custom-allocators
Custom allocator skill for memory allocation strategies. Use when implementing…
GCC compiler skill for C/C++ projects. Use when selecting optimization levels, warning flags, debug builds, LTO, sanitizer instrumentation, or diagnosing compilation errors with GCC. Covers flag selection for debug vs release, ABI concerns, preprocessor macros, profile-guided
$ npx -y skills add mohitmishra786/low-level-dev-skills --skill gcc --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/gccContext preview
The summary Claude sees to decide when to auto-load this skill.
GCC compiler skill for C/C++ projects. Use when selecting optimization levels, warning flags, debug builds, LTO, sanitizer instrumentation, or diagnosing compilation errors with GCC. Covers flag selection for debug vs release, ABI concerns, preprocessor macros, profile-guided
name: gcc description: GCC compiler skill for C/C++ projects. Use when selecting optimization levels, warning flags, debug builds, LTO, sanitizer instrumentation, or diagnosing compilation errors with GCC. Covers flag selection for debug vs release, ABI concerns, preprocessor macros, profile-guided optimization, and integration with build systems. Activates on queries about gcc flags, compilation errors, performance tuning, warning suppression, or cross-standard compilation.
Guide agents through GCC invocation: flag selection, build modes, warning triage, PGO, LTO, and common error patterns. Assume the project uses GNU Make, CMake, or a shell script.
| Goal | Recommended flags | |------|-------------------| | Debug | `-g -O0 -Wall -Wextra` | | Debug + debuggable optimisation | `-g -Og -Wall -Wextra` | | Release | `-O2 -DNDEBUG -Wall` | | Release (max perf, native only) | `-O3 -march=native -DNDEBUG` | | Release (min size) | `-Os -DNDEBUG` | | Sanitizer (dev) | `-g -O1 -fsanitize=address,undefined` |
Always pass `-std=c11` / `-std=c++17` (or the required standard) explicitly. Never rely on the implicit default.
Start with `-Wall -Wextra`. For stricter standards compliance add `-Wpedantic`. To treat all warnings as errors in CI: `-Werror`.
Suppress a specific warning only in a narrow scope:
#pragma GCC diagnostic push #pragma GCC diagnostic ignored "-Wunused-parameter" // ... #pragma GCC diagnostic pop
Do not pass `-w` (silences everything) except as a last resort for third-party headers.
Pair `-g` with `-Og` (not `-O0`) when you need readable optimised code in GDB.
Need max throughput on a fixed machine?
yes -> -O3 -march=native -flto
no -> profiling available?
yes -> -O2 -fprofile-use
no -> -O2
Size-constrained (embedded, shared lib)?
yes -> -Os (or -Oz with clang)`-O3` vs `-O2`: `-O3` adds aggressive loop transformations (`-funswitch-loops`, `-fpeel-loops`, `-floop-interchange`) and more aggressive inlining. Use `-O3` only after benchmarking; it occasionally regresses due to i-cache pressure.
`-Ofast`: enables `-ffast-math` which breaks IEEE 754 semantics (NaN handling, associativity). Avoid unless the numerical domain explicitly permits it.
# Compile gcc -O2 -flto -c foo.c -o foo.o gcc -O2 -flto -c bar.c -o bar.o # Link (must pass -flto again) gcc -O2 -flto foo.o bar.o -o prog
Use `gcc-ar` / `gcc-ranlib` instead of `ar` / `ranlib` when archiving LTO objects into static libs.
For parallel LTO: `-flto=auto` (uses `make`-style jobserver) or `-flto=N`.
See [references/flags.md](references/flags.md) for full flag reference. See `skills/binaries/linkers-lto` for linker-level LTO configuration.
# Step 1: instrument gcc -O2 -fprofile-generate prog.c -o prog_inst # Step 2: run with representative workload ./prog_inst < workload.input # Step 3: optimise with profile gcc -O2 -fprofile-use -fprofile-correction prog.c -o prog
`-fprofile-correction` handles profile data inconsistencies from multi-threaded runs.
| Symptom | Likely cause | Fix | |---------|-------------|-----| | `undefined reference to 'foo'` | Missing `-lfoo` or wrong link order | Add `-lfoo`; move `-l` flags after object files | | `multiple definition of 'x'` | Variable defined (not just declared) in a header | Add `extern` in header, define in one `.c` | | `implicit declaration of function` | Missing `#include` | Add the header | | `warning: incompatible pointer types` | Wrong cast or missing prototype | Fix the type; check headers | | ABI errors with C++ | Mixed `-std=` or different `libstdc++` | Unify `-std=` across all TUs | | `relocation truncated` | Overflow on a 32-bit relative relocation | Use `-mcmodel=large` or restructure code |
For sanitizer reports, use `skills/runtimes/sanitizers`.
# Show all flags enabled at -O2 gcc -Q --help=optimizers -O2 | grep enabled # Preprocess only (check includes/macros) gcc -E -dD src.c -o src.i # Assembly output (Intel syntax) gcc -S -masm=intel -O2 foo.c -o foo.s # Show include search path gcc -v -E - < /dev/null 2>&1 | grep -A20 '#include <...>' # Check if a flag is supported gcc -Q --help=target | grep march
For a complete flag cheatsheet, see [references/flags.md](references/flags.md). For common error patterns and examples, see [references/examples.md](references/examples.md).
A curated suite of AI agent skills for systems and low-level programming — C/C++, Rust, Zig, GPU, bare-metal firmware, Linux kernel/driver development, computer architecture, compiler internals, HPC, and more.
Repo: mohitmishra786/low-level-dev-skills
Custom allocator skill for memory allocation strategies. Use when implementing…
NUMA programming skill for multi-socket memory locality. Use when detecting NUMA topology,…
AF_XDP skill for high-performance XDP sockets. Use when creating AF_XDP sockets, configuring…
DPDK skill for userspace packet I/O. Use when initializing EAL, configuring PMD drivers,…
io_uring skill for Linux async I/O. Use when building high-performance servers with liburing,…
Bare-metal ADC and DAC skill. Use when configuring analog sampling, DMA-driven ADC,…