large fp16 kernel

This commit is contained in:
Richard Yan
2024-09-05 16:22:38 -07:00
parent 4fddca3d1a
commit dd3244fba0

View File

@@ -6,19 +6,29 @@
#include "include/gemmini.h"
#include "gemmini_mmio.h"
#define TILE_M 64
#define TILE_N 64
#define TILE_K 64
#define SMEM_ADDR_Q0 ((float * const) 0xff000000)
#define SMEM_ADDR_Q1 ((float * const) 0xff004000)
#define SMEM_ADDR_Q2 ((float * const) 0xff008000)
#define SMEM_ADDR_Q3 ((float * const) 0xff00c000)
#define SPAD_ADDR_Q0 0x0
#define SPAD_ADDR_Q1 0x200
#define SPAD_ADDR_Q2 0x400
#define SPAD_ADDR_Q3 0x600
// fp16 16x16
#define TILE_M 128
#define TILE_N 128
#define TILE_K 128
#define BOUND_INST 0x800080008ULL
#define NUM_THREADS_IN_CLUSTER 512
// fp32 8x8
// #define TILE_M 64
// #define TILE_N 64
// #define TILE_K 64
// #define SMEM_ADDR_Q0 ((float * const) 0xff000000)
// #define SMEM_ADDR_Q1 ((float * const) 0xff004000)
// #define SMEM_ADDR_Q2 ((float * const) 0xff008000)
// #define SMEM_ADDR_Q3 ((float * const) 0xff00c000)
// #define SPAD_ADDR_Q0 0x0
// #define SPAD_ADDR_Q1 0x200
// #define SPAD_ADDR_Q2 0x400
// #define SPAD_ADDR_Q3 0x600
// #define BOUND_INST 0x800080008ULL
// #define NUM_THREADS_IN_CLUSTER 256
// fp32 4x4
// #define TILE_M 32
// #define TILE_N 32
// #define TILE_K 32
@@ -31,9 +41,9 @@
// #define SPAD_ADDR_Q2 0x100
// #define SPAD_ADDR_Q3 0x180
// #define BOUND_INST 0x400040004ULL
// #define NUM_THREADS_IN_CLUSTER 256
#define NUM_CLUSTERS 1
#define NUM_THREADS_IN_CLUSTER 256 \
// (NUM_CORES * NUM_WARPS * NUM_THREADS)
#define rd_cycles_force(x) asm volatile ("csrr %0, mcycle" : "=r" (x))
@@ -42,7 +52,7 @@
#define PRINTF(...) sprintf(PRINT_BUF, __VA_ARGS__)
// #define PRINTF(...) vx_printf(__VA_ARGS__)
#define SWISH(beta, x) ((x) / (1 + exp(-(beta) * (x))))
// #define POWER
#define POWER
typedef uint16_t smem_elem_t;
// typedef float smem_elem_t;
@@ -83,13 +93,11 @@ void thread_block_matmul_gemmini(kernel_arg_t *__UNIFORM__ arg,
const uint32_t num_tile_rows_per_tb = num_tiles_m / NUM_CLUSTERS;
constexpr scale_t MVIN_SCALE_IDENTITY_HEX = 0x3c00;
if (HW_TID() == 0) {
gemmini_extended3_config_ld(dim_k * sizeof(elem_t), MVIN_SCALE_IDENTITY_HEX, false, 0);
gemmini_extended3_config_ld(dim_n * sizeof(elem_t), MVIN_SCALE_IDENTITY_HEX, false, 1);
gemmini_extended3_config_ld(dim_k * sizeof(elem_t), MVIN_SCALE_IDENTITY, false, 0);
gemmini_extended3_config_ld(dim_n * sizeof(elem_t), MVIN_SCALE_IDENTITY, false, 1);
// gemmini_extended3_config_ld(repeating_bias ? 0 : (stride_D * sizeof_D), D_scale_factor, low_D, 2);
gemmini_extended_config_st(dim_n * sizeof(elem_t), 0, MVIN_SCALE_IDENTITY_HEX);
gemmini_extended_config_st(dim_n * sizeof(elem_t), 0, MVIN_SCALE_IDENTITY);
// gemmini_extended_config_st(stride_C * sizeof_C, act & 3, scale);
}
@@ -187,4 +195,4 @@ int main() {
vx_spawn_tasks_contiguous(grid_size, (vx_spawn_tasks_cb)kernel_body, arg);
#endif
return 0;
}
}