omnitrace-run executable - required for running binary writes (#257)

* omnitrace-run exe

- ensure LD_PRELOAD for libomnitrace-dl.so
- convert config options into command-line options

* Update timemory submodule

- updates to tsettings
- updates to argparser

* common environment update

- throw error if get_env<bool> has empty string

* config updates

- minor tweaks to categories of settings

* core lib update

- add argparse for common handling of argument parsers

* omnitrace-sample update

- fix handling of --trace-file (OMNITRACE_PERFETTO_FILE)

* omnitrace-run update

- updated to use omnitrace::argparse functions

* Tests for omnitrace-run

* argparse core update

- remove choices for --cpu-events and --gpu-events

* remove some debugging prints

* fix timemory include in argparse.cpp

* always provide --hsa-interrupt option

* Update source/lib/core/argparse.cpp

- fix pedantic warning

* Update testing

- remove testing args that may not be there in some builds

* roctracer/pthread_create fix

- disable roctracer_data when roctracer not enabled

* omnitrace-causal tweak

* omnitrace-instrument: module_function tweak

- allow DEFAULT_MODULE and LIBRARY_MODULE

* common environment update

- support get_env for enums

* core: config update

- Add "mode" category to OMNITRACE_MODE

* Update timemory submodule

- remove debug print statement

* omnitrace-sample tweak

- change var init

* omnitrace-run testing update

- use --help instead of -?

* core: common.hpp

- tweak header include style

* core: argparser update

- add_ld_preload func
- launcher and command member variables in parser_data
- support launcher

* omnitrace-run update

- clean up and reworked

* libomnitrace-dl updates

- require LD_PRELOAD with binary rewrite
- dl::InstrumentMode
- dl::get_instrumented()
- verify_instrumented_preloaded()
- omnitrace_set_instrumented(int)
- relocated omnitrace_main from main.c to dl.cpp
- omnitrace_set_env does not dlopen libomnitrace
- omnitrace_set_main(func_ptr) [internal API]
- OMNITRACE_HIDDEN_API -> OMNITRACE_INTERNAL_API

* Update testing to new LD_PRELOAD requirements

* omnitrace-instrument updates

- adhere to LD_PRELOAD requirementsa
- invoke omnitrace_set_instrumented
- binary rewrite does not instrument main
- binary rewrite does not instrument call to omnitrace_init
- runtime instr does not instrument main
- runtime instr does not instrument call to omnitrace_init

* Bump to v1.9.0

- LD_PRELOAD requirement necessitates minor version increment

* common: environment

- fix ambiguous get_env calls

* omnitrace-instrument update

- fix issue with temporaries

* omnitrace-instrument and libomnitrace-dl updates

- runtime instrumentation does not work if libomnitrace-dl is preloaded

* libomnitrace-dl and libpyomnitrace updates

- define dl::InstrumentMode in dl.hpp
- handle instrumentation via setprofile libpyomnitrace
  - do not push trace in omnitrace_init

* omnitrace-instrument and libomnitrace-dl updates

- move header to dl subdirectory
- omnitrace::omnitrace-headers include omnitrace-dl folder
- use InstrumentMode in omnitrace-instrument

* Update workflows and scripts

- Use omnitrace-run on instrumented exes

* Update docs

- add omnitrace-run to examples of running binary rewritten exes

[ROCm/rocprofiler-systems commit: abe35de43a]

projects/rocprofiler-systems/source/docs/getting_started.md

@@ -97,7 +97,7 @@ e.g., omnitrace's meaning of the term "module" when instrumenting Python.
   - Binary rewrites only instrument the functions defined in the target binary, whereas runtime instrumentation can/will instrument functions defined in the shared libraries which are linked into the target binary
 - When using binary instrumentation with MPI, avoid runtime instrumentation
   - Runtime instrumentation requires a fork + ptrace: which is generally incompatible with how MPI applications spawn their processes
-  - Binary rewrite the executable using MPI (and, optionally, libraries used by the executable) and execute the generated instrumented executable instead of the original, e.g. `mpirun -n 2 ./myexe` should be `mpirun -n 2 ./myexe.inst` where `myexe.inst` is the generated instrumented `myexe` executable.
+  - Binary rewrite the executable using MPI (and, optionally, libraries used by the executable) and execute the generated instrumented executable via `omnitrace-run` instead of the original, e.g. `mpirun -n 2 ./myexe` should be `mpirun -n 2 omnitrace-run -- ./myexe.inst` where `myexe.inst` is the generated instrumented `myexe` executable.
 
 ## Data Collection Mode(s)
 

projects/rocprofiler-systems/source/docs/instrumenting.md

@@ -335,7 +335,7 @@ is consistent with the LLVM style of using a standalone double-hyphen (`--`). Al
 are interpreted as belonging to omnitrace and all arguments following the double-hyphen are interpreted as the
 application and it's arguments. In binary rewrite mode, all application arguments after the first argument
 are ignored, i.e. `./omnitrace-instrument -o ls.inst -- ls -l` interprets `ls` as the target to instrument (ignores the `-l` argument)
-and generates a `ls.inst` executable that you can subsequently run `ls.inst -l` with.
+and generates a `ls.inst` executable that you can subsequently run `omnitrace-run -- ls.inst -l` with.
 
 ## Runtime Instrumentation
 
@@ -781,7 +781,7 @@ For example, if the following sequence of commands are run:
 omnitrace-instrument -o ./foo.inst -- ./foo
 export OMNITRACE_USE_SAMPLING=ON
 export OMNITRACE_SAMPLING_FREQ=5
-./foo.inst
+omnitrace-run -- ./foo.inst
 ```
 
 These configuration settings will not be preserved in another session, whereas:
@@ -794,7 +794,7 @@ will preserve those environment variables:
 
 ```shell
 # will sample 5x per second
-./foo.samp
+omnitrace-run -- ./foo.samp
 ```
 
 while still allowing the subsequent session to override those defaults:
@@ -802,7 +802,7 @@ while still allowing the subsequent session to override those defaults:
 ```shell
 # will sample 100x per second
 export OMNITRACE_SAMPLING_FREQ=100
-./foo.samp
+omnitrace-run -- ./foo.samp
 ```
 
 ### Troubleshooting

projects/rocprofiler-systems/source/docs/sampling.md

@@ -15,7 +15,7 @@ Call-stack sampling can be activated with either a binary instrumented via the `
 
 ```console
 omnitrace-instrument -M sampling -o foo.inst -- foo
-./foo.inst
+omnitrace-run -- ./foo.inst
 ```
 
 - Runtime instrumentation with only instrumentation necessary to start/stop sampling

projects/rocprofiler-systems/source/docs/user_api.md

@@ -169,11 +169,7 @@ custom_push_region(const char* name)
 ```console
 $ omnitrace-instrument -l --min-instructions=8 -E custom_push_region -o -- ./user-api
 ...
-$ export OMNITRACE_USE_TIMEMORY=ON
-$ export OMNITRACE_USE_PID=OFF
-$ export OMNITRACE_TIME_OUTPUT=OFF
-$ export OMNITRACE_OUTPUT_PATH=omnitrace-example-output
-$ ./user-api.inst 20 4 100
+$ omnitrace-run --profile --use-pid off --time-output off -- ./user-api.inst 20 4 100
 Pushing custom region :: ./user-api.inst
 [omnitrace][omnitrace_init_tooling] Instrumentation mode: Trace