Cranelift: internal error: entered unreachable code: no rule matched for term emit_fcmp at src/isa/x64/inst.isle line 3595; should it be partial? #12671
Description
.clif Test Case
not reduced at all - sorry
function u0:0(i64, i64, i64, i64) tail {
ss0 = explicit_slot 8, align = 8
sig0 = (i64, i64, i64, i64) tail
sig1 = (i64, i64, i64) -> i8 system_v
sig2 = (i64, i32, i64) system_v
sig3 = (i64, i32, i64) -> i8 system_v
sig4 = (i64, i32, i64) -> i8 system_v
sig5 = (i64, i32, i64) -> i8 system_v
sig6 = (i64, i32, i64) -> i8 system_v
sig7 = (i64, i32, i8) -> i8 system_v
sig8 = (i64, i32, i16) -> i8 system_v
sig9 = (i64, i32, i32) -> i8 system_v
sig10 = (i64, i32, i64) -> i8 system_v
sig11 = (i64, i32, i32, i64) -> i8 system_v
sig12 = (i64, i32, i32, f64) -> i8 system_v
sig13 = (i64, i32) system_v
sig14 = (i64) system_v
sig15 = (i64, i16) system_v
fn0 = u1:0 sig15
fn1 = u0:2 sig1
fn2 = u0:3 sig2
fn3 = u0:4 sig3
fn4 = u0:5 sig4
fn5 = u0:6 sig5
fn6 = u0:7 sig6
fn7 = u0:8 sig7
fn8 = u0:9 sig8
fn9 = u0:10 sig9
fn10 = u0:11 sig10
fn11 = u0:12 sig11
fn12 = u0:13 sig12
fn13 = u0:14 sig13
fn14 = u0:15 sig14
fn15 = u0:16 sig14
fn16 = u0:17 sig14
fn17 = u0:18 sig14
fn18 = u0:19 sig14
fn19 = u0:20 sig14
fn20 = u0:21 sig14
fn21 = u0:22 sig14
fn22 = u0:23 sig14
block0(v0: i64, v1: i64, v2: i64, v3: i64):
@0000 v4 = load.i32 notrap aligned v2+680
@0000 v5 = iconst.i32 13
@0000 v6 = ushr v4, v5 ; v5 = 13
@0000 v7 = band_imm v6, 1
@0000 v8 = ireduce.i8 v7
@0000 brif v8, block2, block1
block1 cold:
@0000 v9 = iconst.i16 2048
@0000 call fn0(v2, v9) ; v9 = 2048
@0000 return
block2:
@0000 v10 = load.f64x2 notrap aligned v2+632
@0000 v11 = load.f64x2 notrap aligned v2+312
@0000 v12 = fadd v10, v11
@0000 v13 = extractlane v12, 0
@0000 v14 = f64const 0.0
@0000 v15 = fcmp lt v13, v14 ; v14 = 0.0
@0000 v16 = fcmp gt v13, v14 ; v14 = 0.0
@0000 v17 = fcmp eq v13, v14 ; v14 = 0.0
@0000 v18 = fcmp uno v13, v14 ; v14 = 0.0
@0000 v19 = load.i32 notrap aligned v2+652
@0000 v20 = uextend.i32 v15
@0000 v21 = uextend.i32 v16
@0000 v22 = uextend.i32 v17
@0000 v23 = uextend.i32 v18
@0000 v24 = ishl_imm v20, 15
@0000 v25 = ishl_imm v21, 14
@0000 v26 = ishl_imm v22, 13
@0000 v27 = ishl_imm v23, 12
@0000 v28 = bor v24, v25
@0000 v29 = bor v28, v26
@0000 v30 = bor v29, v27
@0000 v31 = iconst.i32 0x0001_f000
@0000 v32 = band_not v19, v31 ; v31 = 0x0001_f000
@0000 v33 = bor v32, v30
@0000 v34 = load.i32 notrap aligned v2
@0000 v35 = iadd_imm v34, 4
@0001 v36 = f64const 0.0
@0001 v37 = splat.f64x2 v36 ; v36 = 0.0
@0001 v38 = load.f64x2 notrap aligned v2+280
@0001 v39 = load.f64x2 notrap aligned v2+232
@0001 v40 = load.f64x2 notrap aligned v2+216
@0001 v41 = bitcast.i64x2 v39
@0001 v42 = bitcast.i64x2 v40
@0001 v43 = fcmp ge v38, v37
@0001 v44 = select v43, v42, v41
@0001 v45 = bitcast.f64x2 v44
@0001 v46 = iadd_imm v35, 4
@0002 v47 = fadd v38, v45
@0002 v48 = extractlane v47, 0
@0002 v49 = f64const 0.0
@0002 v50 = fcmp lt v48, v49 ; v49 = 0.0
@0002 v51 = fcmp gt v48, v49 ; v49 = 0.0
@0002 v52 = fcmp eq v48, v49 ; v49 = 0.0
@0002 v53 = fcmp uno v48, v49 ; v49 = 0.0
@0002 v54 = uextend.i32 v50
@0002 v55 = uextend.i32 v51
@0002 v56 = uextend.i32 v52
@0002 v57 = uextend.i32 v53
@0002 v58 = ishl_imm v54, 15
@0002 v59 = ishl_imm v55, 14
@0002 v60 = ishl_imm v56, 13
@0002 v61 = ishl_imm v57, 12
@0002 v62 = bor v58, v59
@0002 v63 = bor v62, v60
@0002 v64 = bor v63, v61
@0002 v65 = iconst.i32 0x0001_f000
@0002 v66 = band_not v33, v65 ; v65 = 0x0001_f000
@0002 v67 = bor v66, v64
@0002 v68 = iadd_imm v46, 4
@0003 v69 = f64const 0.0
@0003 v70 = splat.f64x2 v69 ; v69 = 0.0
@0003 v71 = load.f64x2 notrap aligned v2+264
@0003 v72 = bitcast.i64x2 v39
@0003 v73 = bitcast.i64x2 v40
@0003 v74 = fcmp ge v71, v70
@0003 v75 = select v74, v73, v72
@0003 v76 = bitcast.f64x2 v75
@0003 v77 = iadd_imm v68, 4
store notrap aligned v77, v2
store notrap aligned v67, v2+652
store notrap aligned v76, v2+632
store notrap aligned v47, v2+280
store notrap aligned v12, v2+312
v78 = load.i32 notrap aligned v0
v79 = iadd_imm v78, 4
store notrap aligned v79, v0
v80 = load.i32 notrap aligned v0+4
v81 = iadd_imm v80, 8
store notrap aligned v81, v0+4
return
}
Steps to Reproduce
seems to only happen when running on release..?
Expected Results
should not panic
Actual Results
panics
Versions and Environment
Cranelift version or commit: 0.129
Operating system: Linux 6.19.2-2-cachyos
Architecture: x86_64 (Ryzen 5 5600)