template gemm 2dpipe tiling

sw/blas/gemm/src/gemm.h

@@ -1,6 +1,8 @@
 #pragma once
 
 #include "gemm_kernel.h"
-// #include "gemm_occamy_baseline.h"
-// #include "gemm_occamy_1dpipe.h"
-#include "gemm_occamy_2dpipe.h"
+#ifdef OCCAMY
+// #include "gemm_baseline.h"
+// #include "gemm_1dpipe.h"
+#include "gemm_2dpipe.h"
+#endif

sw/blas/gemm/src/gemm_2dpipe.h

@@ -0,0 +1,17 @@
+#pragma once
+
+#define FLOAT_T fp64
+#include "gemm_2dpipe_tpl.h"
+#undef FLOAT_T
+
+#define FLOAT_T fp32
+#include "gemm_2dpipe_tpl.h"
+#undef FLOAT_T
+
+#define FLOAT_T fp16
+#include "gemm_2dpipe_tpl.h"
+#undef FLOAT_T
+
+#define FLOAT_T fp8
+#include "gemm_2dpipe_tpl.h"
+#undef FLOAT_T

sw/blas/gemm/src/gemm_2dpipe_tpl.h

@@ -0,0 +1,30 @@
+#include "gemm_decls.h"
+
+#ifndef FLOAT_T 
+#error "Define FLOAT_T to use this template."
+#endif
+
+// Instantiate template code
+#define USE_C2C_TILES true
+
+#define IS_DM_CORE true
+#include "gemm_tiling_2dpipe_tpl.h"
+#undef IS_DM_CORE
+
+#define IS_DM_CORE false
+#include "gemm_tiling_2dpipe_tpl.h"
+#undef IS_DM_CORE
+
+#undef USE_C2C_TILES
+
+#ifndef SNBLAS_GEMM
+#define SNBLAS_GEMM(float_t) CONCAT(snblas_gemm_, float_t)
+#endif
+
+void SNBLAS_GEMM(FLOAT_T) (const SnblasGemmInfo info, const SNBLAS_GEMM_ARGS(FLOAT_T) args, bool bench) {
+    if (snrt_is_dm_core()) {
+        SNBLAS_GEMM_TILING(true, FLOAT_T)(info, args, bench);
+    } else {
+        SNBLAS_GEMM_TILING(false, FLOAT_T)(info, args, bench);
+    }
+}

sw/blas/gemm/src/gemm_occamy_baseline.h → sw/blas/gemm/src/gemm_baseline.h

@@ -9,54 +9,7 @@
 #include "gemm.h"
 #include "snrt.h"
 
-#include "dump.h"
-NAMED_DUMP(uint32_t, aIdx, 0x1a)
-NAMED_DUMP(uint32_t, bIdx, 0x1b)
-NAMED_DUMP(uint32_t, cIdx, 0x1c)
-NAMED_DUMP(uint32_t, ib, 0x10)
-NAMED_DUMP(uint32_t, jb, 0x11)
-NAMED_DUMP(uint32_t, kb, 0x12)
-NAMED_DUMP(double, a, 0xa)
-NAMED_DUMP(double, b, 0xb)
-NAMED_DUMP(double, c, 0xc)
-
-/**
- * \brief Implements a reversing loop for an index range
- * \param begin Beginning of the range
- * \param end End of the range
- * \param dir Sets the direction of traversal. True: loop starts at begin.
- * \param i_prev Set the previous index to the first index, must update this
- * manually at the end of the loop. \details i_end_floor will contain the exact
- * end with the stride, s.t. the reversed loop starts at the correct index.
- */
-#define FOR_EACH(i, begin, end, stride, dir, i_prev)                     \
-    dir = !dir;                                                          \
-    const int i##_end_floor =                                            \
-        ((end - begin + stride - 1) / stride) * stride - stride + begin; \
-    const int i##_first = dir ? begin : i##_end_floor;                   \
-    const int i##_last = dir ? i##_end_floor : begin;                    \
-    i = i##_first;                                                       \
-    i_prev = i;                                                          \
-    for (; dir ? i <= i##_last : i >= i##_last;                          \
-         i = dir ? i + stride : i - stride)
-
-#define L1_M 8  // 128;
-#define L1_N 8  // 128;
-#define L1_K 8  // 128;
-#define L1_LDA L1_K
-#define L1_LDB L1_N
-#define L1_LDC L1_N
-
-/**
- * \brief Maps the layout of the TCDM. May be double buffered.
- */
-typedef struct {
-    double A[L1_M * L1_K];
-    double B[L1_K * L1_N];
-    double C[L1_M * L1_N];
-} TcdmLayout;
-
-NAMED_DUMP(TcdmLayout*, l1, 0x8)
+#include "gemm_decls.h"
 
 /**
  * \brief Each cluster performs a GEMM for A, B, C inside each TCDM
@@ -71,16 +24,11 @@ void gemm_cluster_kernel(double alpha, double beta, uint32_t M, uint32_t N,
     for (uint32_t i = p[0]; i < M; i += P[0]) {
         for (uint32_t j = 0; j < N; j++) {
             uint32_t cIdx = i * ldc + j;  // C[i][j]
-            // dump_cIdx(cIdx);
-            // dump_c(C[cIdx]);
             register double c0 = beta * C[cIdx];
+
             for (uint32_t k = 0; k < K; k++) {
                 uint32_t aIdx = i * lda + k;  // A[i][k]
                 uint32_t bIdx = k * ldb + j;  // B[k][j]
-                // dump_aIdx(aIdx);
-                // dump_bIdx(bIdx);
-                // dump_a(A[aIdx]);
-                // dump_b(B[bIdx]);
 
                 c0 += A[aIdx] * B[bIdx];
             }

sw/blas/gemm/src/gemm_tiling_2dpipe_tpl.h

@@ -13,7 +13,7 @@
 #include "gemm_kernel.h"
 
 #ifndef SNBLAS_GEMM_TILING
-#define SNBLAS_GEMM_TILING(is_dm_core, float_t) CONCAT3(gemm_, is_dm_core, float_t)
+#define SNBLAS_GEMM_TILING(is_dm_core, float_t) CONCAT3(snblas_gemm_, is_dm_core, float_t)
 #endif
 
 void SNBLAS_GEMM_TILING(IS_DM_CORE, FLOAT_T) (const SnblasGemmInfo info, const SNBLAS_GEMM_ARGS(FLOAT_T) args, const bool bench) {

sw/blas/gemm/src/main.c

@@ -29,7 +29,7 @@ int main() {
     uint32_t ldb = N;
     uint32_t ldc = N;
 
-    GemmInfo gemmInfo = {0};
+    SnblasGemmInfo gemmInfo = {0};
     gemmInfo.M   = M;
     gemmInfo.N   = N;
     gemmInfo.K   = K;
@@ -39,7 +39,7 @@ int main() {
     gemmInfo.ta  = TA;
     gemmInfo.tb  = TB;
 
-    GemmArgs gemmArgs = {0};
+    SNBLAS_GEMM_ARGS(DTYPE) gemmArgs = {0};
     gemmArgs.A     = a;
     gemmArgs.B     = b;
     gemmArgs.C     = c;
@@ -48,7 +48,7 @@ int main() {
 
     for (volatile int i = iters; i > 0; --i) {
         // if (i == 1) snrt_mcycle(); // start
-        gemm_oc(gemmInfo, gemmArgs, i == 1);
+        SNBLAS_GEMM(DTYPE)(gemmInfo, gemmArgs, i == 1);
         // gemm_oc(data_dtype_size, data_expand, setup_ssr, data_TA, data_TB, data_M, data_N, data_K, 1,
         //         data_a, lda, data_b, ldb, data_BETA, data_c, ldc);
         if (i == 1) snrt_mcycle(); // end