🎯 Tingkatan Paralelisme

⚡ Paralelisme dalam Komputer

Paralelisme adalah eksekusi multiple operasi secara bersamaan untuk meningkatkan throughput dan performance.

Level Paralelisme Contoh Scope
Bit-level 32-bit vs 64-bit processor Instruction
Instruction-level Pipelining, superscalar Basic block
Data-level SIMD, vector processing Loop iterations
Task-level Multicore, multiprocessor Program/thread

🔍 Flynn's Taxonomy

Klasifikasi arsitektur komputer berdasarkan instruction stream dan data stream:

Flynn's Taxonomy
SISD
Single Instruction, Single Data
(Traditional uniprocessor)
SIMD
Single Instruction, Multiple Data
(Vector processors, GPU)
MISD
Multiple Instruction, Single Data
(Rarely used)
MIMD
Multiple Instruction, Multiple Data
(Multicore, clusters)

🖥️ Arsitektur Multicore

🔧 Multicore vs Manycore

Aspek Multicore Manycore
Jumlah Core 2-64 cores 64+ cores
Kompleksitas Core Complex cores (out-of-order) Simple cores (in-order)
Target Aplikasi General purpose Highly parallel workloads
Contoh Intel Core i7, AMD Ryzen NVIDIA GPU, Intel Xeon Phi

🏗️ Hierarki Memori Multicore

Quad-Core Processor dengan Shared L3 Cache
Core 1
L1
L1 Cache
Core 2
L1
L1 Cache
Core 3
L1
L1 Cache
Core 4
L1
L1 Cache
L2 Cache
Shared L2
L3 Cache
Shared L3
Main Memory

📊 Amdahl's Law

📈 Konsep Amdahl's Law

Amdahl's Law memprediksi maximum speedup yang bisa dicapai ketika hanya sebagian program yang dioptimasi:

Speedup = 1 / [(1 - P) + (P / N)] Dimana: P = Fraction of program that can be parallelized N = Number of processors

🎮 Simulator Amdahl's Law

Maximum Speedup: Calculating...

⚖️ Perbandingan CPU vs GPU

🔍 Arsitektur CPU vs GPU

Karakteristik CPU GPU
Core Count 4-64 cores 1000+ cores
Core Complexity Complex (out-of-order) Simple (in-order)
Cache Hierarchy Large, complex Small, simple
Clock Speed 2-5 GHz 1-2 GHz
Power Efficiency Lower (per core) Higher (per FLOP)
Ideal Workload Sequential, branchy code Data-parallel, regular code

🎯 Parallel Task Execution

Simulasi eksekusi task paralel pada multicore system:

Completed: 0/16
Time: 0 cycles
Efficiency: 0%

🚀 Arsitektur Paralel Modern

💡 Heterogeneous Computing

Kombinasi different processor types untuk optimal performance:

CPU Cores
Performance
Big Cores
GPU Cores
Throughput
Many Cores
Accelerators
Specialized
AI/ML

🔮 Trend Masa Depan

  • Chiplet Technology: Modular processor design
  • 3D Stacking: Vertical integration untuk density
  • Domain-Specific Architecture: Optimized untuk workload tertentu
  • Quantum Computing: Parallelisme quantum
  • Neuromorphic Computing: Brain-inspired architecture