Open
Description
Git commit
Operating systems
Linux
GGML backends
CUDA
Problem description & steps to reproduce
when compiling for all cuda architectures (50,52,60,61,70,75,80,86,89,90) , the generated ggml-cuda static/dynamic libraries are too large to be linked correctly
delete one target cuda architecture to make the generated file less than 2gb can compile correctly
First Bad Commit
No response
Compile command
cmake -B build -DBUILD_SHARED_LIBS=OFF -DLLAMA_CURL=ON -DGGML_CUDA=ON -DGGML_CUDA_FA_ALL_QUANTS=ON -DCMAKE_CUDA_ARCHITECTURES="50;52;60;61;70;75;80;86;89;90"
cmake --build build --config Release -j --target llama-serverRelevant log output
Consolidate compiler generated dependencies of target build_info
Consolidate compiler generated dependencies of target ggml-base
[ 0%] Built target build_info
[ 3%] Built target ggml-base
Consolidate compiler generated dependencies of target ggml-cpu
[ 10%] Built target ggml-cpu
Consolidate compiler generated dependencies of target ggml-cuda
[ 84%] Built target ggml-cuda
Consolidate compiler generated dependencies of target ggml
[ 84%] Building CXX object ggml/src/CMakeFiles/ggml.dir/ggml-backend-reg.cpp.o
[ 84%] Linking CXX static library libggml.a
[ 84%] Built target ggml
Consolidate compiler generated dependencies of target llama
[ 84%] Building CXX object src/CMakeFiles/llama.dir/llama.cpp.o
[ 84%] Building CXX object src/CMakeFiles/llama.dir/llama-adapter.cpp.o
[ 86%] Building CXX object src/CMakeFiles/llama.dir/llama-arch.cpp.o
[ 86%] Building CXX object src/CMakeFiles/llama.dir/llama-batch.cpp.o
[ 86%] Building CXX object src/CMakeFiles/llama.dir/llama-chat.cpp.o
[ 86%] Building CXX object src/CMakeFiles/llama.dir/llama-context.cpp.o
[ 87%] Building CXX object src/CMakeFiles/llama.dir/llama-grammar.cpp.o
[ 87%] Building CXX object src/CMakeFiles/llama.dir/llama-graph.cpp.o
[ 87%] Building CXX object src/CMakeFiles/llama.dir/llama-hparams.cpp.o
[ 87%] Building CXX object src/CMakeFiles/llama.dir/llama-impl.cpp.o
[ 89%] Building CXX object src/CMakeFiles/llama.dir/llama-io.cpp.o
[ 89%] Building CXX object src/CMakeFiles/llama.dir/llama-kv-cache.cpp.o
[ 89%] Building CXX object src/CMakeFiles/llama.dir/llama-memory.cpp.o
[ 89%] Building CXX object src/CMakeFiles/llama.dir/llama-mmap.cpp.o
[ 91%] Building CXX object src/CMakeFiles/llama.dir/llama-model-loader.cpp.o
[ 91%] Building CXX object src/CMakeFiles/llama.dir/llama-model.cpp.o
[ 91%] Building CXX object src/CMakeFiles/llama.dir/llama-quant.cpp.o
[ 91%] Building CXX object src/CMakeFiles/llama.dir/llama-sampling.cpp.o
[ 93%] Building CXX object src/CMakeFiles/llama.dir/llama-vocab.cpp.o
[ 93%] Building CXX object src/CMakeFiles/llama.dir/unicode.cpp.o
[ 93%] Building CXX object src/CMakeFiles/llama.dir/unicode-data.cpp.o
[ 94%] Linking CXX static library libllama.a
[ 94%] Built target llama
Consolidate compiler generated dependencies of target common
[ 94%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o
[ 96%] Building CXX object common/CMakeFiles/common.dir/chat.cpp.o
[ 96%] Building CXX object common/CMakeFiles/common.dir/common.cpp.o
[ 96%] Building CXX object common/CMakeFiles/common.dir/console.cpp.o
[ 96%] Building CXX object common/CMakeFiles/common.dir/json-schema-to-grammar.cpp.o
[ 98%] Building CXX object common/CMakeFiles/common.dir/llguidance.cpp.o
[ 98%] Building CXX object common/CMakeFiles/common.dir/log.cpp.o
[ 98%] Building CXX object common/CMakeFiles/common.dir/ngram-cache.cpp.o
[100%] Building CXX object common/CMakeFiles/common.dir/sampling.cpp.o
[100%] Building CXX object common/CMakeFiles/common.dir/speculative.cpp.o
[100%] Linking CXX static library libcommon.a
[100%] Built target common
Consolidate compiler generated dependencies of target llama-bench
[100%] Building CXX object tools/llama-bench/CMakeFiles/llama-bench.dir/llama-bench.cpp.o
[100%] Linking CXX executable ../../bin/llama-bench
/usr/lib/gcc/x86_64-linux-gnu/9/../../../x86_64-linux-gnu/Scrt1.o: in function `_start':
(.text+0x2a): relocation truncated to fit: R_X86_64_GOTPCRELX against symbol `__libc_start_main@@GLIBC_2.2.5' defined in .text section in /lib/x86_64-linux-gnu/libc.so.6
/usr/lib/gcc/x86_64-linux-gnu/9/../../../x86_64-linux-gnu/Scrt1.o:(.eh_frame+0x20): relocation truncated to fit: R_X86_64_PC32 against `.text'
/usr/lib/gcc/x86_64-linux-gnu/9/../../../x86_64-linux-gnu/crti.o: in function `_init':
(.init+0xb): relocation truncated to fit: R_X86_64_REX_GOTPCRELX against undefined symbol `__gmon_start__'
CMakeFiles/llama-bench.dir/llama-bench.cpp.o: in function `create_printer(output_formats)':
llama-bench.cpp:(.text+0x51): relocation truncated to fit: R_X86_64_PC32 against symbol `vtable for markdown_printer' defined in .data.rel.ro.local._ZTV16markdown_printer[_ZTV16markdown_printer] section in CMakeFiles/llama-bench.dir/llama-bench.cpp.o
llama-bench.cpp:(.text+0x6b): relocation truncated to fit: R_X86_64_PC32 against symbol `vtable for sql_printer' defined in .data.rel.ro.local._ZTV11sql_printer[_ZTV11sql_printer] section in CMakeFiles/llama-bench.dir/llama-bench.cpp.o
llama-bench.cpp:(.text+0x96): relocation truncated to fit: R_X86_64_PC32 against symbol `vtable for csv_printer' defined in .data.rel.ro.local._ZTV11csv_printer[_ZTV11csv_printer] section in CMakeFiles/llama-bench.dir/llama-bench.cpp.o
llama-bench.cpp:(.text+0xb8): relocation truncated to fit: R_X86_64_PC32 against symbol `vtable for json_printer' defined in .data.rel.ro.local._ZTV12json_printer[_ZTV12json_printer] section in CMakeFiles/llama-bench.dir/llama-bench.cpp.o
llama-bench.cpp:(.text+0xde): relocation truncated to fit: R_X86_64_PC32 against symbol `vtable for jsonl_printer' defined in .data.rel.ro.local._ZTV13jsonl_printer[_ZTV13jsonl_printer] section in CMakeFiles/llama-bench.dir/llama-bench.cpp.o
CMakeFiles/llama-bench.dir/llama-bench.cpp.o: in function `pair_str(std::pair<int, int> const&)':
llama-bench.cpp:(.text+0x801): relocation truncated to fit: R_X86_64_PC32 against `.bss'
llama-bench.cpp:(.text+0x834): relocation truncated to fit: R_X86_64_PC32 against `.bss'
llama-bench.cpp:(.text+0x88e): additional relocation overflows omitted from the output
../../bin/llama-bench: PC-relative offset overflow in PLT entry for `__printf_chk@@GLIBC_2.3.4'
collect2: error: ld returned 1 exit status
make[3]: *** [tools/llama-bench/CMakeFiles/llama-bench.dir/build.make:109: bin/llama-bench] Error 1
make[2]: *** [CMakeFiles/Makefile2:3549: tools/llama-bench/CMakeFiles/llama-bench.dir/all] Error 2
make[1]: *** [CMakeFiles/Makefile2:3556: tools/llama-bench/CMakeFiles/llama-bench.dir/rule] Error 2
make: *** [Makefile:1323: llama-bench] Error 2