llvm

Name: llvm
Author: mohitmishra786

600installs19forks

How to install llvm?

mohitmishra786/low-level-dev-skills/llvm

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

Installs into the current project.

Summary

LLVM IR and pass pipeline skill. Use when working directly with LLVM Intermediate Representation (IR), running opt passes, generating IR with llc, inspecting or writing LLVM IR for custom passes, or understanding how the LLVM backend lowers IR to assembly. Activates on queries ab

Files · 1

View on GitHub

SKILL.md

1	# LLVM IR and Tooling
2
3	## Purpose
4
5	Guide agents through LLVM as a user: generating and inspecting IR, running existing optimisation passes with `opt`, lowering to assembly with `llc`, and diagnosing missed optimisations. For writing new LLVM passes (PassPlugin, llvm-lit testing), use `skills/compiler-internals/llvm-passes` instead.
6
7	## Triggers
8
9	- "Show me the LLVM IR for this function"
10	- "How do I run an LLVM optimisation pass?"
11	- "What does this LLVM IR instruction mean?"
12	- "How do I write a custom LLVM pass?"
13	- "Why isn't auto-vectorisation happening in LLVM?"
14
15	## Workflow
16
17	### 1. Generate LLVM IR
18
19	```bash
20	# Emit textual IR (.ll)
21	clang -O0 -emit-llvm -S src.c -o src.ll
22
23	# Emit bitcode (.bc)
24	clang -O2 -emit-llvm -c src.c -o src.bc
25
26	# Disassemble bitcode to text
27	llvm-dis src.bc -o src.ll
28	```
29
30	### 2. Run optimisation passes with `opt`
31
32	```bash
33	# Apply a specific pass
34	opt -passes='mem2reg,instcombine,simplifycfg' src.ll -S -o out.ll
35
36	# Standard optimisation pipelines
37	opt -passes='default<O2>' src.ll -S -o out.ll
38	opt -passes='default<O3>' src.ll -S -o out.ll
39
40	# List available passes
41	opt --print-passes 2>&1 \| less
42
43	# Print IR before and after a pass
44	opt -passes='instcombine' --print-before=instcombine --print-after=instcombine src.ll -S -o out.ll 2>&1 \| less
45	```
46
47	### 3. Lower IR to assembly with `llc`
48
49	```bash
50	# Compile IR to object file
51	llc -filetype=obj src.ll -o src.o
52
53	# Compile to assembly
54	llc -filetype=asm -masm-syntax=intel src.ll -o src.s
55
56	# Target a specific CPU
57	llc -mcpu=skylake -mattr=+avx2 src.ll -o src.s
58
59	# Show available targets
60	llc --version
61	```
62
63	### 4. Inspect IR
64
65	Key IR constructs to understand:
66
67	\| Construct \| Meaning \|
68	\|-----------\|---------\|
69	\| `alloca` \| Stack allocation (pre-SSA; `mem2reg` promotes to registers) \|
70	\| `load`/`store` \| Memory access \|
71	\| `getelementptr` (GEP) \| Pointer arithmetic / field access \|
72	\| `phi` \| SSA φ-node: merges values from predecessor blocks \|
73	\| `call`/`invoke` \| Function call (`invoke` has exception edges) \|
74	\| `icmp`/`fcmp` \| Integer/float comparison \|
75	\| `br` \| Branch (conditional or unconditional) \|
76	\| `ret` \| Return \|
77	\| `bitcast` \| Reinterpret bits (no-op in codegen) \|
78	\| `ptrtoint`/`inttoptr` \| Pointer↔integer (avoid where possible) \|
79
80	### 5. Key passes
81
82	\| Pass \| Effect \|
83	\|------\|--------\|
84	\| `mem2reg` \| Promote alloca to SSA registers \|
85	\| `instcombine` \| Instruction combining / peephole \|
86	\| `simplifycfg` \| CFG cleanup, dead block removal \|
87	\| `loop-vectorize` \| Auto-vectorisation \|
88	\| `slp-vectorize` \| Superword-level parallelism (straight-line vectorisation) \|
89	\| `inline` \| Function inlining \|
90	\| `gvn` \| Global value numbering (common subexpression elimination) \|
91	\| `licm` \| Loop-invariant code motion \|
92	\| `loop-unroll` \| Loop unrolling \|
93	\| `argpromotion` \| Promote pointer args to values \|
94	\| `sroa` \| Scalar Replacement of Aggregates \|
95
96	### 6. Debugging missed optimisations
97
98	```bash
99	# Why was a loop not vectorised?
100	clang -O2 -Rpass-missed=loop-vectorize -Rpass-analysis=loop-vectorize src.c
101
102	# Dump pass pipeline
103	clang -O2 -mllvm -debug-pass=Structure src.c -o /dev/null 2>&1 \| less
104
105	# Print IR after each pass (very verbose)
106	opt -passes='default<O2>' -print-after-all src.ll -S 2>&1 \| less
107	```
108
109	### 7. Useful llvm tools
110
111	\| Tool \| Purpose \|
112	\|------\|---------\|
113	\| `llvm-dis` \| Bitcode → textual IR \|
114	\| `llvm-as` \| Textual IR → bitcode \|
115	\| `llvm-link` \| Link multiple bitcode files \|
116	\| `llvm-lto` \| Standalone LTO \|
117	\| `llvm-nm` \| Symbols in bitcode/object \|
118	\| `llvm-objdump` \| Disassemble objects \|
119	\| `llvm-profdata` \| Merge/show PGO profiles \|
120	\| `llvm-cov` \| Coverage reporting \|
121	\| `llvm-mca` \| Machine code analyser (throughput/latency) \|
122
123	For binutils equivalents, see `skills/binaries/binutils`.
124
125	## Related skills
126
127	- Use `skills/compiler-internals/llvm-passes` for writing and testing custom LLVM passes
128	- Use `skills/compiler-internals/compiler-frontend` for generating LLVM IR from an AST
129	- Use `skills/compiler-internals/jit-compilation` for ORC JIT execution of LLVM IR
130	- Use `skills/compilers/clang` for source-level Clang flags
131	- Use `skills/binaries/linkers-lto` for LTO at link time
132	- Use `skills/profilers/linux-perf` combined with `llvm-mca` for micro-architectural analysis

People who liked this also viewed

6 picks

Type

llvm

600installs19forks

View on GitHub

How to install llvm?

mohitmishra786/low-level-dev-skills/llvm

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

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# LLVM IR and Tooling
2
3	## Purpose
4
5	Guide agents through LLVM as a user: generating and inspecting IR, running existing optimisation passes with `opt`, lowering to assembly with `llc`, and diagnosing missed optimisations. For writing new LLVM passes (PassPlugin, llvm-lit testing), use `skills/compiler-internals/llvm-passes` instead.
6
7	## Triggers
8
9	- "Show me the LLVM IR for this function"
10	- "How do I run an LLVM optimisation pass?"
11	- "What does this LLVM IR instruction mean?"
12	- "How do I write a custom LLVM pass?"
13	- "Why isn't auto-vectorisation happening in LLVM?"
14
15	## Workflow
16
17	### 1. Generate LLVM IR
18
19	```bash
20	# Emit textual IR (.ll)
21	clang -O0 -emit-llvm -S src.c -o src.ll
22
23	# Emit bitcode (.bc)
24	clang -O2 -emit-llvm -c src.c -o src.bc
25
26	# Disassemble bitcode to text
27	llvm-dis src.bc -o src.ll
28	```
29
30	### 2. Run optimisation passes with `opt`
31
32	```bash
33	# Apply a specific pass
34	opt -passes='mem2reg,instcombine,simplifycfg' src.ll -S -o out.ll
35
36	# Standard optimisation pipelines
37	opt -passes='default<O2>' src.ll -S -o out.ll
38	opt -passes='default<O3>' src.ll -S -o out.ll
39
40	# List available passes
41	opt --print-passes 2>&1 \| less
42
43	# Print IR before and after a pass
44	opt -passes='instcombine' --print-before=instcombine --print-after=instcombine src.ll -S -o out.ll 2>&1 \| less
45	```
46
47	### 3. Lower IR to assembly with `llc`
48
49	```bash
50	# Compile IR to object file
51	llc -filetype=obj src.ll -o src.o
52
53	# Compile to assembly
54	llc -filetype=asm -masm-syntax=intel src.ll -o src.s
55
56	# Target a specific CPU
57	llc -mcpu=skylake -mattr=+avx2 src.ll -o src.s
58
59	# Show available targets
60	llc --version
61	```
62
63	### 4. Inspect IR
64
65	Key IR constructs to understand:
66
67	\| Construct \| Meaning \|
68	\|-----------\|---------\|
69	\| `alloca` \| Stack allocation (pre-SSA; `mem2reg` promotes to registers) \|
70	\| `load`/`store` \| Memory access \|
71	\| `getelementptr` (GEP) \| Pointer arithmetic / field access \|
72	\| `phi` \| SSA φ-node: merges values from predecessor blocks \|
73	\| `call`/`invoke` \| Function call (`invoke` has exception edges) \|
74	\| `icmp`/`fcmp` \| Integer/float comparison \|
75	\| `br` \| Branch (conditional or unconditional) \|
76	\| `ret` \| Return \|
77	\| `bitcast` \| Reinterpret bits (no-op in codegen) \|
78	\| `ptrtoint`/`inttoptr` \| Pointer↔integer (avoid where possible) \|
79
80	### 5. Key passes
81
82	\| Pass \| Effect \|
83	\|------\|--------\|
84	\| `mem2reg` \| Promote alloca to SSA registers \|
85	\| `instcombine` \| Instruction combining / peephole \|
86	\| `simplifycfg` \| CFG cleanup, dead block removal \|
87	\| `loop-vectorize` \| Auto-vectorisation \|
88	\| `slp-vectorize` \| Superword-level parallelism (straight-line vectorisation) \|
89	\| `inline` \| Function inlining \|
90	\| `gvn` \| Global value numbering (common subexpression elimination) \|
91	\| `licm` \| Loop-invariant code motion \|
92	\| `loop-unroll` \| Loop unrolling \|
93	\| `argpromotion` \| Promote pointer args to values \|
94	\| `sroa` \| Scalar Replacement of Aggregates \|
95
96	### 6. Debugging missed optimisations
97
98	```bash
99	# Why was a loop not vectorised?
100	clang -O2 -Rpass-missed=loop-vectorize -Rpass-analysis=loop-vectorize src.c
101
102	# Dump pass pipeline
103	clang -O2 -mllvm -debug-pass=Structure src.c -o /dev/null 2>&1 \| less
104
105	# Print IR after each pass (very verbose)
106	opt -passes='default<O2>' -print-after-all src.ll -S 2>&1 \| less
107	```
108
109	### 7. Useful llvm tools
110
111	\| Tool \| Purpose \|
112	\|------\|---------\|
113	\| `llvm-dis` \| Bitcode → textual IR \|
114	\| `llvm-as` \| Textual IR → bitcode \|
115	\| `llvm-link` \| Link multiple bitcode files \|
116	\| `llvm-lto` \| Standalone LTO \|
117	\| `llvm-nm` \| Symbols in bitcode/object \|
118	\| `llvm-objdump` \| Disassemble objects \|
119	\| `llvm-profdata` \| Merge/show PGO profiles \|
120	\| `llvm-cov` \| Coverage reporting \|
121	\| `llvm-mca` \| Machine code analyser (throughput/latency) \|
122
123	For binutils equivalents, see `skills/binaries/binutils`.
124
125	## Related skills
126
127	- Use `skills/compiler-internals/llvm-passes` for writing and testing custom LLVM passes
128	- Use `skills/compiler-internals/compiler-frontend` for generating LLVM IR from an AST
129	- Use `skills/compiler-internals/jit-compilation` for ORC JIT execution of LLVM IR
130	- Use `skills/compilers/clang` for source-level Clang flags
131	- Use `skills/binaries/linkers-lto` for LTO at link time
132	- Use `skills/profilers/linux-perf` combined with `llvm-mca` for micro-architectural analysis

People who liked this also viewed

6 picks

Type