Recursive jmp instruction produced by icx

[c-algua]

Describe the bug

Compilation of CLAPACK's dlamch.c using icx produces a recursive jmp instruction on itself, for optimization levels >= -O1

The same problem is not present with the clang binary distributed alongside icx.

To reproduce

1. Unpack dlamch.tar.gz
2. CC=${CMPLR_ROOT}/bin/icx make run: this command will hang at runtime (after printing Start) due to the recursive jmp produced by icx
3. CC=${CMPLR_ROOT}/bin/icx make dlamch.s: you will find the jmp instruction shortly after the dlamc2_ label, around line 242 for -O1, e.g.:

.LBB1_4:                                # =>This Inner Loop Header: Depth=1
        jmp     .LBB1_4

4. make clean
5. CC=${CMPLR_ROOT}/bin/compiler/clang make run: this will run correctly (printing End after Start)
6. CC=${CMPLR_ROOT}/bin/compiler/clang make dlamch.s: inspecting the assembly there's no such jmp

Environment

• OS: tested on Ubuntu 24.04 using docker's image intel/oneapi:2026.1.0-devel-ubuntu24.04
• Target device and vendor: x86
• DPC++ version: Intel(R) oneAPI DPC++/C++ Compiler 2026.1.0 (2026.1.0.20260617)
• Dependencies version: [opencl:cpu][opencl:0] Intel(R) OpenCL, AMD Ryzen 5 7535HS with Radeon Graphics OpenCL 3.0 (Build 0) [2026.21.6.0.17_160000]

Additional context

No response