IR support for switch instructions

[vaivaswatha]

Description

This PR adds a new IR instruction switch, that supports both exhaustive (for exhaustive matchs, verified by the front-end) and non-exhaustive switches.

Since matchs in the frontend still translate to an if-else ladder, a switch is never generated yet. So for testing, the PR adds support in the test harness to compile and execute (in the VM) sway-ir files directly.

To be able to snapshot-test the generated ASM output, we will add compile-ir to forc as well. This will be done in a follow up PR in which we will also add snapshot tests for should_pass/language/ir_switch tests added in this PR

This is the first step on the backend in implementing #7503.

[codspeed-hq]

Merging this PR will not alter performance

✅ 25 untouched benchmarks

---

_{Comparing vaivaswatha/ir_switch (62f3856) with master (f9de545)}

[cursor]

PR Summary

Medium Risk
Adds a new IR terminator and threads it through parsing, verification, CFG/optimizer utilities, and Fuel ASM generation, which could affect compilation correctness and control-flow handling. EVM codegen stubs switch as todo!(), so any accidental use on that backend would panic.

Overview
Adds a new Sway-IR switch terminator (optional default; None means exhaustive) with parser/printer support, verifier rules (u64 discriminant, no duplicates, exhaustive-range enforcement), and updated CFG utilities/optimizations to understand switch successors and branch arguments.

Implements Fuel backend lowering for switch by introducing an organizational Switch op that expands into a jump-table in the data section (Datum::WordArray) and a computed jump sequence; constant propagation and other ASM passes are updated for the new control-flow op. EVM backend plumbing is added but compile_switch remains todo!().

Adds IR compilation + test coverage by exposing forc_pkg::compile_ir (requires new sway-ir dep) and extending the e2e harness with a new ir_run category that compiles main.ir and executes it in the VM, along with new IR switch test programs and simplify-cfg tests.

^{Written by Cursor Bugbot for commit 62f3856. This will update automatically on new commits. Configure here.}

[cursor]

In-place SUBI corrupts discriminant register for successor blocks

High Severity

The SUBI writes the adjusted discriminant back into discrim_reg, the same virtual register returned by value_to_register(discriminant). This register is shared by all uses of that IR value. If any successor block references the discriminant directly (not via phi/block arguments), value_to_register returns the same register, and the successor sees the subtracted value instead of the original. A fresh temporary register from self.reg_seqr.next() is needed for the adjusted value.

 

[ironcev]

Great catch! This is indeed a bug!

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[ironcev]

Putting to draft until https://github.com/FuelLabs/sway/pull/7536/changes#r2821795856 is fixed and tested.