Skip to content

Commit

Permalink
sw: modified mx benchmarks to use L1 cache
Browse files Browse the repository at this point in the history
  • Loading branch information
@Navaneeth-KunhiPurayil
Navaneeth-KunhiPurayil committed Oct 22, 2024
1 parent a7c3923 commit adc027b
Show file tree
Hide file tree
Showing 6 changed files with 150 additions and 0 deletions.
25 changes: 25 additions & 0 deletions sw/spatzBenchmarks/dp-mxfmatmul-m4n4k4-b2/main.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,7 @@ double *c;

#define CHECK
//#define PRINT_RESULT
#define USE_CACHE

int main() {
const unsigned int num_cores = snrt_cluster_core_num();
Expand All @@ -119,16 +120,38 @@ int main() {
// Todo: we need to clarify the vl
// It can be M*K, K*N, or M*N
unsigned int vl = KERNEL_M * KERNEL_K;

#ifdef USE_CACHE
if (cid == 0) {
// Init the cache
l1d_init();
l1d_wait();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifdef USE_CACHE
if (cid == 0) {
// configure the cache
uint32_t spm_size = 32;
l1d_spm_config(spm_size);
}
#endif

#ifndef USE_CACHE
// Allocate the matrices in the local tile
if (cid == 0) {
a = (double *)snrt_l1alloc(gemm_l.M * gemm_l.K * sizeof(double));
b = (double *)snrt_l1alloc(gemm_l.K * gemm_l.N * sizeof(double));
c = (double *)snrt_l1alloc(gemm_l.M * gemm_l.N * sizeof(double));
}
#else
a = gemm_A_dram;
b = gemm_B_dram;
c = gemm_C_dram;
#endif

// Reset timer
timer = (unsigned int)-1;
Expand All @@ -150,13 +173,15 @@ int main() {
// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifndef USE_CACHE
// Initialize matrices
if (cid == 0) {
snrt_dma_start_1d(a, gemm_A_dram, gemm_l.M * gemm_l.K * sizeof(double));
snrt_dma_start_1d(b, gemm_B_dram, gemm_l.K * gemm_l.N * sizeof(double));
snrt_dma_start_1d(c, gemm_C_dram, gemm_l.M * gemm_l.N * sizeof(double));
snrt_dma_wait_all();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();
Expand Down
25 changes: 25 additions & 0 deletions sw/spatzBenchmarks/dp-mxfmatmul-m4n4k4-b4/main.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,7 @@ double *c;

#define CHECK
//#define PRINT_RESULT
#define USE_CACHE

int main() {
const unsigned int num_cores = snrt_cluster_core_num();
Expand All @@ -120,15 +121,37 @@ int main() {
// It can be M*K, K*N, or M*N
unsigned int vl = KERNEL_M * KERNEL_K;

#ifdef USE_CACHE
if (cid == 0) {
// Init the cache
l1d_init();
l1d_wait();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifdef USE_CACHE
if (cid == 0) {
// configure the cache
uint32_t spm_size = 32;
l1d_spm_config(spm_size);
}
#endif

#ifndef USE_CACHE
// Allocate the matrices in the local tile
if (cid == 0) {
a = (double *)snrt_l1alloc(gemm_l.M * gemm_l.K * sizeof(double));
b = (double *)snrt_l1alloc(gemm_l.K * gemm_l.N * sizeof(double));
c = (double *)snrt_l1alloc(gemm_l.M * gemm_l.N * sizeof(double));
}
#else
a = gemm_A_dram;
b = gemm_B_dram;
c = gemm_C_dram;
#endif

// Reset timer
timer = (unsigned int)-1;
Expand All @@ -150,13 +173,15 @@ int main() {
// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifndef USE_CACHE
// Initialize matrices
if (cid == 0) {
snrt_dma_start_1d(a, gemm_A_dram, gemm_l.M * gemm_l.K * sizeof(double));
snrt_dma_start_1d(b, gemm_B_dram, gemm_l.K * gemm_l.N * sizeof(double));
snrt_dma_start_1d(c, gemm_C_dram, gemm_l.M * gemm_l.N * sizeof(double));
snrt_dma_wait_all();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();
Expand Down
25 changes: 25 additions & 0 deletions sw/spatzBenchmarks/dp-mxfmatmul-m4n4k8-b2/main.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,7 @@ double *c;

#define CHECK
//#define PRINT_RESULT
#define USE_CACHE

int main() {
const unsigned int num_cores = snrt_cluster_core_num();
Expand All @@ -120,15 +121,37 @@ int main() {
// It can be M*K, K*N, or M*N
unsigned int vl = KERNEL_M * KERNEL_K;

#ifdef USE_CACHE
if (cid == 0) {
// Init the cache
l1d_init();
l1d_wait();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifdef USE_CACHE
if (cid == 0) {
// configure the cache
uint32_t spm_size = 32;
l1d_spm_config(spm_size);
}
#endif

#ifndef USE_CACHE
// Allocate the matrices in the local tile
if (cid == 0) {
a = (double *)snrt_l1alloc(gemm_l.M * gemm_l.K * sizeof(double));
b = (double *)snrt_l1alloc(gemm_l.K * gemm_l.N * sizeof(double));
c = (double *)snrt_l1alloc(gemm_l.M * gemm_l.N * sizeof(double));
}
#else
a = gemm_A_dram;
b = gemm_B_dram;
c = gemm_C_dram;
#endif

// Reset timer
timer = (unsigned int)-1;
Expand All @@ -150,13 +173,15 @@ int main() {
// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifndef USE_CACHE
// Initialize matrices
if (cid == 0) {
snrt_dma_start_1d(a, gemm_A_dram, gemm_l.M * gemm_l.K * sizeof(double));
snrt_dma_start_1d(b, gemm_B_dram, gemm_l.K * gemm_l.N * sizeof(double));
snrt_dma_start_1d(c, gemm_C_dram, gemm_l.M * gemm_l.N * sizeof(double));
snrt_dma_wait_all();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();
Expand Down
25 changes: 25 additions & 0 deletions sw/spatzBenchmarks/dp-mxfmatmul-m4n4k8-b4/main.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,7 @@ double *c;

#define CHECK
//#define PRINT_RESULT
#define USE_CACHE

int main() {
const unsigned int num_cores = snrt_cluster_core_num();
Expand All @@ -120,15 +121,37 @@ int main() {
// It can be M*K, K*N, or M*N
unsigned int vl = KERNEL_M * KERNEL_K;

#ifdef USE_CACHE
if (cid == 0) {
// Init the cache
l1d_init();
l1d_wait();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifdef USE_CACHE
if (cid == 0) {
// configure the cache
uint32_t spm_size = 32;
l1d_spm_config(spm_size);
}
#endif

#ifndef USE_CACHE
// Allocate the matrices in the local tile
if (cid == 0) {
a = (double *)snrt_l1alloc(gemm_l.M * gemm_l.K * sizeof(double));
b = (double *)snrt_l1alloc(gemm_l.K * gemm_l.N * sizeof(double));
c = (double *)snrt_l1alloc(gemm_l.M * gemm_l.N * sizeof(double));
}
#else
a = gemm_A_dram;
b = gemm_B_dram;
c = gemm_C_dram;
#endif

// Reset timer
timer = (unsigned int)-1;
Expand All @@ -150,13 +173,15 @@ int main() {
// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifndef USE_CACHE
// Initialize matrices
if (cid == 0) {
snrt_dma_start_1d(a, gemm_A_dram, gemm_l.M * gemm_l.K * sizeof(double));
snrt_dma_start_1d(b, gemm_B_dram, gemm_l.K * gemm_l.N * sizeof(double));
snrt_dma_start_1d(c, gemm_C_dram, gemm_l.M * gemm_l.N * sizeof(double));
snrt_dma_wait_all();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();
Expand Down
25 changes: 25 additions & 0 deletions sw/spatzBenchmarks/dp-mxfmatmul-m8n4k4-b2/main.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,7 @@ double *c;

#define CHECK
//#define PRINT_RESULT
#define USE_CACHE

int main() {
const unsigned int num_cores = snrt_cluster_core_num();
Expand All @@ -120,15 +121,37 @@ int main() {
// It can be M*K, K*N, or M*N
unsigned int vl = KERNEL_M * KERNEL_K;

#ifdef USE_CACHE
if (cid == 0) {
// Init the cache
l1d_init();
l1d_wait();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifdef USE_CACHE
if (cid == 0) {
// configure the cache
uint32_t spm_size = 32;
l1d_spm_config(spm_size);
}
#endif

#ifndef USE_CACHE
// Allocate the matrices in the local tile
if (cid == 0) {
a = (double *)snrt_l1alloc(gemm_l.M * gemm_l.K * sizeof(double));
b = (double *)snrt_l1alloc(gemm_l.K * gemm_l.N * sizeof(double));
c = (double *)snrt_l1alloc(gemm_l.M * gemm_l.N * sizeof(double));
}
#else
a = gemm_A_dram;
b = gemm_B_dram;
c = gemm_C_dram;
#endif

// Reset timer
timer = (unsigned int)-1;
Expand All @@ -150,13 +173,15 @@ int main() {
// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifndef USE_CACHE
// Initialize matrices
if (cid == 0) {
snrt_dma_start_1d(a, gemm_A_dram, gemm_l.M * gemm_l.K * sizeof(double));
snrt_dma_start_1d(b, gemm_B_dram, gemm_l.K * gemm_l.N * sizeof(double));
snrt_dma_start_1d(c, gemm_C_dram, gemm_l.M * gemm_l.N * sizeof(double));
snrt_dma_wait_all();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();
Expand Down
25 changes: 25 additions & 0 deletions sw/spatzBenchmarks/dp-mxfmatmul-m8n4k4-b4/main.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,7 @@ double *c;

#define CHECK
//#define PRINT_RESULT
#define USE_CACHE

int main() {
const unsigned int num_cores = snrt_cluster_core_num();
Expand All @@ -120,15 +121,37 @@ int main() {
// It can be M*K, K*N, or M*N
unsigned int vl = KERNEL_M * KERNEL_K;

#ifdef USE_CACHE
if (cid == 0) {
// Init the cache
l1d_init();
l1d_wait();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifdef USE_CACHE
if (cid == 0) {
// configure the cache
uint32_t spm_size = 32;
l1d_spm_config(spm_size);
}
#endif

#ifndef USE_CACHE
// Allocate the matrices in the local tile
if (cid == 0) {
a = (double *)snrt_l1alloc(gemm_l.M * gemm_l.K * sizeof(double));
b = (double *)snrt_l1alloc(gemm_l.K * gemm_l.N * sizeof(double));
c = (double *)snrt_l1alloc(gemm_l.M * gemm_l.N * sizeof(double));
}
#else
a = gemm_A_dram;
b = gemm_B_dram;
c = gemm_C_dram;
#endif

// Reset timer
timer = (unsigned int)-1;
Expand All @@ -150,13 +173,15 @@ int main() {
// Wait for all cores to finish
snrt_cluster_hw_barrier();

#ifndef USE_CACHE
// Initialize matrices
if (cid == 0) {
snrt_dma_start_1d(a, gemm_A_dram, gemm_l.M * gemm_l.K * sizeof(double));
snrt_dma_start_1d(b, gemm_B_dram, gemm_l.K * gemm_l.N * sizeof(double));
snrt_dma_start_1d(c, gemm_C_dram, gemm_l.M * gemm_l.N * sizeof(double));
snrt_dma_wait_all();
}
#endif

// Wait for all cores to finish
snrt_cluster_hw_barrier();
Expand Down

0 comments on commit adc027b

Please sign in to comment.