File size: 4,142 Bytes
7a0c684 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 |
"""
Massively parallel multicore system simulation.
Each core supports 700K hardware threads for a total of 35B threads across the system.
"""
import threading
from typing import List, Dict, Any
import time
from core import ThreadedCore
class MultiCoreSystem:
def __init__(self, num_cores=50000, threads_per_core=700000, threads_per_block=32, bits=2, num_registers=2):
self.cores = [
ThreadedCore(
num_threads=threads_per_core,
threads_per_block=threads_per_block,
bits=bits,
num_registers=num_registers
) for _ in range(num_cores)
]
self.num_cores = num_cores
self.threads_per_core = threads_per_core
self.total_threads = num_cores * threads_per_core
self.scheduler_lock = threading.Lock()
def _execute_core(self, core_id: int, inputs: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
"""Execute operations on a single core's threads"""
return self.cores[core_id].execute_parallel(inputs)
def execute_all(self, inputs: List[Dict[str, Any]], sync_blocks: bool = True):
"""
Execute operations across all cores and their threads in parallel.
inputs: List of operation inputs (one per thread)
sync_blocks: Whether to synchronize thread blocks after execution
Returns: List of results from all threads
"""
if len(inputs) > self.total_threads:
inputs = inputs[:self.total_threads]
# Split inputs across cores
inputs_per_core = self.threads_per_core
core_inputs = [
inputs[i:i + inputs_per_core]
for i in range(0, len(inputs), inputs_per_core)
]
# Create and start a thread for each core
threads = []
results = []
for core_id, core_input in enumerate(core_inputs):
thread = threading.Thread(
target=lambda: results.extend(self._execute_core(core_id, core_input))
)
threads.append(thread)
thread.start()
# Wait for all core threads to complete
for thread in threads:
thread.join()
# Optionally synchronize thread blocks across all cores
if sync_blocks:
for core in self.cores:
core.barrier_all_threads()
return results
def execute_same_all(self, operation: Dict[str, Any], sync_blocks: bool = True):
"""
Execute the same operation across all cores and their threads.
operation: Single operation to replicate across all threads
sync_blocks: Whether to synchronize thread blocks after execution
"""
inputs = [operation] * self.total_threads
return self.execute_all(inputs, sync_blocks)
if __name__ == "__main__":
print("\n--- Massively Parallel MultiCore System Simulation ---")
# Create system with 50K cores, 700K threads per core
system = MultiCoreSystem(
num_cores=50000,
threads_per_core=700000,
threads_per_block=32,
bits=2,
num_registers=2
)
print(f"Total cores: {system.num_cores:,}")
print(f"Threads per core: {system.threads_per_core:,}")
print(f"Total threads: {system.total_threads:,}")
# Example: Execute same ADD operation across all threads
print("\nExecuting same operation across all threads...")
operation = {
'a': [0.7, 0.0],
'b': [0.7, 0.7],
'cin': 0.0,
'opcode': 0b10,
'reg_sel': 0
}
start_time = time.time()
results = system.execute_same_all(operation)
end_time = time.time()
print(f"Executed {len(results):,} thread operations")
print(f"First thread result: {results[0]}")
print(f"Execution time: {end_time - start_time:.4f} seconds")
print(f"Operations per second: {len(results) / (end_time - start_time):,.0f}")
|