April 2026 - Wine-GitLab - list.winehq.org

[PATCH v2 0/1] MR10528: oleaut32: Add fast paths for VarAdd and VarCmp with integer operands.
by Francis De Brabandere (＠francisdb) April 1, 2026

April 1, 2026

Add fast paths for `VarAdd` and `VarCmp` when both operands are simple integer types (VT_I2, VT_I4, VT_EMPTY), skipping the expensive `VariantChangeType`/`VariantCopyInd` machinery. Profiling VBScript For-loops showed that `VarAdd`/`VarCmp` are called every iteration for the loop counter, and the full VARIANT conversion pipeline (including locale allocation in ucrtbase) was the dominant cost. With this patch, a VBScript For-loop inside a Sub goes from 3,041ms to 1,712ms for 10M iterations (**1.8× faster**). Benchmark results (VBScript `cscript`, 10M iterations unless noted): | Benchmark | Before | After | Speedup | |-----------|--------|-------|---------| | For-loop in Sub | 3,041 ms | 1,712 ms | **1.8×** | | Empty For (global) | 6,125 ms | 4,792 ms | **1.3×** | | For Step 2 (5M) | 3,058 ms | 2,404 ms | **1.3×** | | For R8 counter | 10,824 ms | 10,878 ms | ~same (R8 not fast-pathed) | Global-scope loops show smaller gains because VBScript's string-based identifier lookup still dominates there (see !10515). Combined with !10515, integer For-loops reach 1.2–1.5× of native Windows performance (up to 51× faster than current master). -- v2: oleaut32: Add fast paths for VarAdd and VarCmp with integer operands. https://gitlab.winehq.org/wine/wine/-/merge_requests/10528

2 1

[PATCH 0/1] MR10528: oleaut32: Add fast paths for VarAdd and VarCmp with integer operands.
by Francis De Brabandere (＠francisdb) April 1, 2026

April 1, 2026

Skip the expensive VariantChangeType/VariantCopyInd machinery when both operands are simple integer types (VT_I2, VT_I4, VT_EMPTY). For VarAdd, handle overflow promotion (I2+I2 that overflows to I4, I4+I4 that overflows to R8) matching native Windows behavior. For VarCmp, compare via LONGLONG to handle mixed I2/I4 correctly. This eliminates the dominant cost in VBScript For-loop step/compare operations, where VarAdd and VarCmp are called every iteration with trivial I2 operands but previously went through the full VARIANT conversion pipeline including locale allocation in ucrtbase. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10528

2 1

Re: [PATCH v9 0/2] MR10460: msvcrt: Add support for %Z printf specifier.
by Piotr Caban (＠piotr) April 1, 2026

April 1, 2026

On Tue Mar 31 18:55:56 2026 +0000, Elizabeth Figura wrote: > I believe we have some of those structures elsewhere already. Is there a > known compiler that doesn't support them? I think that it's not supported by MSVC. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10460#note_134611

1 0

[PATCH v3 0/5] MR10515: vbscript: Bind local variables, arguments, and class properties at compile time.
by Francis De Brabandere (＠francisdb) April 1, 2026

April 1, 2026

Resolve identifiers to direct indices during compilation instead of doing runtime wcsicmp string scans in lookup_identifier(). This adds new opcodes (OP_local, OP_assign_local, OP_set_local, OP_step_local, OP_incc_local, OP_local_prop, OP_assign_local_prop, OP_set_local_prop) that access local variables, function arguments, and class properties by index, eliminating the linear string comparison overhead for the most common cases. This mirrors jscript's OP_local / bind_local() optimization (dlls/jscript/compile.c) and addresses the FIXME in lookup_identifier() noting that class property access should be bound at compile time. Names that cannot be resolved at compile time (globals, dynamic vars, host objects) continue to use the existing string-based path. It's still 10x slower compared to windows but I have another related MR coming for that and it's related to the loop Windows ``` Local vars: 156.3 ms Arguments: 156.3 ms Class props: 390.6 ms Iterations: 1,000,000 ``` | Phase | Local vars | Arguments | Class props | |-------|-----------|-----------|-------------| | Before (baseline) | 10,200 ms | 6,700 ms | 6,710 ms | | `OP_local` | 5,130 ms | 2,035 ms | 5,920 ms | 2.0× / 3.3× | | `OP_assign_local` | 1,880 ms | 1,125 ms | 5,000 ms | 5.4× / 6.0× | | `OP_step_local`, `OP_incc_local` | 960 ms | 968 ms | 4,830 ms | **10.6× / 6.9× | | `OP_local_prop`, `OP_assign_local_prop` | 968 ms | 972 ms | 968 ms | **10.5× / 6.9× | -- v3: vbscript: Bind global-scope Dim variables at compile time. https://gitlab.winehq.org/wine/wine/-/merge_requests/10515

2 5