wu-arch/kernels
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

sgemm_tcore: Replace hardcoded NUM_LANES with NUM_THREADS

Browse Source
This commit is contained in:
Hansung Kim
2024-06-12 21:01:37 -07:00
parent 91efc0fc14
commit ce4f3a24e3
4 changed files with 18 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/regression/sgemm_tcore/kernel.cpp
View File

@@ -7,7 +7,7 @@
#include "include/gemmini.h" #include "include/gemmini.h"
#include "gemmini_mmio.h" #include "gemmini_mmio.h"
#define GEMMINI_DMA 1 #define GEMMINI_DMA 0
#if SMEM_SIZE == 0x4000 #if SMEM_SIZE == 0x4000
#define SMEM_ADDR_Q0 ((float * const) 0xff000000) #define SMEM_ADDR_Q0 ((float * const) 0xff000000)
#define SMEM_ADDR_Q1 ((float * const) 0xff001000) #define SMEM_ADDR_Q1 ((float * const) 0xff001000)
@@ -273,10 +273,10 @@ inline void thread_block_gemm(kernel_arg_t *__UNIFORM__ arg,
// no double-buffering // no double-buffering
const uint32_t threads_per_warpgroup = threads_per_threadblock; const uint32_t threads_per_warpgroup = threads_per_threadblock;
const uint32_t warp_id_in_warpgroup = tid_in_threadblock / NUM_LANES; const uint32_t warp_id_in_warpgroup = tid_in_threadblock / NUM_THREADS;
const uint32_t warp_row = warp_id_in_warpgroup / (BN / WN); const uint32_t warp_row = warp_id_in_warpgroup / (BN / WN);
const uint32_t warp_col = warp_id_in_warpgroup % (BN / WN); const uint32_t warp_col = warp_id_in_warpgroup % (BN / WN);
const uint32_t tid_in_warp = tid_in_threadblock % NUM_LANES; const uint32_t tid_in_warp = tid_in_threadblock % NUM_THREADS;
volatile float *local_a = sharedmem_per_threadblock; volatile float *local_a = sharedmem_per_threadblock;
constexpr size_t local_a_elems = (BM * BK); constexpr size_t local_a_elems = (BM * BK);

6
tests/regression/sgemm_tcore/kernel.warpspecial.cpp
View File

@@ -9,7 +9,7 @@
#define DOUBLE_BUFFER 1 #define DOUBLE_BUFFER 1
#undef ELEM_PER_THREAD #undef ELEM_PER_THREAD
#define ELEM_PER_THREAD (WMITER * WNITER * ((TCM * TCN) / NUM_LANES) / (DOUBLE_BUFFER ? 2 : 1)) #define ELEM_PER_THREAD (WMITER * WNITER * ((TCM * TCN) / NUM_THREADS) / (DOUBLE_BUFFER ? 2 : 1))
// FIXME: NUM_THREADS and NUM_WARPS hardcoded // FIXME: NUM_THREADS and NUM_WARPS hardcoded
#if ((BM * BN / ELEM_PER_THREAD) > (CORES_PER_CLUSTER * 8 * 8)) #if ((BM * BN / ELEM_PER_THREAD) > (CORES_PER_CLUSTER * 8 * 8))
@@ -291,11 +291,11 @@ inline void thread_block_gemm(kernel_arg_t *__UNIFORM__ arg,
const uint32_t threads_per_warpgroup = threads_per_threadblock / (DOUBLE_BUFFER ? 2 : 1); const uint32_t threads_per_warpgroup = threads_per_threadblock / (DOUBLE_BUFFER ? 2 : 1);
const uint32_t warpgroup_id = tid_in_threadblock / threads_per_warpgroup; const uint32_t warpgroup_id = tid_in_threadblock / threads_per_warpgroup;
const uint32_t tid_in_warpgroup = tid_in_threadblock % threads_per_warpgroup; // FIXME const uint32_t tid_in_warpgroup = tid_in_threadblock % threads_per_warpgroup; // FIXME
const uint32_t warp_in_warpgroup = tid_in_warpgroup / NUM_LANES; const uint32_t warp_in_warpgroup = tid_in_warpgroup / NUM_THREADS;
// FIXME: warp_row / BN should be warp-specialized? // FIXME: warp_row / BN should be warp-specialized?
const uint32_t warp_row = warp_in_warpgroup / (BN / WN); const uint32_t warp_row = warp_in_warpgroup / (BN / WN);
const uint32_t warp_col = warp_in_warpgroup % (BN / WN); const uint32_t warp_col = warp_in_warpgroup % (BN / WN);
const uint32_t tid_in_warp = tid_in_threadblock % NUM_LANES; const uint32_t tid_in_warp = tid_in_threadblock % NUM_THREADS;
volatile float *local_a = sharedmem_per_threadblock; volatile float *local_a = sharedmem_per_threadblock;
// const size_t local_a_elems = threadblock_dim_x * threadblock_dim_y; // const size_t local_a_elems = threadblock_dim_x * threadblock_dim_y;

16
tests/regression/sgemm_tcore/kernel.warpspecial_dma.cpp
View File

@@ -8,8 +8,6 @@
#include "include/gemmini.h" #include "include/gemmini.h"
#include "gemmini_mmio.h" #include "gemmini_mmio.h"
#define NUM_LANES 8
#define SMEM_ADDR_Q0 ((float * const) 0xff000000) #define SMEM_ADDR_Q0 ((float * const) 0xff000000)
#define SMEM_ADDR_Q1 ((float * const) 0xff001000) #define SMEM_ADDR_Q1 ((float * const) 0xff001000)
#define SMEM_ADDR_Q2 ((float * const) 0xff002000) #define SMEM_ADDR_Q2 ((float * const) 0xff002000)
@@ -52,7 +50,7 @@
#define TCK 8 #define TCK 8
#define WMITER (WM / TCM) #define WMITER (WM / TCM)
#define WNITER (WN / TCN) #define WNITER (WN / TCN)
#define ELEM_PER_THREAD (WMITER * WNITER * ((TCM * TCN) / NUM_LANES) / (DOUBLE_BUFFER ? 2 : 1)) #define ELEM_PER_THREAD (WMITER * WNITER * ((TCM * TCN) / NUM_THREADS) / (DOUBLE_BUFFER ? 2 : 1))
// FIXME: NUM_THREADS and NUM_WARPS hardcoded // FIXME: NUM_THREADS and NUM_WARPS hardcoded
#if ((BM * BN / ELEM_PER_THREAD) > (CORES_PER_CLUSTER * 8 * 8)) #if ((BM * BN / ELEM_PER_THREAD) > (CORES_PER_CLUSTER * 8 * 8))
@@ -101,9 +99,9 @@ inline constexpr void map_operand_8lanes(const int tid, int &row, int &col) {
} }
inline constexpr void map_operand(const int tid, int &row, int &col) { inline constexpr void map_operand(const int tid, int &row, int &col) {
if constexpr (NUM_LANES == 32) { if constexpr (NUM_THREADS == 32) {
map_operand_32lanes(tid, row, col); map_operand_32lanes(tid, row, col);
} else if constexpr (NUM_LANES == 8) { } else if constexpr (NUM_THREADS == 8) {
map_operand_8lanes(tid, row, col); map_operand_8lanes(tid, row, col);
} else { } else {
// FIXME: not allowed // FIXME: not allowed
@@ -137,9 +135,9 @@ inline constexpr void map_c_8lanes(const int tid, int &row, int &col) {
} }
inline constexpr void map_c(const int tid, int &row, int &col) { inline constexpr void map_c(const int tid, int &row, int &col) {
if constexpr (NUM_LANES == 32) { if constexpr (NUM_THREADS == 32) {
map_c_32lanes(tid, row, col); map_c_32lanes(tid, row, col);
} else if constexpr (NUM_LANES == 8) { } else if constexpr (NUM_THREADS == 8) {
map_c_8lanes(tid, row, col); map_c_8lanes(tid, row, col);
} else { } else {
// FIXME: not allowed // FIXME: not allowed
@@ -571,12 +569,12 @@ inline void thread_block_gemm(kernel_arg_t *__UNIFORM__ arg,
const uint32_t threads_per_warpgroup = threads_per_threadblock / 1; const uint32_t threads_per_warpgroup = threads_per_threadblock / 1;
const uint32_t warpgroup_id = tid_in_threadblock / threads_per_warpgroup; const uint32_t warpgroup_id = tid_in_threadblock / threads_per_warpgroup;
const uint32_t tid_in_warpgroup = tid_in_threadblock % threads_per_warpgroup; // FIXME const uint32_t tid_in_warpgroup = tid_in_threadblock % threads_per_warpgroup; // FIXME
const uint32_t warp_in_warpgroup = tid_in_warpgroup / NUM_LANES; const uint32_t warp_in_warpgroup = tid_in_warpgroup / NUM_THREADS;
// FIXME: warp_row / BN should be warp-specialized? // FIXME: warp_row / BN should be warp-specialized?
const uint32_t warp_row = warp_in_warpgroup / (BN / WN); const uint32_t warp_row = warp_in_warpgroup / (BN / WN);
const uint32_t warp_col = warp_in_warpgroup % (BN / WN); const uint32_t warp_col = warp_in_warpgroup % (BN / WN);
const uint32_t tid_in_warp = tid_in_threadblock % NUM_LANES; const uint32_t tid_in_warp = tid_in_threadblock % NUM_THREADS;
volatile float *local_a = sharedmem_per_threadblock; volatile float *local_a = sharedmem_per_threadblock;
// const size_t local_a_elems = threadblock_dim_x * threadblock_dim_y; // const size_t local_a_elems = threadblock_dim_x * threadblock_dim_y;

12
tests/regression/sgemm_tcore/util.hpp
View File

@@ -6,8 +6,6 @@
#include "include/gemmini.h" #include "include/gemmini.h"
#include "gemmini_mmio.h" #include "gemmini_mmio.h"
#define NUM_LANES 8
// Constraints on parameters: // Constraints on parameters:
// * Memory: // * Memory:
// (BM + BN) * BK * sizeof(float) <= sharedmem size. // (BM + BN) * BK * sizeof(float) <= sharedmem size.
@@ -30,7 +28,7 @@
#define TCK 8 #define TCK 8
#define WMITER (WM / TCM) #define WMITER (WM / TCM)
#define WNITER (WN / TCN) #define WNITER (WN / TCN)
#define ELEM_PER_THREAD (WMITER * WNITER * (TCM * TCN) / NUM_LANES) #define ELEM_PER_THREAD (WMITER * WNITER * (TCM * TCN) / NUM_THREADS)
// number of loop around the inner 0..TCK..BK loop to simulate perfect-DRAM // number of loop around the inner 0..TCK..BK loop to simulate perfect-DRAM
// scenario // scenario
@@ -91,9 +89,9 @@ inline constexpr void map_operand_8lanes(const int tid, int &row, int &col) {
} }
inline constexpr void map_operand(const int tid, int &row, int &col) { inline constexpr void map_operand(const int tid, int &row, int &col) {
if constexpr (NUM_LANES == 32) { if constexpr (NUM_THREADS == 32) {
map_operand_32lanes(tid, row, col); map_operand_32lanes(tid, row, col);
} else if constexpr (NUM_LANES == 8) { } else if constexpr (NUM_THREADS == 8) {
map_operand_8lanes(tid, row, col); map_operand_8lanes(tid, row, col);
} else { } else {
// FIXME: not allowed // FIXME: not allowed
@@ -127,9 +125,9 @@ inline constexpr void map_c_8lanes(const int tid, int &row, int &col) {
} }
inline constexpr void map_c(const int tid, int &row, int &col) { inline constexpr void map_c(const int tid, int &row, int &col) {
if constexpr (NUM_LANES == 32) { if constexpr (NUM_THREADS == 32) {
map_c_32lanes(tid, row, col); map_c_32lanes(tid, row, col);
} else if constexpr (NUM_LANES == 8) { } else if constexpr (NUM_THREADS == 8) {
map_c_8lanes(tid, row, col); map_c_8lanes(tid, row, col);
} else { } else {
// FIXME: not allowed // FIXME: not allowed