, address arithmetic, absolute variables, ASM, pragmas, and many extensions — but deliberately permissive (multi-pass resolution accepts forward references).
• V3 is a PIM/ISO compiler with the opposite bias: it rejects gm2 leniencies (declaration order), implements the common ISO/PIM constructs (generics, coroutines, foreign modules, ISO strings/SYSTEM/ procedure types, empty string as CHAR), and documents its chosen gm2 compatibility (__FILE__/__LINE__, scalar→ARRAY OF BYTE, same-size VAR, bare proc values, variant extensions, set CASE labels, libc prototypes). Its hard gaps are the large ISO features (exceptions, complex arithmetic, PACKEDSET, finalization/protection, multi-dimensional open arrays, ISO OO) and its 32-bit word model (BYTE granularity, SHORTREAL, REAL width), plus a 65536-bit set span limit.

V3-specific deviations and limits

• Strict data declaration order (PIM-conforming); forward procedure calls are allowed where strict PIM would require FORWARD.
• 32-bit word model: INTEGER/CARDINAL/WORD/BYTE are 32-bit; LONGINT/LONGCARD are 64-bit; every REAL/SHORTREAL is 8 bytes; COMPLEX* is a placeholder. ARRAY OF BYTE is a compatibility descriptor, not an 8-bit buffer.
• Sets: bases limited to bool/char/enum/small subranges; total span ≤ 65536; no SET OF CARDINAL.
• No runtime range checks on subrange assignment (V3 is already lenient for the integer family).
• Multi-dimensional open arrays are not supported: V3 represents an array level as its own descriptor, whereas gm2 passes a flat block — accepting the corpus files would compile but miscompile.
• Single-invocation compilation: definition, implementation and program units are compiled together into one image; there is no incremental object/archive format.
• OO is the Clarion-form CLASS dialect (single inheritance, virtual dispatch), not ISO 10514-3.

Sources

• V3: docs/STATUS.md, docs/features.md, docs/gaps/README.md, docs/analysis_gaps.md, docs/summary_session-2026-10-22.md; the gates in this document were run at v3-proc-types.
• gm2: the GCC Modula-2 testsuite and gm2-libs/gm2-libs-iso definition modules in the sibling gcc-git tree.
• PIM4: N. Wirth, Programming in Modula-2, 4th edition.