fix: ensure that purity analysis pass explores all functions

[TomAFrench]

Description

Problem*

From discussion with @michaeljklein

Fixes #8625

Summary*

This PR adds some sanity checks to SSA passes so that, for instance, the SSA pass to remove bitshifts does not leave behind any bitshifts in ACIR functions.

Additional Context

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[TomAFrench]

Huh, seems like we have a legit issue with purity analysis?

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Execution Time'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 98851e8	Previous: b2bfdb7	Ratio
semaphore-depth-10	0.029 s	0.021 s	1.38

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench

[TomAFrench]

Here's the SSA for `array_sort` after the purity analysis as built from master with settings which panic on this branch.

After Purity Analysis (1):
brillig(inline) impure fn main f0 {
  b0(v0: [u8; 3]):
    inc_rc v0
    v1 = allocate -> &mut [u8; 3]
    store v0 at v1
    inc_rc v0
    inc_rc v0
    call f2(v1, u32 0, u32 2, Field 3)
    v6 = load v1 -> [u8; 3]
    v7 = array_get v6, index u32 0 -> u8
    constrain v7 == u8 1
    v10 = array_get v6, index u32 1 -> u8
    constrain v10 == u8 2
    v12 = array_get v6, index u32 2 -> u8
    constrain v12 == u8 3
    return
}
brillig(inline) fn cmp f1 {
  b0(v0: u8, v1: u8):
    v4 = lt v0, v1
    jmpif v4 then: b1, else: b2
  b1():
    jmp b3(Field 0)
  b2():
    v5 = lt v1, v0
    jmpif v5 then: b4, else: b5
  b3(v2: Field):
    return v2
  b4():
    jmp b6(Field 2)
  b5():
    jmp b6(Field 1)
  b6(v3: Field):
    jmp b3(v3)
}
brillig(inline) impure fn quicksort_loop f2 {
  b0(v0: &mut [u8; 3], v1: u32, v2: u32, v3: Field):
    v5 = make_array [v1, v2] : [(u32, u32)]
    v6 = allocate -> &mut u32
    store u32 1 at v6
    v8 = allocate -> &mut [(u32, u32)]
    store v5 at v8
    jmp b1(u32 0)
  b1(v4: u32):
    v11 = lt v4, u32 6
    jmpif v11 then: b2, else: b3
  b2():
    v12 = load v6 -> u32
    v13 = load v8 -> [(u32, u32)]
    inc_rc v13
    v14 = eq v12, u32 0
    jmpif v14 then: b4, else: b5
  b3():
    return
  b4():
    jmp b3()
  b5():
    v15 = load v6 -> u32
    v16 = load v8 -> [(u32, u32)]
    inc_rc v16
    v18, v19, v20, v21 = call slice_pop_back(v15, v16) -> (u32, [(u32, u32)], u32, u32)
    inc_rc v19
    store v18 at v6
    store v19 at v8
    v22 = add v20, u32 1
    v23 = lt v21, v22
    jmpif v23 then: b6, else: b7
  b6():
    v35 = unchecked_add v4, u32 1
    jmp b1(v35)
  b7():
    v25 = call f3(v0, v20, v21, v3) -> u32
    inc_rc v19
    v26 = add v25, u32 1
    v28, v29 = call slice_push_back(v18, v19, v26, v21) -> (u32, [(u32, u32)])
    store v28 at v6
    store v29 at v8
    v30 = lt u32 0, v25
    jmpif v30 then: b8, else: b9
  b8():
    inc_rc v29
    v31 = sub v25, u32 1
    v32, v33 = call slice_push_back(v28, v29, v20, v31) -> (u32, [(u32, u32)])
    store v32 at v6
    store v33 at v8
    jmp b9()
  b9():
    v34 = unchecked_add v4, u32 1
    jmp b1(v34)
}
brillig(inline) fn partition f3 {
  b0(v0: &mut [u8; 3], v1: u32, v2: u32, v3: Field):
    v5 = allocate -> &mut u32
    store v1 at v5
    jmp b1(v1)
  b1(v4: u32):
    v6 = lt v4, v2
    jmpif v6 then: b2, else: b3
  b2():
    v19 = load v0 -> [u8; 3]
    v20 = lt v4, u32 3
    constrain v20 == u1 1, "Index out of bounds"
    v21 = array_get v19, index v4 -> u8
    v22 = lt v2, u32 3
    constrain v22 == u1 1, "Index out of bounds"
    v23 = array_get v19, index v2 -> u8
    v25 = call f1(v21, v23) -> Field
    v27 = eq v25, Field 2
    v28 = not v27
    jmpif v27 then: b5, else: b4
  b3():
    v7 = load v0 -> [u8; 3]
    v8 = load v5 -> u32
    v10 = lt v8, u32 3
    constrain v10 == u1 1, "Index out of bounds"
    v12 = array_get v7, index v8 -> u8
    v13 = lt v2, u32 3
    constrain v13 == u1 1, "Index out of bounds"
    v14 = array_get v7, index v2 -> u8
    v15 = lt v8, u32 3
    constrain v15 == u1 1, "Index out of bounds"
    v16 = array_set v7, index v8, value v14
    v17 = lt v2, u32 3
    constrain v17 == u1 1, "Index out of bounds"
    v18 = array_set v16, index v2, value v12
    store v18 at v0
    return v8
  b4():
    v29 = load v0 -> [u8; 3]
    v30 = load v5 -> u32
    v31 = lt v30, u32 3
    constrain v31 == u1 1, "Index out of bounds"
    v32 = array_get v29, index v30 -> u8
    v33 = lt v4, u32 3
    constrain v33 == u1 1, "Index out of bounds"
    v34 = array_get v29, index v4 -> u8
    v35 = lt v30, u32 3
    constrain v35 == u1 1, "Index out of bounds"
    v36 = array_set v29, index v30, value v34
    v37 = lt v4, u32 3
    constrain v37 == u1 1, "Index out of bounds"
    v38 = array_set v36, index v4, value v32
    store v38 at v0
    v40 = add v30, u32 1
    store v40 at v5
    jmp b5()
  b5():
    v41 = unchecked_add v4, u32 1
    jmp b1(v41)
}

[github-actions]

Changes to number of Brillig opcodes executed

Generated at commit: db97feee5e43f8845bc031f47558f02be8517e32, compared to commit: d992ad59d4c787aa12b0176e6f9d12202e4d9f71

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
derive_inliner_min	-36 ✅	-10.26%
derive_inliner_zero	-38 ✅	-13.15%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
hashmap_inliner_min	73,232 (-103)	-0.14%
slice_regex_inliner_zero	3,878 (-112)	-2.81%
derive_inliner_min	315 (-36)	-10.26%
derive_inliner_zero	251 (-38)	-13.15%

[github-actions]

Changes to Brillig bytecode sizes

Generated at commit: db97feee5e43f8845bc031f47558f02be8517e32, compared to commit: d992ad59d4c787aa12b0176e6f9d12202e4d9f71

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
derive_inliner_min	-13 ✅	-3.63%
derive_inliner_zero	-13 ✅	-4.59%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
hashmap_inliner_min	8,749 (-11)	-0.13%
slice_regex_inliner_zero	1,632 (-12)	-0.73%
derive_inliner_min	345 (-13)	-3.63%
derive_inliner_zero	270 (-13)	-4.59%

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Test Suite Duration'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: b2efbfa	Previous: d992ad5	Ratio
test_report_AztecProtocol_aztec-packages_noir-projects_aztec-nr	4 s	3 s	1.33

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench

[TomAFrench]

This is now ready for review

[TomAFrench]

@aakoshh Can you take all the fuzz stuff out of the merge queue CI? It seems like every other PR gets blocked on this.