Compile bug: SYCL error: CHECK_TRY_ERROR

[HP-L]

Git commit

(llm) ➜ build-nuc git:(master) git rev-parse --short HEAD
8be759e

I have two log.(gdb-log.txt,build-log.tx)

gdb-log.txt
build-log.txt

Operating systems

Linux

GGML backends

SYCL

Problem description & steps to reproduce

Preliminary warning: ZES_ENABLE_SYSMAN=1 is not enabled, so the Intel GPU cannot retrieve accurate free memory statistics.
Fatal error: could not create a memory object — the SYCL backend failed to allocate GPU memory on the integrated graphics card.
Crash location: ggml-sycl matrix multiplication operator. Insufficient VRAM/device memory for the 35B large model caused the SYCL memory allocation failure.
The program ultimately crashed with the SIGABRT signal and generated a core dump file.
Supplementary GDB warnings: Missing program debugging symbols, missing glibc source code, and the SYCL debugging script was blocked by security policy restrictions.

First Bad Commit

Reproduction Steps

1. Check out the commit with hash 8be759e6f.
2. Compile with the following CMake command:

cmake -DGGML_SYCL=ON \
      -DGGML_SYCL_F16=ON \
      -DCMAKE_CXX_COMPILER=icpx \
      -DCMAKE_C_COMPILER=icx \
      -DGGML_MKL=ON \
      ..

3. After building completes, run the program:

./bin/llama-cli -m /home/nuc/gguf/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q3_K_P.gguf

Initiate a conversation with the AI.
4. The issue can be reproduced.

Compile command

source /opt/intel/oneapi/setvars.sh

cmake -DGGML_SYCL=ON \
      -DGGML_SYCL_F16=ON \
      -DCMAKE_CXX_COMPILER=icpx \
      -DCMAKE_C_COMPILER=icx \
      -DGGML_MKL=ON \
      ..

make -j$(nproc)

Relevant log output

Exception caught at file:/home/nuc/git_code/llama.cpp/ggml/src/ggml-sycl/ggml-sycl.cpp, line:3055, func:operator()
SYCL error: CHECK_TRY_ERROR(op(ctx, src0, src1, dst, src0_dd_i, src1_ddf_i, src1_ddq_i, dst_dd_i, dev[i].row_low, dev[i].row_high, src1_ncols, src1_padded_col_size, stream)): Exception caught in this line of code.
  in function ggml_sycl_op_mul_mat at /home/nuc/git_code/llama.cpp/ggml/src/ggml-sycl/ggml-sycl.cpp:3055
/home/nuc/git_code/llama.cpp/ggml/src/ggml-sycl/../ggml-sycl/common.hpp:150: SYCL error                                                                                                                      /[New LWP 380519]
[New LWP 380518]
[New LWP 380288]
[New LWP 380275]
[New LWP 378858]
[New LWP 378857]
warning: File "/home/nuc/intel/oneapi/compiler/2025.3/lib/libsycl.so.8.0.0-gdb.py" auto-loading has been declined by your `auto-load safe-path' set to "$debugdir:$datadir/auto-load".
To enable execution of this file add
        add-auto-load-safe-path /home/nuc/intel/oneapi/compiler/2025.3/lib/libsycl.so.8.0.0-gdb.py
line to your configuration file "/home/nuc/.config/gdb/gdbinit".
To completely disable this security protection add
        set auto-load safe-path /
line to your configuration file "/home/nuc/.config/gdb/gdbinit".
For more information about this security protection see the
"Auto-loading safe path" section in the GDB manual.  E.g., run from the shell:
        info "(gdb)Auto-loading safe path"
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
0x00007baca3898d71 in __futex_abstimed_wait_common64 (private=0, cancel=true, abstime=0x7ffeff151138, op=137, expected=0, futex_word=0x4c2804c) at ./nptl/futex-internal.c:57
warning: 57     ./nptl/futex-internal.c: No such file or directory
#0  0x00007baca3898d71 in __futex_abstimed_wait_common64 (private=0, cancel=true, abstime=0x7ffeff151138, op=137, expected=0, futex_word=0x4c2804c) at ./nptl/futex-internal.c:57
57      in ./nptl/futex-internal.c
#1  __futex_abstimed_wait_common (cancel=true, private=0, abstime=0x7ffeff151138, clockid=-4, expected=0, futex_word=0x4c2804c) at ./nptl/futex-internal.c:87
87      in ./nptl/futex-internal.c
#2  __GI___futex_abstimed_wait_cancelable64 (futex_word=futex_word@entry=0x4c2804c, expected=expected@entry=0, clockid=clockid@entry=1, abstime=abstime@entry=0x7ffeff151138, private=private@entry=0) at ./nptl/futex-internal.c:139
139     in ./nptl/futex-internal.c
#3  0x00007baca389c116 in __pthread_cond_wait_common (abstime=<optimized out>, clockid=<optimized out>, mutex=0x4c27ff8, cond=0x4c28020) at ./nptl/pthread_cond_wait.c:503
warning: 503    ./nptl/pthread_cond_wait.c: No such file or directory
#4  ___pthread_cond_clockwait64 (abstime=<optimized out>, clockid=<optimized out>, mutex=0x4c27ff8, cond=0x4c28020) at ./nptl/pthread_cond_wait.c:691
691     in ./nptl/pthread_cond_wait.c
#5  ___pthread_cond_clockwait64 (cond=0x4c28020, mutex=0x4c27ff8, clockid=<optimized out>, abstime=<optimized out>) at ./nptl/pthread_cond_wait.c:679
679     in ./nptl/pthread_cond_wait.c
#6  0x00007baca7d06e2b in server_response::recv_with_timeout(std::unordered_set<int, std::hash<int>, std::equal_to<int>, std::allocator<int> > const&, int) () from /home/nuc/git_code/llama.cpp/build-nuc/bin/libllama-cli-impl.so
#7  0x00007baca7d07f6b in server_response_reader::next(std::function<bool ()> const&) () from /home/nuc/git_code/llama.cpp/build-nuc/bin/libllama-cli-impl.so
#8  0x00007baca7c91760 in cli_context::generate_completion[abi:cxx11](result_timings&) () from /home/nuc/git_code/llama.cpp/build-nuc/bin/libllama-cli-impl.so
#9  0x00007baca7c88a3c in llama_cli(int, char**) () from /home/nuc/git_code/llama.cpp/build-nuc/bin/libllama-cli-impl.so
#10 0x00007baca382a1ca in __libc_start_call_main (main=main@entry=0x4064f0 <main>, argc=argc@entry=3, argv=argv@entry=0x7ffeff154698) at ../sysdeps/nptl/libc_start_call_main.h:58
warning: 58     ../sysdeps/nptl/libc_start_call_main.h: No such file or directory
#11 0x00007baca382a28b in __libc_start_main_impl (main=0x4064f0 <main>, argc=3, argv=0x7ffeff154698, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7ffeff154688) at ../csu/libc-start.c:360
warning: 360    ../csu/libc-start.c: No such file or directory
#12 0x0000000000406425 in _start ()
[Inferior 1 (process 378854) detached]                                                                                                                                                                       \[1]    378854 IOT instruction (core dumped)  ./bin/llama-cli -m
(llm) ➜  build-nuc git:(master) ls
app  CMakeCache.txt  cmake_install.cmake  compile_commands.json  CTestTestfile.cmake    examples  license.cpp         llama.pc             Makefile  src      tests  vendor
bin  CMakeFiles      common               core.378854            DartConfiguration.tcl  ggml      llama-config.cmake  llama-version.cmake  pocs      Testing  tools

[HP-L]

This is my device

get_memory_info: [warning] ext_intel_free_memory is not supported (export/set ZES_ENABLE_SYSMAN=1 to support), use total memory as free memory
0.00.126.721 I log_info: verbosity = 3 (adjust with the `-lv N` CLI arg)
0.00.126.724 I device_info:
0.00.126.784 I   - SYCL0   : Intel(R) Arc(TM) Graphics (29469 MiB, 29469 MiB free)
0.00.126.789 I   - CPU     : Intel(R) Core(TM) Ultra 5 125H (31829 MiB, 31829 MiB free)
0.00.126.835 I system_info: n_threads = 18 (n_threads_batch = 18) / 18 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX_VNNI = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 |
0.00.126.838 I srv  llama_server: n_parallel is set to auto, using n_parallel = 4 and kv_unified = true
0.00.127.217 I srv          init: using 17 threads for HTTP server
0.00.127.413 I srv         start: binding port with default address family
0.00.128.577 I srv  llama_server: loading model
0.00.128.623 I srv    load_model: loading model '/home/nuc/gguf/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q3_K_P.gguf'

[HP-L]

Add -verbosity 5 log

llama-server--verbosity5.txt

[arthw]

@HP-L
1.
I feel it's driver issue.
Check them:

lspci -nnk | grep -i vga -A3
dpkg -l | grep libze-intel-gpu1

Upgrade the driver to latest and try again: https://github.com/intel/compute-runtime/releases

Maybe less free memory.
SYCL can't detect free mem of all GPUs.
It use the all mem to replace the free mem in code.
Reboot the PC and try again.

Thank you!