clang

Name: clang
Author: mohitmishra786

368installs19forks

How to install clang?

mohitmishra786/low-level-dev-skills/clang

$npx -y skills add mohitmishra786/low-level-dev-skills --skill clang

Installs into the current project.

Summary

Clang/LLVM compiler skill for C/C++ projects. Use when working with clang or clang++ for diagnostics, sanitizer instrumentation, optimization remarks, static analysis with clang-tidy, LTO via lld, or when migrating from GCC to Clang. Activates on queries about clang flags, clang-

Files · 1

View on GitHub

SKILL.md

1	# Clang
2
3	## Purpose
4
5	Guide agents through Clang-specific features: superior diagnostics, sanitizer integration, optimization remarks, static analysis, and LLVM tooling. Covers divergences from GCC and Apple/FreeBSD specifics.
6
7	## Triggers
8
9	- "I want better compiler diagnostics/errors"
10	- "How do I use clang-tidy / clang-format?"
11	- "How do I see what the compiler optimised or didn't?"
12	- "I'm on macOS / FreeBSD using clang"
13	- "clang-cl for MSVC-compatible builds" — see `skills/compilers/msvc-cl`
14	- Sanitizer queries — see `skills/runtimes/sanitizers`
15
16	## Workflow
17
18	### 1. Build mode flags (identical to GCC)
19
20	Clang accepts most GCC flags. Key differences:
21
22	\| Feature \| GCC \| Clang \|
23	\|---------\|-----\|-------\|
24	\| Min size \| `-Os` \| `-Os` or `-Oz` (more aggressive) \|
25	\| Optimise only hot \| — \| `-fprofile-instr-use` (LLVM PGO) \|
26	\| Thin LTO \| `-flto` \| `-flto=thin` (faster) \|
27	\| Static analyser \| `-fanalyzer` \| `clang --analyze` or `clang-tidy` \|
28
29	### 2. Clang-specific diagnostic flags
30
31	```bash
32	# Show fix-it hints inline
33	clang -Wall -Wextra --show-fixits src.c
34
35	# Limit error count
36	clang -ferror-limit=5 src.c
37
38	# Verbose template errors (disable elision)
39	clang -fno-elide-type src.cpp
40
41	# Show tree diff for template mismatch
42	clang -fdiagnostics-show-template-tree src.cpp
43	```
44
45	Clang's diagnostics include exact range highlighting and fix-it suggestions that GCC lacks.
46
47	### 3. Optimization remarks
48
49	Optimization remarks let you see what Clang did or refused to do:
50
51	```bash
52	# Inliner decisions
53	clang -O2 -Rpass=inline src.c
54
55	# Missed vectorisation
56	clang -O2 -Rpass-missed=loop-vectorize src.c
57
58	# Why a loop was not vectorized
59	clang -O2 -Rpass-analysis=loop-vectorize src.c
60
61	# Save all remarks to YAML for post-processing
62	clang -O2 -fsave-optimization-record src.c
63	# Produces src.opt.yaml
64	```
65
66	Interpret remarks:
67
68	- `remark: foo inlined into bar` — inlining happened; good for hot paths
69	- `remark: loop not vectorized: loop control flow is not understood` — restructure the loop
70	- `remark: not vectorized: cannot prove it is safe to reorder...` — add `__restrict__` or `#pragma clang loop vectorize(assume_safety)`
71
72	### 4. Static analysis
73
74	```bash
75	# Built-in analyser (CSA)
76	clang --analyze -Xanalyzer -analyzer-output=text src.c
77
78	# clang-tidy (separate tool, richer checks)
79	clang-tidy src.c -- -std=c++17 -I/usr/include
80
81	# Enable specific check families
82	clang-tidy -checks='clang-analyzer-,modernize-,bugprone-*' src.cpp --
83
84	# Apply fixits automatically
85	clang-tidy -fix src.cpp --
86	```
87
88	Common `clang-tidy` check families:
89
90	- `bugprone-*`: real bugs (use-after-move, dangling, etc.)
91	- `clang-analyzer-*`: CSA checks (memory, null deref)
92	- `modernize-*`: C++11/14/17 modernisation
93	- `performance-*`: unnecessary copies, move candidates
94	- `readability-*`: naming, complexity
95
96	### 5. LTO with lld
97
98	```bash
99	# Full LTO
100	clang -O2 -flto -fuse-ld=lld src.c -o prog
101
102	# Thin LTO (faster link, nearly same quality)
103	clang -O2 -flto=thin -fuse-ld=lld src.c -o prog
104
105	# Check lld is available
106	clang -fuse-ld=lld -Wl,--version 2>&1 \| head -1
107	```
108
109	For large projects, ThinLTO is preferred: link times 5-10x faster than full LTO with comparable code quality.
110
111	### 6. PGO (LLVM instrumentation)
112
113	```bash
114	# Step 1: instrument
115	clang -O2 -fprofile-instr-generate prog.c -o prog_inst
116
117	# Step 2: run with representative input
118	./prog_inst < workload.input
119	# Generates default.profraw
120
121	# Step 3: merge profiles
122	llvm-profdata merge -output=prog.profdata default.profraw
123
124	# Step 4: use profile
125	clang -O2 -fprofile-instr-use=prog.profdata prog.c -o prog
126	```
127
128	AutoFDO (sampling-based, less intrusive): collect with `perf`, convert with `create_llvm_prof`, use with `-fprofile-sample-use`. See `skills/profilers/linux-perf`.
129
130	### 7. GCC compatibility
131
132	Clang is intentionally GCC-compatible for driver flags. Key differences:
133
134	- Clang does not support all GCC-specific attributes; check with `__has_attribute(foo)`
135	- `-Weverything` enables all Clang warnings (no GCC equivalent); too noisy for production, useful for one-off audits
136	- Some GCC intrinsics need `#include <x86intrin.h>` on Clang too
137	- `__int128` is supported; `__float128` requires `-lquadmath` on some targets
138
139	### 8. macOS specifics
140
141	On macOS, `clang` is the system compiler (Apple LLVM). Key points:
142
143	- `ld64` is the default linker; `lld` requires explicit `-fuse-ld=lld` and Homebrew LLVM
144	- Use `-mmacosx-version-min=X.Y` to set deployment target
145	- Sanitizers on macOS use `DYLD_INSERT_LIBRARIES`; do not strip the binary
146	- `xcrun clang` resolves to the Xcode toolchain c

People who liked this also viewed

6 picks

Type

clang

368installs19forks

View on GitHub

How to install clang?

mohitmishra786/low-level-dev-skills/clang

$npx -y skills add mohitmishra786/low-level-dev-skills --skill clang

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Clang
2
3	## Purpose
4
5	Guide agents through Clang-specific features: superior diagnostics, sanitizer integration, optimization remarks, static analysis, and LLVM tooling. Covers divergences from GCC and Apple/FreeBSD specifics.
6
7	## Triggers
8
9	- "I want better compiler diagnostics/errors"
10	- "How do I use clang-tidy / clang-format?"
11	- "How do I see what the compiler optimised or didn't?"
12	- "I'm on macOS / FreeBSD using clang"
13	- "clang-cl for MSVC-compatible builds" — see `skills/compilers/msvc-cl`
14	- Sanitizer queries — see `skills/runtimes/sanitizers`
15
16	## Workflow
17
18	### 1. Build mode flags (identical to GCC)
19
20	Clang accepts most GCC flags. Key differences:
21
22	\| Feature \| GCC \| Clang \|
23	\|---------\|-----\|-------\|
24	\| Min size \| `-Os` \| `-Os` or `-Oz` (more aggressive) \|
25	\| Optimise only hot \| — \| `-fprofile-instr-use` (LLVM PGO) \|
26	\| Thin LTO \| `-flto` \| `-flto=thin` (faster) \|
27	\| Static analyser \| `-fanalyzer` \| `clang --analyze` or `clang-tidy` \|
28
29	### 2. Clang-specific diagnostic flags
30
31	```bash
32	# Show fix-it hints inline
33	clang -Wall -Wextra --show-fixits src.c
34
35	# Limit error count
36	clang -ferror-limit=5 src.c
37
38	# Verbose template errors (disable elision)
39	clang -fno-elide-type src.cpp
40
41	# Show tree diff for template mismatch
42	clang -fdiagnostics-show-template-tree src.cpp
43	```
44
45	Clang's diagnostics include exact range highlighting and fix-it suggestions that GCC lacks.
46
47	### 3. Optimization remarks
48
49	Optimization remarks let you see what Clang did or refused to do:
50
51	```bash
52	# Inliner decisions
53	clang -O2 -Rpass=inline src.c
54
55	# Missed vectorisation
56	clang -O2 -Rpass-missed=loop-vectorize src.c
57
58	# Why a loop was not vectorized
59	clang -O2 -Rpass-analysis=loop-vectorize src.c
60
61	# Save all remarks to YAML for post-processing
62	clang -O2 -fsave-optimization-record src.c
63	# Produces src.opt.yaml
64	```
65
66	Interpret remarks:
67
68	- `remark: foo inlined into bar` — inlining happened; good for hot paths
69	- `remark: loop not vectorized: loop control flow is not understood` — restructure the loop
70	- `remark: not vectorized: cannot prove it is safe to reorder...` — add `__restrict__` or `#pragma clang loop vectorize(assume_safety)`
71
72	### 4. Static analysis
73
74	```bash
75	# Built-in analyser (CSA)
76	clang --analyze -Xanalyzer -analyzer-output=text src.c
77
78	# clang-tidy (separate tool, richer checks)
79	clang-tidy src.c -- -std=c++17 -I/usr/include
80
81	# Enable specific check families
82	clang-tidy -checks='clang-analyzer-,modernize-,bugprone-*' src.cpp --
83
84	# Apply fixits automatically
85	clang-tidy -fix src.cpp --
86	```
87
88	Common `clang-tidy` check families:
89
90	- `bugprone-*`: real bugs (use-after-move, dangling, etc.)
91	- `clang-analyzer-*`: CSA checks (memory, null deref)
92	- `modernize-*`: C++11/14/17 modernisation
93	- `performance-*`: unnecessary copies, move candidates
94	- `readability-*`: naming, complexity
95
96	### 5. LTO with lld
97
98	```bash
99	# Full LTO
100	clang -O2 -flto -fuse-ld=lld src.c -o prog
101
102	# Thin LTO (faster link, nearly same quality)
103	clang -O2 -flto=thin -fuse-ld=lld src.c -o prog
104
105	# Check lld is available
106	clang -fuse-ld=lld -Wl,--version 2>&1 \| head -1
107	```
108
109	For large projects, ThinLTO is preferred: link times 5-10x faster than full LTO with comparable code quality.
110
111	### 6. PGO (LLVM instrumentation)
112
113	```bash
114	# Step 1: instrument
115	clang -O2 -fprofile-instr-generate prog.c -o prog_inst
116
117	# Step 2: run with representative input
118	./prog_inst < workload.input
119	# Generates default.profraw
120
121	# Step 3: merge profiles
122	llvm-profdata merge -output=prog.profdata default.profraw
123
124	# Step 4: use profile
125	clang -O2 -fprofile-instr-use=prog.profdata prog.c -o prog
126	```
127
128	AutoFDO (sampling-based, less intrusive): collect with `perf`, convert with `create_llvm_prof`, use with `-fprofile-sample-use`. See `skills/profilers/linux-perf`.
129
130	### 7. GCC compatibility
131
132	Clang is intentionally GCC-compatible for driver flags. Key differences:
133
134	- Clang does not support all GCC-specific attributes; check with `__has_attribute(foo)`
135	- `-Weverything` enables all Clang warnings (no GCC equivalent); too noisy for production, useful for one-off audits
136	- Some GCC intrinsics need `#include <x86intrin.h>` on Clang too
137	- `__int128` is supported; `__float128` requires `-lquadmath` on some targets
138
139	### 8. macOS specifics
140
141	On macOS, `clang` is the system compiler (Apple LLVM). Key points:
142
143	- `ld64` is the default linker; `lld` requires explicit `-fuse-ld=lld` and Homebrew LLVM
144	- Use `-mmacosx-version-min=X.Y` to set deployment target
145	- Sanitizers on macOS use `DYLD_INSERT_LIBRARIES`; do not strip the binary
146	- `xcrun clang` resolves to the Xcode toolchain c

People who liked this also viewed

6 picks

Type