Support IVL_EX_ARRAY_PATTERN in object evaluation for queue-typed class properties: build an empty queue and append pattern elements, matching darray array-pattern object handling. Fix vec4 stack use so %queue/append_word/v is not double-popped. Update class queue locator regression to use direct c.q = '{...} assignment.
Made-with: Cursor
Route class queue-property min/max calls with `with (...)` through predicate-based locator elaboration so filtered extrema behave consistently with queue and darray locator semantics.
Made-with: Cursor
Enable queue locator elaboration for class properties typed as queues (not only dynamic arrays), and add regression coverage for class queue property find/unique/min/max behavior.
Made-with: Cursor
Enable unique and unique_index locator methods with `with (...)` for queues, dynamic arrays, and class dynamic-array properties, including VVP opcode/runtime support and regression coverage.
Made-with: Cursor
Handle IVL_EX_ARRAY_PATTERN in object evaluation by constructing a dynamic-array object and filling elements, so direct class property assignments like c.d = '{...}; codegen correctly without requiring a temporary variable.
Made-with: Cursor
Add class dynamic-array property support for locator methods (find*, unique*, min/max, with predicates) by elaborating dynamic-array property method calls to queue-method sfuns. Extend VVP property paths to treat property values as queue-or-darray sources, not queue-only, and fix object-stack handling in property locator loops so temporary accumulator objects do not hide the class object.
Add regression coverage for class dynamic-array property locators and update locator-method test documentation/listing.
Made-with: Cursor
Add dynamic-array unique() and unique_index() support and extend queue/darray min() and max() to accept locator with(...) predicates. Generate dedicated _with sfuns for min/max, emit VVP code for filtered scans, and reduce through new object-stack min/max opcodes.
Cover the new behavior with ivtest regressions for darray unique and queue/darray min/max with predicates, and update locator method documentation.
Made-with: Cursor
Implement queue and dynamic-array min()/max() for integral vec4 element types, with VVP support for both queue-backed and darray-backed storage and regressions in ivtest. Keep min/max as no-arg methods and document current behavior alongside existing locator tests.
Made-with: Cursor
Implement array locator methods for SystemVerilog queues and dynamic arrays:
find, find_index, find_first, find_first_index, find_last, find_last_index,
unique, and unique_index. Support both value arguments (equality) and
with (predicate) forms using implicit item and index in the iterator scope.
Elaboration returns queue-typed results for first/last/index locators per LRM
(empty queue when no match). Fix assignment compatibility between queues and
dynamic arrays where element types match.
VVP: extend %queue/size, %queue/word, %queue/find*, and %queue/unique* paths
so nets holding vvp_darray (including atom-backed int[]) are handled, not
only vvp_queue_vec4. Queue types still subclass vvp_darray; resolve vec4
queues before plain dynamic arrays. Fall back to the legacy get_queue_object
path for non-vec4 queues. Document opcode and source file touchpoints in
ivtest/ivltests/README_sv_queue_locators.txt.
Add vvp regressions: sv_queue_find*, sv_queue_unique, sv_queue_find_locators_ext,
sv_darray_find_locators.
Made-with: Cursor
Implement call_chain_expr in the parser, PECallFunction chain prefixes,
elaboration for class methods on returned handles, and prefix class resolution
for width checks on multi-hop chains. Place call_chain_expr before bare
hierarchy_identifier in expr_primary so id ( is not reduced as a lone ident.
Document behavior and developer notes in devel/sv_call_chain.md (including
that iverilog uses the installed lib/ivl/ivl binary).
Add ivtest sv_call_chain_method1 normal test with -g2012.
Made-with: Cursor
Add vvp regression for push_front, push_back, pop_front, pop_back on
queue-typed class fields. Catches vvp opcode_table sort regressions that
surface as Invalid opcode at runtime.
Made-with: Cursor
Internally, the maximum address space of a vector is 31 bits + a sign bit
to signal invalid addresses (out of bounds or has one or more x or z bits).
This commit ensures that unsigned part-select bit addresses which would
otherwise overflow and wrap around within this address space are correctly
handled as out of bounds.
mjoekhan
Martin Whitaker
Cary R
Wilson Snyder
Lars-Peter Clausen
Dag Lem