[Bug]: Compilation Errors Due to Namespace Conflicts and Missing Type Definitions

[beatlesforever]

What happened?

When compiling the TACOS project, I encountered multiple compilation errors related to namespace conflicts and missing type definitions. The errors primarily indicate that the compiler cannot recognize types such as std::string, std::shared_ptr, and std::vector within the tacos::std namespace. Additionally, there are issues with ambiguous references to the std namespace.

Project Version

• Operating System: Ubuntu 24.04.1 LTS (Noble Numbat)

• Docker Version: Docker version 24.0.7, build 24.0.7-0ubuntu2~22.04.1

• TACOS Version: Commit 00f3873 (Merge pull request Develop #11 from astra-sim/develop)

• Protobuf Version: libprotoc 3.6.1

• Boost Version:

  • libboost-dev: 1.74.0.3ubuntu7
  • libboost-filesystem1.74.0: 1.74.0-14ubuntu3
  • libboost-iostreams1.74.0: 1.74.0-14ubuntu3
  • libboost-locale1.74.0: 1.74.0-14ubuntu3
  • libboost-program-options1.74.0: 1.74.0-14ubuntu3
  • libboost-regex1.74.0: 1.74.0-14ubuntu3
  • libboost-thread1.74.0: 1.74.0-14ubuntu3
  • libboost1.74-dev: 1.74.0-14ubuntu3

Operating System

Linux - Debian-based (e.g., Ubuntu)

How to reproduce the bug

1. Clone the TACOS repository:
   git clone --recurse-submodules https://github.com/astra-sim/tacos.git
2. Pull the TACOS Docker Image:
   docker pull astrasim/tacos:latest
3. Start the Docker Container:
   ./utils/start_docker_container.sh
4. Run TACOS with the provided script:
   [docker] ./tacos.sh

Relevant log output

/app/tacos/include/tacos/writer/xml_writer.h:33:10: error: 'string' in namespace 'tacos::std' does not name a type
   33 |     std::string path;
      |          ^~~~~~
/app/tacos/include/tacos/writer/xml_writer.h:35:10: error: 'shared_ptr' in namespace 'tacos::std' does not name a template type
   35 |     std::shared_ptr<Topology> topology_;
      |          ^~~~~~~~~~
/app/tacos/include/tacos/writer/xml_writer.h:36:10: error: 'shared_ptr' in namespace 'tacos::std' does not name a template type
   36 |     std::shared_ptr<Collective> collective_;
      |          ^~~~~~~~~~
/app/tacos/src/main.cpp: In function 'int main()':
/app/tacos/src/main.cpp:29:9: error: reference to 'std' is ambiguous
   29 |         std::make_shared<Mesh2D>(width, height, latency, bandwidth);
      |         ^~~
/usr/include/c++/13/istream:43:15: note: candidates are: 'namespace tacos::std { }'
   43 | namespace std _GLIBCXX_VISIBILITY(default)
      |               ^~~~~~~~~~~~~~~~~~~
/usr/include/x86_64-linux-gnu/c++/13/bits/c++config.h:306:11: note:                 'namespace std { }'
  306 | namespace std
      |           ^~~
/app/tacos/src/main.cpp:29:32: error: expected primary-expression before '>' token
   29 |         std::make_shared<Mesh2D>(width, height, latency, bandwidth);
      |                                ^
/app/tacos/src/main.cpp:41:9: error: reference to 'std' is ambiguous
   41 |         std::make_shared<AllGather>(npusCount, chunkSize, initChunksPerNpu);
      |         ^~~
/usr/include/c++/13/istream:43:15: note: candidates are: 'namespace tacos::std { }'
   43 | namespace std _GLIBCXX_VISIBILITY(default)
      |               ^~~~~~~~~~~~~~~~~~~
/usr/include/x86_64-linux-gnu/c++/13/bits/c++config.h:306:11: note:                 'namespace std { }'
  306 | namespace std
      |           ^~~
/app/tacos/src/main.cpp:41:35: error: expected primary-expression before '>' token
   41 |         std::make_shared<AllGather>(npusCount, chunkSize, initChunksPerNpu);
      |                                   ^
/app/tacos/src/writer/xml_writer.cpp: In member function 'void tacos::XmlWriter::writeNpu(NpuID)':
/app/tacos/src/writer/xml_writer.cpp:54:40: error: 'collective_' was not declared in this scope
   54 |     npu.append_attribute("o_chunks") = collective_->chunksPerNpu();
      |                                        ^~~~~~~~~~~
/app/tacos/src/writer/xml_writer.cpp:58:64: error: 'class tacos::NpuResult' has no member named 'ingressLinks'
   58 |     for (const auto& [src, link] : synthesisResult_.npu(npuID).ingressLinks()) {
      |                                                                ^~~~~~~~~~~~
/app/tacos/src/writer/xml_writer.cpp:63:65: error: 'class tacos::NpuResult' has no member named 'egressLinks'
   63 |     for (const auto& [dest, link] : synthesisResult_.npu(npuID).egressLinks()) {
      |                                                                 ^~~~~~~~~~~
/app/tacos/src/writer/xml_writer.cpp: In member function 'void tacos::XmlWriter::save()':
/app/tacos/src/writer/xml_writer.cpp:69:23: error: 'path' was not declared in this scope; did you mean 'boost::filesystem::path'?
   69 |     if (xml.save_file(path.c_str(), PUGIXML_TEXT("\t"),
      |                       ^~~~
      |                       boost::filesystem::path
/usr/include/boost/filesystem/path.hpp:155:7: note: 'boost::filesystem::path' declared here
  155 | class path :
      |       ^~~~
/app/tacos/src/writer/xml_writer.cpp: In member function 'void tacos::XmlWriter::writeIngressLink(pugi::xml_node&, NpuID, NpuID, const tacos::LinkResult&)':
/app/tacos/src/writer/xml_writer.cpp:88:40: error: 'const class tacos::LinkResult' has no member named 'ops'
   88 |     for (const auto& [opID, op] : link.ops()) {
      |                                        ^~~
/app/tacos/src/writer/xml_writer.cpp: In member function 'void tacos::XmlWriter::writeEgressLink(pugi::xml_node&, NpuID, NpuID, const tacos::LinkResult&)':
/app/tacos/src/writer/xml_writer.cpp:117:40: error: 'const class tacos::LinkResult' has no member named 'ops'
  117 |     for (const auto& [opID, op] : link.ops()) {
      |                                        ^~~
gmake[2]: *** [CMakeFiles/tacos.dir/build.make:300: CMakeFiles/tacos.dir/src/writer/xml_writer.cpp.o] Error 1
gmake[2]: *** [CMakeFiles/tacos.dir/build.make:314: CMakeFiles/tacos.dir/src/main.cpp.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:189: CMakeFiles/tacos.dir/all] Error 2
gmake: *** [Makefile:136: all] Error 2

Miscellaneous

No response

[linsj20]

Hi @beatlesforever, can you apply this patch (#14) to the current codebase and try again? It seems there is a namespace error that causes all the mess.

[beatlesforever]

Hi @beatlesforever, can you apply this patch (#14) to the current codebase and try again? It seems there is a namespace error that causes all the mess.

Hi @linsj20,
Thank you very much! I will try again later.
By the way, did you manage to compile successfully after applying this patch?

[linsj20]

Yes, I've successfully compiled both in container and in my own environment.

[beatlesforever]

Yes, I've successfully compiled both in container and in my own environment.

Wonderful! Thanks a ton, @linsj20. Knowing that you've had success gives me a lot of confidence. I really appreciate you taking the time to let me know. I'll get on it right away.

[beatlesforever]

Hi @linsj20, @willjwon

Thank you again for your help! I successfully ran the example after applying the patch and obtained the following results:

[2025-03-08 09:35:06.151] [TACOS] [info] TACOS
[2025-03-08 09:35:06.151] [TACOS] [info] Log file path: /app/tacos/results/tacos.log
[2025-03-08 09:35:06.151] [TACOS] [info] 
[2025-03-08 09:35:06.151] [TACOS] [info] Topology Information
[2025-03-08 09:35:06.151] [TACOS] [info] 	- NPUs Count: 9
[2025-03-08 09:35:06.151] [TACOS] [info] 
[2025-03-08 09:35:06.151] [TACOS] [info] Collective Information
[2025-03-08 09:35:06.151] [TACOS] [info] 	- Chunks Count: 9
[2025-03-08 09:35:06.151] [TACOS] [info] 	- Chunk Size: 1 MB
[2025-03-08 09:35:06.151] [TACOS] [info] 
[2025-03-08 09:35:06.151] [TACOS] [info] Synthesis Process
[2025-03-08 09:35:06.152] [TACOS] [info] 
[2025-03-08 09:35:06.152] [TACOS] [info] Synthesis Result
[2025-03-08 09:35:06.152] [TACOS] [info] 	- Time to solve: 164.91 us (0.00 s)
[2025-03-08 09:35:06.152] [TACOS] [info] 	- Synthesized Collective Time: 80125000 ps (80.12 us)
[2025-03-08 09:35:06.152] [TACOS] [info] 
[2025-03-08 09:35:06.152] [TACOS] [info] XML file written at: /app/tacos/results/tacos.xml
[2025-03-08 09:35:06.152] [TACOS] [info] Done!

One question I have is about the synthesized collective time calculation. From my understanding, this timing seems to be computed independently within TACOS, without directly relying on Astra-Sim. However, the TACOS paper mentions that experiments were conducted using Astra-Sim.
So, does this mean that I need to take the synthesized algorithm from TACOS and input it into Astra-Sim for further evaluation? Specifically, should I convert tacos.xml into a format that Astra-Sim accepts in order to properly simulate the synthesized collective algorithm within Astra-Sim?
Looking forward to your insights. Thanks again!