Error when building in codespaces

[gitboy16]

Hi,

Please see below.

Thank you in advance for your help.

sandbox:${WORKSPACE}/srcdir # ls -lusah --color
total 12K    
     4 drwxr-xrw-    1 root     root        4.0K Aug 26 08:55 .
     4 drwxr-xrw-    6 root     root        4.0K Aug 26 08:50 ..
     4 drwxr-xrw-   12 root     root        4.0K Aug 26 08:50 x86-simd-sort
sandbox:${WORKSPACE}/srcdir # cd x86-simd-sort/
sandbox:${WORKSPACE}/srcdir/x86-simd-sort # ls -lusah --color
total 108K   
     4 drwxr-xrw-   12 root     root        4.0K Aug 26 08:50 .
     4 drwxr-xrw-    1 root     root        4.0K Aug 26 08:55 ..
     4 -rw-r--r--    1 root     root        2.8K Aug 26 08:50 .clang-format
     4 drwxr-xrw-    7 root     root        4.0K Aug 26 08:50 .git
     4 drwxr-xr--    3 root     root        4.0K Aug 26 08:50 .github
     4 -rw-r--r--    1 root     root         375 Aug 26 08:50 .gitignore
     8 -rw-r--r--    1 root     root        5.3K Aug 26 08:50 CODE_OF_CONDUCT.md
     4 -rw-r--r--    1 root     root        2.3K Aug 26 08:50 CONTRIBUTING.md
     4 -rw-r--r--    1 root     root        1.5K Aug 26 08:50 LICENSE.md
     4 -rw-r--r--    1 root     root        1.1K Aug 26 08:50 Makefile
    12 -rw-r--r--    1 root     root        9.5K Aug 26 08:50 README.md
     4 -rw-r--r--    1 root     root         298 Aug 26 08:50 SECURITY.md
     4 drwxr-xr--    2 root     root        4.0K Aug 26 08:50 benchmarks
     4 -rw-r--r--    1 root     root        2.3K Aug 26 08:50 example.c
     4 drwxr-xr--    2 root     root        4.0K Aug 26 08:50 examples
     4 drwxr-xr--    2 root     root        4.0K Aug 26 08:50 lib
     4 -rw-r--r--    1 root     root        3.9K Aug 26 08:50 meson.build
     4 -rw-r--r--    1 root     root        1.0K Aug 26 08:50 meson_options.txt
     4 drwxr-xr--    2 root     root        4.0K Aug 26 08:50 misc
     4 -rw-r--r--    1 root     root        2.6K Aug 26 08:50 run-bench.py
     4 drwxr-xr--    2 root     root        4.0K Aug 26 08:50 scripts
     4 drwxr-xr--    2 root     root        4.0K Aug 26 08:50 src
     4 drwxr-xr--    2 root     root        4.0K Aug 26 08:50 tests
     4 drwxr-xr--    2 root     root        4.0K Aug 26 08:50 utils
sandbox:${WORKSPACE}/srcdir/x86-simd-sort # meson setup --buildtype release builddir
The Meson build system
Version: 1.4.0
Source dir: /workspace/srcdir/x86-simd-sort
Build dir: /workspace/srcdir/x86-simd-sort/builddir
Build type: native build
Project name: x86-simd-sort
Project version: 7.0.x
C++ compiler for the host machine: x86_64-linux-musl-c++ (gcc 14.2.0 "x86_64-linux-musl-g++ (GCC) 14.2.0")
C++ linker for the host machine: x86_64-linux-musl-c++ ld.bfd 2.43.1
Host machine cpu family: x86_64
Host machine cpu: x86_64
Compiler for C++ supports arguments -march=haswell: NO 
Compiler for C++ supports arguments -march=skylake-avx512: NO 
Compiler for C++ supports arguments -march=icelake-client: NO 
Found pkg-config: YES (/usr/bin/pkg-config) 1.8.1
Build targets in project: 1

x86-simd-sort 7.0.x

  Configuration
    Can compile AVX-512 FP16 ISA: NO
    Build test content          : NO
    Build benchmarks            : NO

  User defined options
    buildtype                   : release

Found ninja-1.9 at /usr/bin/ninja
sandbox:${WORKSPACE}/srcdir/x86-simd-sort # cat /proc/cpuinfo  | grep 'name'| uniq
model name      : Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
sandbox:${WORKSPACE}/srcdir/x86-simd-sort # grep avx2 /proc/cpuinfo
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology tsc_reliable nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq vmx ssse3 fma cx16 pcid sse4_1 sse4_2 movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch tpr_shadow ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves vnmi avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid fsrm arch_capabilities
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology tsc_reliable nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq vmx ssse3 fma cx16 pcid sse4_1 sse4_2 movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch tpr_shadow ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves vnmi avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid fsrm arch_capabilities
sandbox:${WORKSPACE}/srcdir/x86-simd-sort # cd builddir
sandbox:${WORKSPACE}/srcdir/x86-simd-sort/builddir # meson compile
INFO: autodetecting backend as ninja
INFO: calculating backend command to run: /usr/bin/ninja
[1/2] Compiling C++ object libx86simdsortcpp.so.1.p/lib_x86simdsort.cpp.o
[2/2] Linking target libx86simdsortcpp.so.1
ninja: job failed: x86_64-linux-musl-c++  -o libx86simdsortcpp.so.1 libx86simdsortcpp.so.1.p/lib_x86simdsort.cpp.o -Wl,--as-needed -Wl,--no-undefined -Wl,-O1 -shared -fPIC -Wl,-soname,libx86simdsortcpp.so.1
/opt/x86_64-linux-musl/bin/../lib/gcc/x86_64-linux-musl/14.2.0/../../../../x86_64-linux-musl/bin/ld: libx86simdsortcpp.so.1.p/lib_x86simdsort.cpp.o: in function `x86simdsort::resolve_qsort_Float16()':
x86simdsort.cpp:(.text.startup+0x215): undefined reference to `void xss::fp16_icl::qsort<_Float16>(_Float16*, unsigned long, bool, bool)'
/opt/x86_64-linux-musl/bin/../lib/gcc/x86_64-linux-musl/14.2.0/../../../../x86_64-linux-musl/bin/ld: libx86simdsortcpp.so.1.p/lib_x86simdsort.cpp.o: relocation R_X86_64_PC32 against undefined hidden symbol `_ZN3xss8fp16_icl5qsortIDF16_EEvPT_mbb' can not be used when making a shared object
/opt/x86_64-linux-musl/bin/../lib/gcc/x86_64-linux-musl/14.2.0/../../../../x86_64-linux-musl/bin/ld: final link failed: bad value
collect2: error: ld returned 1 exit status
ninja: subcommand failed
sandbox:${WORKSPACE}/srcdir/x86-simd-sort/builddir # 

[r-devulap]

Could you try adding -fPIC to CXXFLAGS and rebuild?

[gitboy16]

Do you mean something like this?

CXXFLAGS=-fPIC meson setup --buildtype release builddir

Sorry I am not familiar with how meson works.

[gitboy16]

I tried the above and got the same error message.

[gitboy16]

I am wondering if the issue is not that in the file lib/meson.build no option is set properly because it does not satisfy any of the following conditions:

if cpp.has_argument('-march=haswell')
if cpp.has_argument('-march=skylake-avx512')
if cpp.has_argument('-march=icelake-client')
if cancompilefp16

Also regarding -fPIC it seems to be already included as shown in my error message on the following line:

ninja: job failed: x86_64-linux-musl-c++  -o libx86simdsortcpp.so.1 libx86simdsortcpp.so.1.p/lib_x86simdsort.cpp.o -Wl,--as-needed -Wl,--no-undefined -Wl,-O1 -shared -fPIC -Wl,-soname,libx86simdsortcpp.so.1

when I use CXXFLAGS=-fPIC meson setup --buildtype release builddir it adds it twice in the error message.

[r-devulap]

@gitboy16 yeah looks like your compiler can't build with -march=haswell, which I think is a pretty basic requirement for this library.

[gitboy16]

It does work when I compile a simple code using the templates in lib and -mavx2 so building the library should also work or am I wrong?

[r-devulap]

We use -march=haswell in the meson.build file https://github.com/intel/x86-simd-sort/blob/058f9132b87bb11ab7dd1aa27e07b070c6ba0f4b/lib/meson.build?utm_source=chatgpt.com#L9
which isn’t supported by your compiler. While this could technically be replaced with -mavx2, we prefer -march=haswell since it enables the compiler to apply a broader set of optimizations.

[r-devulap]

@gitboy16 is there a reason your compiler doesn't support -march options? If you would like, in meson.build we could

if x86_64-linux-musl-c++
use -mavx2

Not sure how to do that, but perhaps that is possible.

[gitboy16]

On my computer everything works perfectly, it is not working on github codespaces. It should be fully reproducible on your side. I can give you the steps if you want.
Just to give you a bit of context, I am trying to build the x86-simd-sort library using BinaryBuilder.jl (julia language) in order to make it accessible to Julia users.
I am just trying to spread the amazing work that you did.
Fyi, I am able to use it in Julia on my side, and I also manage to use it alongside google highway sort.

[r-devulap]

Just to give you a bit of context, I am trying to build the x86-simd-sort library using BinaryBuilder.jl (julia language) in order to make it accessible to Julia users.

That's great to hear!

On my computer everything works perfectly, it is not working on github codespaces. It should be fully reproducible on your side. I can give you the steps if you want.

I think I have an easier fix. I will try to submit that as soon as I can.

[gitboy16]

If you look at the details that I include in the first message, you will see the CPU type as well as the instruction set which are supported. What is strange to me is that avx2 wnd avx512 are supported bybthe processor but for some reasons it seems not to be picked up by the build system.

[r-devulap]

What is strange to me is that avx2 wnd avx512 are supported bybthe processor but for some reasons it seems not to be picked up by the build system.

What the build system supports is separate from what the processor supports. As long as the compiler supports it, I can compile for AVX-512F using -mavx512f on a haswell CPU which has just AVX2 on it.

[gitboy16]

In github codespace I am using gcc 14.2, so I think it is supported. I wrote a small program in src folder usimg the template library and manage to compile it. I am not sure why meson compile fails.