Stream: git-wasmtime

Topic: wasmtime / issue #12170 Cranelift: unimplemented for > 64...

Wasmtime GitHub notifications bot (Dec 16 2025 at 12:12):

abc767234318 opened issue #12170:

.clif Test Case

test run

set enable_nan_canonicalization=true
set preserve_frame_pointers=true
set enable_multi_ret_implicit_sret=true
set opt_level=none

target x86_64 sse42 has_avx
target aarch64
target s390x
target riscv64gc has_zcd has_zbkb has_zbc has_zbs has_zicond has_zvl32b has_zvl64b has_zvl128b has_zvl1024b has_zvl2048b has_zvl4096b has_zvl8192b has_zvl16384b has_zvl32768b

function %main() -> f64 fast {
    ss0 = explicit_slot 32
    ss1 = explicit_slot 32
    const0 = 0xfa2675c080000000e8a433230a7479e5

block0:
    v1 = iconst.i16 1348
    v2 = f64const -0x1.e10d21f69a163p995
    v3 = vconst.i32x4 const0
    v4 = smin.i32x4 v3, v3  ; v3 = const0, v3 = const0
    v5 = clz.i16 v1  ; v1 = 1348
    v6 = icmp.i32x4 slt v3, v4  ; v3 = const0

    return v2
}

; print: %main()

Steps to Reproduce

This test case processed successfully when the opt_level=none, but if I change it to the opt_level=speed, it will throw thread 'main' panicked at cranelift/codegen/src/opts.rs:191:5: unimplemented for > 64 bits.

Expected Results

%main() -> -0x1.e10d21f69a163p995

Actual Results

thread 'main' panicked at cranelift/codegen/src/opts.rs:191:5:
unimplemented for > 64 bits
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Wasmtime GitHub notifications bot (Dec 16 2025 at 12:12):

abc767234318 added the bug label to Issue #12170.

Wasmtime GitHub notifications bot (Dec 16 2025 at 12:12):

abc767234318 added the cranelift label to Issue #12170.

Wasmtime GitHub notifications bot (Dec 18 2025 at 18:11):

alexcrichton commented on issue #12170:

@abc767234318 it looks like reporting this may have gone awry by accident as the source here says set opt_level=none, but I believe set opt_level=speed is needed to reproduce. After changing that I can reproduce myself.

@cfallin w.r.t. this comment personally I think we still want to investigate these sorts of issues. For example this isn't an esoteric instruction, this input:

(module
  (func (export "hi") (result v128)
    (local $i v128)
    v128.const i64x2 0xfa2675c080000000 0xe8a433230a7479e5
    local.set $i
    local.get $i
    local.get $i
    local.get $i
    i32x4.min_s
    i32x4.lt_s
  )
)

fails with:

$ cargo run foo.wat
...

thread '<unnamed>' (7768935) panicked at cranelift/codegen/src/opts.rs:191:5:
unimplemented for > 64 bits
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

so this is not a case of an obscure cranelift instruction on an obscure platform, but definitely reachable from wasm and is something we should fix.

Wasmtime GitHub notifications bot (Dec 18 2025 at 18:12):

alexcrichton added the fuzz-bug label to Issue #12170.

Wasmtime GitHub notifications bot (Dec 18 2025 at 18:16):

cfallin commented on issue #12170:

Ah, I missed that this was reachable from Wasm -- yes, it'd still be nice to build the table of missing corners in the full CLIF-space but this is higher priority, agreed.

Wasmtime GitHub notifications bot (Dec 18 2025 at 18:53):

cfallin closed issue #12170:

.clif Test Case

test run

set enable_nan_canonicalization=true
set preserve_frame_pointers=true
set enable_multi_ret_implicit_sret=true
set opt_level=none

target x86_64 sse42 has_avx
target aarch64
target s390x
target riscv64gc has_zcd has_zbkb has_zbc has_zbs has_zicond has_zvl32b has_zvl64b has_zvl128b has_zvl1024b has_zvl2048b has_zvl4096b has_zvl8192b has_zvl16384b has_zvl32768b

function %main() -> f64 fast {
    ss0 = explicit_slot 32
    ss1 = explicit_slot 32
    const0 = 0xfa2675c080000000e8a433230a7479e5

block0:
    v1 = iconst.i16 1348
    v2 = f64const -0x1.e10d21f69a163p995
    v3 = vconst.i32x4 const0
    v4 = smin.i32x4 v3, v3  ; v3 = const0, v3 = const0
    v5 = clz.i16 v1  ; v1 = 1348
    v6 = icmp.i32x4 slt v3, v4  ; v3 = const0

    return v2
}

; print: %main()

Steps to Reproduce

This test case processed successfully when the opt_level=none, but if I change it to the opt_level=speed, it will throw thread 'main' panicked at cranelift/codegen/src/opts.rs:191:5: unimplemented for > 64 bits.

Expected Results

%main() -> -0x1.e10d21f69a163p995

Actual Results

thread 'main' panicked at cranelift/codegen/src/opts.rs:191:5:
unimplemented for > 64 bits
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Last updated: Jan 09 2026 at 13:15 UTC