WIP:[MemPool-Spatz]

1. Clean the tests and benchmarks for MemPool-Spatz.
2. Add TeraPool-Spatz4 configuration.

config/terapool_spatz4_fpu.mk

@@ -0,0 +1,62 @@
+# Copyright 2021 ETH Zurich and University of Bologna.
+# Licensed under the Apache License, Version 2.0, see LICENSE for details.
+# SPDX-License-Identifier: Apache-2.0
+
+# Author: Matheus Cavalcante, ETH Zurich
+
+################
+##  TeraPool  ##
+################
+
+# Global Control
+terapool ?= 1
+
+# Number of cores
+num_cores ?= 256
+
+# Number of groups
+num_groups ?= 4
+
+# Number of cores per Terapool tile
+num_cores_per_tile ?= 2
+
+# Number of sub groups per Terapool group
+num_sub_groups_per_group ?= 4
+
+# L1 scratchpad banking factor
+banking_factor ?= 4
+
+# Access latency between remote groups
+# Options: "7", "9" or "11":
+remote_group_latency_cycles ?= 7
+
+# Radix for hierarchical AXI interconnect
+axi_hier_radix ?= 9
+
+# Number of AXI masters per group
+axi_masters_per_group ?= 4
+
+# Number of DMA backends in each group
+dmas_per_group ?= 4
+
+# L2 Banks/Channels
+l2_banks = 16
+
+# Makefile RTL Filtering Control
+subgroup_rtl = 1
+
+# Activate Spatz and RVV
+spatz ?= 1
+
+# Lenght of single vector register
+vlen ?= 512
+
+# Number of IPUs
+n_ipu ?= 4
+
+n_fpu ?= 4
+
+# Deactivate the XpulpIMG extension
+xpulpimg ?= 0
+
+rvf ?= 1

hardware/Makefile

@@ -102,6 +102,9 @@ vlog_defs += -DDMAS_PER_GROUP=$(dmas_per_group)
 vlog_defs += -DAXI_HIER_RADIX=$(axi_hier_radix) -DAXI_MASTERS_PER_GROUP=$(axi_masters_per_group)
 vlog_defs += -DSEQ_MEM_SIZE=$(seq_mem_size) -DXQUEUE_SIZE=$(xqueue_size)
 # This parameter is only used for TeraPool configurations
+ifdef terapool
+	vlog_defs += -DTERAPOOL=$(terapool)
+endif
 vlog_defs += -DNUM_SUB_GROUPS_PER_GROUP=$(num_sub_groups_per_group) -DREMOTE_GROUP_LATENCY_CYCLES=$(remote_group_latency_cycles)
 
 ifeq ($(spatz), 1)

hardware/scripts/questa/wave.tcl

@@ -30,15 +30,15 @@ add wave /mempool_tb/wfi
 
 # Add all cores from group 0 tile 0
 for {set core 0}  {$core < [examine -radix dec mempool_pkg::NumCoresPerTile]} {incr core} {
-    if {$config == {terapool}} {
+    if {$config == {terapool_spatz4_fpu} | $config == {terapool}} {
         do ../scripts/questa/wave_core.tcl 0 0 0 $core
     } else {
         do ../scripts/questa/wave_core.tcl 0 0 $core
     }
 }
 
 # Add specific cores from different tiles
-if {$config == {terapool}} {
+if {$config == {terapool_spatz4_fpu} | $config == {terapool}} {
     do ../scripts/questa/wave_core.tcl 1 0 0 0
 } else {
     do ../scripts/questa/wave_core.tcl 1 0 0
@@ -47,7 +47,7 @@ if {$config == {terapool}} {
 # Add groups
 for {set group 0} {$group < [examine -radix dec /mempool_pkg::NumGroups]} {incr group} {
     # Add tiles
-    if {$config == {terapool}} {
+    if {$config == {terapool_spatz4_fpu} | $config == {terapool}} {
         for {set subgroup 0} {$subgroup < [expr min(4,[examine -radix dec /mempool_pkg::NumSubGroupsPerGroup])]} {incr subgroup} {
             for {set tile 0} {$tile < [expr min(4,[examine -radix dec /mempool_pkg::NumTilesPerSubGroup])]} {incr tile} {
                 do ../scripts/questa/wave_tile.tcl $group $subgroup $tile
@@ -63,14 +63,14 @@ for {set group 0} {$group < [examine -radix dec /mempool_pkg::NumGroups]} {incr
     for {set tgtgroup 0} {$tgtgroup < [examine -radix dec /mempool_pkg::NumGroups]} {incr tgtgroup} {
         if {$tgtgroup != $group} {
             set interco_idx [expr $group ^ $tgtgroup]
-            if {$config == {terapool}} {
+            if {$config == {terapool_spatz4_fpu} | $config == {terapool}} {
                 add wave -group group_[$group] -group interconnect_to_group[$tgtgroup] /mempool_tb/dut/i_mempool_cluster/gen_groups[$group]/gen_rtl_group/i_group/gen_remote_interco[$interco_idx]/i_remote_interco/*
             } else {
                 add wave -group group_[$group] -group interconnect_to_group[$tgtgroup] /mempool_tb/dut/i_mempool_cluster/gen_groups[$group]/i_group/gen_remote_interco[$interco_idx]/i_remote_interco/*
             }
         }
     }
-    if {$config != {terapool}} {
+    if {$config != {terapool} & $config != {terapool_spatz4_fpu}} {
         add wave -group group_[$group] -group interconnect_local /mempool_tb/dut/i_mempool_cluster/gen_groups[$group]/i_group/i_local_interco/*
     }
 }
@@ -83,7 +83,7 @@ add wave -Group Control_Registers /mempool_tb/dut/i_ctrl_registers/*
 add wave -Group DMA /mempool_tb/dut/i_mempool_dma/*
 add wave -Group DMA -Group Reg /mempool_tb/dut/i_mempool_dma/i_mempool_dma_frontend_reg_top/*
 for {set group 0} {$group < [examine -radix dec /mempool_pkg::NumGroups]} {incr group} {
-  if  {$config == {terapool}} {
+  if  {$config == {terapool_spatz4_fpu} | $config == {terapool}} {
     add wave -Group DMA_midend_${group} /mempool_tb/dut/i_mempool_cluster/gen_groups[$group]/gen_rtl_group/i_group/i_idma_distributed_midend/NoMstPorts
     add wave -Group DMA_midend_${group} /mempool_tb/dut/i_mempool_cluster/gen_groups[$group]/gen_rtl_group/i_group/i_idma_distributed_midend/DmaRegionWidth
     add wave -Group DMA_midend_${group} /mempool_tb/dut/i_mempool_cluster/gen_groups[$group]/gen_rtl_group/i_group/i_idma_distributed_midend/DmaRegionStart
@@ -121,7 +121,7 @@ add wave -Group DMA_midend_cluster /mempool_tb/dut/i_mempool_cluster/i_idma_dist
 
 add wave -Group DMA_split /mempool_tb/dut/i_mempool_cluster/i_idma_split_midend/*
 
-if {$config == {terapool}} {
+if {$config == {terapool_spatz4_fpu} | $config == {terapool}} {
   do ../scripts/questa/wave_cache.tcl 0 0 0 0
 } else {
   do ../scripts/questa/wave_cache.tcl 0 0 0

hardware/scripts/questa/wave_cache.tcl

@@ -4,7 +4,7 @@
 
 # Create cache for core $3 from group $1 tile $2 (core_id=NUM_CORES_PER_group*$1+NUM_CORES_PER_TILE*$2+$3)
 
-if {$config == {terapool}} {
+if {$config == {terapool_spatz4_fpu} | $config == {terapool}} {
   add wave -noupdate -group cache[$1][$2][$3][$4] -divider Parameters
   add wave -noupdate -group cache[$1][$2][$3][$4] /mempool_tb/dut/i_mempool_cluster/gen_groups[$1]/gen_rtl_group/i_group/gen_sub_groups[$2]/gen_rtl_sg/i_sub_group/gen_tiles[$3]/i_tile/gen_caches[$4]/i_snitch_icache/NR_FETCH_PORTS
   add wave -noupdate -group cache[$1][$2][$3][$4] /mempool_tb/dut/i_mempool_cluster/gen_groups[$1]/gen_rtl_group/i_group/gen_sub_groups[$2]/gen_rtl_sg/i_sub_group/gen_tiles[$3]/i_tile/gen_caches[$4]/i_snitch_icache/L0_LINE_COUNT

hardware/scripts/questa/wave_core.tcl

@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: SHL-0.51
 
 # Create group for core $3 from group $1 tile $2 (core_id=NUM_CORES_PER_group*$1+NUM_CORES_PER_TILE*$2+$3)
-if {$config == {terapool}} {
+if {$config == {terapool_spatz4_fpu} | $config == {terapool}} {
   add wave -noupdate -group core[$1][$2][$3][$4] -group Params /mempool_tb/dut/i_mempool_cluster/gen_groups[$1]/gen_rtl_group/i_group/gen_sub_groups[$2]/gen_rtl_sg/i_sub_group/gen_tiles[$3]/i_tile/gen_cores[$4]/gen_mempool_cc/riscv_core/BootAddr
   add wave -noupdate -group core[$1][$2][$3][$4] -group Params /mempool_tb/dut/i_mempool_cluster/gen_groups[$1]/gen_rtl_group/i_group/gen_sub_groups[$2]/gen_rtl_sg/i_sub_group/gen_tiles[$3]/i_tile/gen_cores[$4]/gen_mempool_cc/riscv_core/MTVEC
   add wave -noupdate -group core[$1][$2][$3][$4] -group Params /mempool_tb/dut/i_mempool_cluster/gen_groups[$1]/gen_rtl_group/i_group/gen_sub_groups[$2]/gen_rtl_sg/i_sub_group/gen_tiles[$3]/i_tile/gen_cores[$4]/gen_mempool_cc/riscv_core/RVE

hardware/scripts/questa/wave_tile.tcl

@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: SHL-0.51
 
 # Create group for group $1 tile $2
-if {$config == {terapool}} {
+if {$config == {terapool_spatz4_fpu} | $config == {terapool}} {
     add wave -noupdate -group group_[$1] -group sub_group_[$2] -group Tile_[$3] -group Params /mempool_pkg::NumBanksPerTile
     add wave -noupdate -group group_[$1] -group sub_group_[$2] -group Tile_[$3] -group Params /mempool_pkg::NumTiles
     add wave -noupdate -group group_[$1] -group sub_group_[$2] -group Tile_[$3] -group Params /mempool_pkg::NumBanks

hardware/src/mempool_group.sv

@@ -111,18 +111,18 @@ module mempool_group
   end: gen_tcdm_struct
 
   `ifdef TERAPOOL
-  /*********************
-   *  TeraPool Section *
-   *********************/
-  // ----------------------------------------------------- //
-  // SubGroups Instantiation.
-  // PostLayout Simulation:
-  //   - Only SubGroup_0 Replaced by Netlist for simulation.
-  // TCDM interconnections:
-  //   - Local SubGroups;
-  //   - Remote Groups.
-  // Most of AXI/DMA logics put into SubGroup level.
-  // ----------------------------------------------------- //
+    /*********************
+     *  TeraPool Section *
+     *********************/
+    // ----------------------------------------------------- //
+    // SubGroups Instantiation.
+    // PostLayout Simulation:
+    //   - Only SubGroup_0 Replaced by Netlist for simulation.
+    // TCDM interconnections:
+    //   - Local SubGroups;
+    //   - Remote Groups.
+    // Most of AXI/DMA logics put into SubGroup level.
+    // ----------------------------------------------------- //
 
     /****************
      *  Sub_Groups  *
@@ -579,16 +579,16 @@ module mempool_group
     );
 
   `else
-  /******************************
-   *  MemPool / MinPool Section *
-   ******************************/
-  // ----------------------------------------------------- //
-  // Tile Instantiation.
-  // TCDM interconnections:
-  //   - Local Tiles;
-  //   - Remote Groups;
-  // Most of AXI/DMA logics put on Group level.
-  // ----------------------------------------------------- //
+    /******************************
+     *  MemPool / MinPool Section *
+     ******************************/
+    // ----------------------------------------------------- //
+    // Tile Instantiation.
+    // TCDM interconnections:
+    //   - Local Tiles;
+    //   - Remote Groups;
+    // Most of AXI/DMA logics put on Group level.
+    // ----------------------------------------------------- //
 
     /***********
      *  Tiles  *

software/apps/sp-axpy/main.c → software/apps/spatz_apps/sp-axpy/main.c

@@ -29,8 +29,8 @@
 #endif
 
 // dump(result, 1);
-const unsigned int csize = 256;
-const unsigned int core_count = 16;
+const unsigned int csize = 64;
+const unsigned int core_count = 64;
 const unsigned int esize = csize*core_count;
 
 float x[esize];

software/apps/spatz_apps/sp-fmatmul/data/.gitignore

@@ -0,0 +1 @@
+data_gemm.h