Programmazione di processori massivamente paralleli: un approccio pratico: il modello di esecuzione SIMT e la partizione in warp

Il SIMT (Single-Instruction, Multiple-Thread) modello è il battito cardiaco dell'architettura GPU. Mentre programmi thread individuali, l'hardware li coordina in una gerarchia a due livelli di griglie e blocchi. Per massimizzare l'efficienza, l'hardware suddivide ulteriormente questi blocchi in unità da 32 thread chiamate warp.

1. SIMT vs. SIMD

A differenza dell'SIMD per CPU (come SSE/AVX), dove si impacchetta manualmente i dati nei registri, il SIMT permette ai thread di apparire indipendenti. L'hardware raggruppa automaticamente i thread in warp, recuperando un'unica istruzione per tutti i 32 thread in modo da eseguirla simultaneamente.

2. Regola di linearizzazione

I programmatori usano threadIdx.x, y, z per la logica, ma l'hardware lo trasforma in una sequenza unidimensionale per la pianificazione:

Indice = x + (y × blockDim.x) + (z × blockDim.x × blockDim.y)

Poiché la dimensione x è l'indice più veloce da variare, i thread con valori consecutivi di threadIdx.x valori si trovano solitamente nello stesso warp, il che è fondamentale per coalescimento della memoria.

TERMINALbash — 80x24

> Ready. Click "Run" to execute.

QUESTION 1

What is the physical scheduling 'atom' (minimum unit) in NVIDIA's SIMT model?

A single Thread

A Warp (32 threads)

A Streaming Multiprocessor

A Grid

QUESTION 2

In an 8x8 thread block, which threads will be assigned to Warp 0?

Threads with linear IDs 0 through 31

Threads with y=0 and y=1

Only threads where x=0

All 64 threads belong to Warp 0

QUESTION 3

How does SIMT differ from traditional SIMD (like SSE/AVX)?

SIMT requires manual packing of vector registers.

SIMT allows threads to operate independently at the software level while hardware manages vectorization.

SIMD is only for GPUs; SIMT is for CPUs.

There is no difference; they are synonymous.

QUESTION 4

Using the linearization formula, what is the ID of T(2, 1, 0) in a block with blockDim.x=16 and blockDim.y=16?

QUESTION 5

What happens if threads within a warp take different execution paths (e.g., an if-else statement)?

The warp executes all paths in parallel without penalty.

The warp splits into two warps.

The hardware serializes the paths, disabling threads not on the current path.

The kernel crashes.

Architectural Analysis: Linearization and Bandwidth

Applying Warp Partitioning to Matrix Addition

You are optimizing a matrix addition kernel on a 2D grid. The threads are organized into 8x8 blocks. You are considering using shared memory to improve performance.

1. [Reading Context: Figure 6.1 shows an example of placing threads of a two-dimensional (2D) block into linear order.] Draw out/Represent the partitioning of an 8x8 thread block into warps. Which thread index (x,y) marks the end of Warp 0?

Solution:
In an 8x8 block (64 threads), threads are linearized using $Index = tx + (ty \times 8)$. Warp 0 covers Linear IDs 0 to 31. T(0,0) is ID 0. ID 31 is calculated as $tx=7, ty=3$ ($7 + 3 \times 8 = 31$). Therefore, Warp 0 starts at $T(0,0)$ and ends with $T(7,3)$. Warp 1 begins at $T(0,4)$ (ID 32) and ends at $T(7,7)$ (ID 63).

2. Consider the matrix addition where each element of the output matrix is the sum of the corresponding elements of the two input matrices. Can one use shared memory to reduce the global memory bandwidth consumption? Explain with details (approx. 150 words).

Solution:
No, shared memory cannot reduce global memory bandwidth for matrix addition. In matrix addition, each output element $C[i][j]$ is the sum of $A[i][j] + B[i][j]$. Each element from the input matrices $A$ and $B$ is accessed exactly once by exactly one thread to compute its unique output. Shared memory is a high-speed scratchpad designed for data reuse—where multiple threads within a block read the same global memory address repeatedly (as seen in matrix multiplication). Since there is zero data commonality between threads in matrix addition, loading data into shared memory first would actually increase overhead. It would require one global load to shared memory, a `__syncthreads()` barrier, and then a shared memory load before the addition. This adds instruction count and synchronization latency without decreasing the number of global memory transactions. To optimize matrix addition, one should focus on global memory coalescing rather than shared memory utilization.