Pinned Tweet
Built a niche site on compilers (compilersutra.com)
Last 12 months:
→ 1.09M impressions
→ 11K clicks
Ranking for LLVM, OpenCL, TVM.
Just consistent deep-tech content.
Still early. Long way to go
#llvm #ai #compiler #gcc
Last 12 months:
→ 1.09M impressions
→ 11K clicks
Ranking for LLVM, OpenCL, TVM.
Just consistent deep-tech content.
Still early. Long way to go
#llvm #ai #compiler #gcc
1
1
2
102
An M.Tech student is still waiting for their June, July & August stipend ₹36,000 pending despite repeated follow-ups. For a student, this isn’t just money. It means rent, food, travel and basic expenses.
@AICTE_INDIA
@EduMinOfIndia
please look into it
@AICTE_INDIA
@EduMinOfIndia
please look into it
2
2
5
321
Building CompilerSutraPerf (csperf) — a tool for reproducible hardware performance benchmarking.
It measures performance counters, runtime, energy/power, and compares runs to generate reports.
Would love feedback from the community!
compilersutra.com/docs/project/c…
#toolchain #llvm
It measures performance counters, runtime, energy/power, and compares runs to generate reports.
Would love feedback from the community!
compilersutra.com/docs/project/c…
#toolchain #llvm
0
1
1
29
Most devs write code.
Few know how it becomes machine instructions.
Building VELOX:
Frontend → LLVM IR → opt → llc → RISC-V → QEMU
👉 pulse = unit of computation
Read: compilersutra.com/docs/project/l…
#LLVM #Compilers #RISCV
Few know how it becomes machine instructions.
Building VELOX:
Frontend → LLVM IR → opt → llc → RISC-V → QEMU
👉 pulse = unit of computation
Read: compilersutra.com/docs/project/l…
#LLVM #Compilers #RISCV
0
0
1
63
Just published a hands-on guide to building your own LLVM Pass as a plugin!
Learn how to:
✅ Create a custom pass
Perfect for anyone diving into compiler internals 🔥
👉 compilersutra.com/docs/llvm/llvm…
#LLVM #Compilers #Programming #Systems #OpenSource #CPlusPlus
Learn how to:
✅ Create a custom pass
Perfect for anyone diving into compiler internals 🔥
👉 compilersutra.com/docs/llvm/llvm…
#LLVM #Compilers #Programming #Systems #OpenSource #CPlusPlus
0
0
1
22
Compiler MCQs for Engineers & Students 🚀
Test your fundamentals (IR, passes, backend).
Concept-based, interview-focused.
🔗 compilersutra.com/docs/mcq/quest…
#mcq #cpp #compilersutra #llvm #mlir #ai #gpu
Test your fundamentals (IR, passes, backend).
Concept-based, interview-focused.
🔗 compilersutra.com/docs/mcq/quest…
#mcq #cpp #compilersutra #llvm #mlir #ai #gpu
0
0
0
25
Everyone talks about AI models.
No one talks about how they actually run.
That’s where compilers come in.
compilersutra.com/docs/AI/
#ai #college #gpu #cpu
No one talks about how they actually run.
That’s where compilers come in.
compilersutra.com/docs/AI/
#ai #college #gpu #cpu
0
0
1
70
CompilerSutra where code meets execution.
C: #Code |
O: #Optimize |
M: Modular |
P: #Performance
I: #instrumental |
L: #Link |
E: #Execute |
R: #Reliable
S: Scale |
U: Understand |
T: Tool |
R: Refine |
A: Architecture
compilersutra.com/about_us
C: #Code |
O: #Optimize |
M: Modular |
P: #Performance
I: #instrumental |
L: #Link |
E: #Execute |
R: #Reliable
S: Scale |
U: Understand |
T: Tool |
R: Refine |
A: Architecture
compilersutra.com/about_us
0
0
1
46
6 lines of Triton → thousands of GPU instructions.
Python → TTIR → TTGIR → LLVM → AMDGCN → HSACO
a + b → buffer_load_b128
mask → v_cmp
Your code ≠ what GPU runs.
Full breakdown 👇
compilersutra.com/docs/ml-compil…
#Compilers #GPU #LLVM #Triton
Python → TTIR → TTGIR → LLVM → AMDGCN → HSACO
a + b → buffer_load_b128
mask → v_cmp
Your code ≠ what GPU runs.
Full breakdown 👇
compilersutra.com/docs/ml-compil…
#Compilers #GPU #LLVM #Triton
0
0
1
34
Just launched CompilerSutra on Vercel
Working on a resource for:
Compilers (LLVM, IR, optimization)
ML Compilers (MLIR, TVM, GPU kernels)
📊 Early traffic already coming in from Google + users reading docs
Check it out:
compilersutra.com/docs/ml-compil…
Working on a resource for:
Compilers (LLVM, IR, optimization)
ML Compilers (MLIR, TVM, GPU kernels)
📊 Early traffic already coming in from Google + users reading docs
Check it out:
compilersutra.com/docs/ml-compil…
0
0
1
25
Sharing a short write-up on measuring CPU behavior using perf.
Covers:
cycles, instructions, IPC
cache misses and their impact
sequential vs random access
Practical way to debug why a C++ loop is slow.
📖 compilersutra.com/docs/coa/measu…
#llvm #compiler #ai
Covers:
cycles, instructions, IPC
cache misses and their impact
sequential vs random access
Practical way to debug why a C++ loop is slow.
📖 compilersutra.com/docs/coa/measu…
#llvm #compiler #ai
0
0
1
33
Most compiler content is too theoretical.
I’m building CompilerSutra to fix that — practical deep dives on LLVM, MLIR, and performance ⚙️
If you're interested 👇
docs.google.com/forms/d/e/1FAI…
What topics do you want next?
#llvm #compiler #tutorial
I’m building CompilerSutra to fix that — practical deep dives on LLVM, MLIR, and performance ⚙️
If you're interested 👇
docs.google.com/forms/d/e/1FAI…
What topics do you want next?
#llvm #compiler #tutorial
0
0
1
34
Because it’s not just codegen it’s backend execution strategy.
In my benchmark
👉 GCC is ~25% faster than Clang
👉 Due to better AGU utilization + instruction scheduling
👉 Lower pipeline stalls in hot loops
Same instructions ≠ same performance.
#gcc #clang #llvm #Compilers
In my benchmark
👉 GCC is ~25% faster than Clang
👉 Due to better AGU utilization + instruction scheduling
👉 Lower pipeline stalls in hot loops
Same instructions ≠ same performance.
#gcc #clang #llvm #Compilers
Most people think:
👉 Performance = fewer instructions
❌ Wrong
Same code. Same instructions.
Still GCC is ~25% faster than Clang
Why? 👇
Comment down the reason we will discuss tommorow
#IPLIndia #cpplus2026 #layoffs2026
0
0
1
33
#CPU doesn’t run code.
It runs #bytes.
#Fetch → #Decode → #Execute → #Writeback
#Perf ≠ #instructions
It’s:
• #cache
• #deps
• #pipeline
#RAM miss = 💀 (200+ cycles)
👉 #CPU fast. #Memory is slow.
#compilers #llvm #systems #performance
compilersutra.com/docs/coa/
It runs #bytes.
#Fetch → #Decode → #Execute → #Writeback
#Perf ≠ #instructions
It’s:
• #cache
• #deps
• #pipeline
#RAM miss = 💀 (200+ cycles)
👉 #CPU fast. #Memory is slow.
#compilers #llvm #systems #performance
compilersutra.com/docs/coa/
0
0
1
32
Most people think:
👉 Performance = fewer instructions
❌ Wrong
Same code. Same instructions.
Still GCC is ~25% faster than Clang
Why? 👇
Comment down the reason we will discuss tommorow
#IPLIndia #cpplus2026 #layoffs2026
👉 Performance = fewer instructions
❌ Wrong
Same code. Same instructions.
Still GCC is ~25% faster than Clang
Why? 👇
Comment down the reason we will discuss tommorow
#IPLIndia #cpplus2026 #layoffs2026
Most people think performance = instruction count.
Not true.
Same instructions.
Same code.
GCC is ~25% faster than Clang here.
Reason?
AGU pressure + scheduling.
Compilers don’t just generate code
they decide how the CPU gets fed.
godbolt.org/z/E4havxWcW
#clang #gcc #llvm
0
0
1
89
Just launched an MCQ section for Compiler & LLVM learners!
If you're preparing for compiler interviews or revising fundamentals, this might help 👇
🔗 compilersutra.com/docs/mcq/
feedback is highly welcome 🙌
#compilers #llvm #programming #cpp #SystemDesign
If you're preparing for compiler interviews or revising fundamentals, this might help 👇
🔗 compilersutra.com/docs/mcq/
feedback is highly welcome 🙌
#compilers #llvm #programming #cpp #SystemDesign
0
0
1
30
Most people think performance = instruction count.
Not true.
Same instructions.
Same code.
GCC is ~25% faster than Clang here.
Reason?
AGU pressure + scheduling.
Compilers don’t just generate code
they decide how the CPU gets fed.
godbolt.org/z/E4havxWcW
#clang #gcc #llvm
Not true.
Same instructions.
Same code.
GCC is ~25% faster than Clang here.
Reason?
AGU pressure + scheduling.
Compilers don’t just generate code
they decide how the CPU gets fed.
godbolt.org/z/E4havxWcW
#clang #gcc #llvm
1
0
3
98