Software updates since tapeout

environment.mk

@@ -44,7 +44,7 @@ __BSG_ENVIRONMENT_MK := 1
 # Name of this project
 PROJECT = cl_manycore
 
-CL_DIR           := $(shell git rev-parse --show-toplevel)
+CL_DIR           ?= $(shell git rev-parse --show-toplevel)
 HARDWARE_PATH    := $(CL_DIR)/hardware
 LIBRARIES_PATH   := $(CL_DIR)/libraries
 MACHINES_PATH    := $(CL_DIR)/machines

examples/.gitignore

@@ -24,3 +24,4 @@ ucli.key
 *.saifgen
 *.json
 dramsim3.txt
+*.a

examples/cuda/hw_barrier/Makefile

@@ -0,0 +1,141 @@
+# Copyright (c) 2021, University of Washington All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without modification,
+# are permitted provided that the following conditions are met:
+#
+# Redistributions of source code must retain the above copyright notice, this list
+# of conditions and the following disclaimer.
+#
+# Redistributions in binary form must reproduce the above copyright notice, this
+# list of conditions and the following disclaimer in the documentation and/or
+# other materials provided with the distribution.
+#
+# Neither the name of the copyright holder nor the names of its contributors may
+# be used to endorse or promote products derived from this software without
+# specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
+# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR
+# ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
+# ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+# This Makefile compiles, links, and executes examples Run `make help`
+# to see the available targets for the selected platform.
+
+################################################################################
+# environment.mk verifies the build environment and sets the following
+# makefile variables:
+#
+# LIBRAIRES_PATH: The path to the libraries directory
+# HARDWARE_PATH: The path to the hardware directory
+# EXAMPLES_PATH: The path to the examples directory
+# BASEJUMP_STL_DIR: Path to a clone of BaseJump STL
+# BSG_MANYCORE_DIR: Path to a clone of BSG Manycore
+###############################################################################
+
+REPLICANT_PATH:=$(shell git rev-parse --show-toplevel)
+
+include $(REPLICANT_PATH)/environment.mk
+
+# TEST_NAME is the basename of the executable
+TEST_NAME = main
+# KERNEL_NAME is the name of the CUDA-Lite Kernel
+KERNEL_NAME = barrier_test
+
+###############################################################################
+# Host code compilation flags and flow
+###############################################################################
+# TEST_SOURCES is a list of source files that need to be compiled
+TEST_SOURCES  = main.cpp
+
+TEST_HEADERS =
+
+DEFINES += -D_XOPEN_SOURCE=500 -D_BSD_SOURCE -D_DEFAULT_SOURCE
+CDEFINES += 
+CXXDEFINES += 
+
+FLAGS     = -O3 -g -Wall -Wno-unused-function -Wno-unused-variable
+CFLAGS   += -std=c99 $(FLAGS)
+CXXFLAGS += -std=c++11 $(FLAGS)
+
+# compilation.mk defines rules for compilation of C/C++
+include $(EXAMPLES_PATH)/compilation.mk
+
+###############################################################################
+# Host code link flags and flow
+###############################################################################
+
+LDFLAGS =
+
+# link.mk defines rules for linking of the final execution binary.
+include $(EXAMPLES_PATH)/link.mk
+
+$(TEST_OBJECTS): $(TEST_HEADERS)
+
+###############################################################################
+# Device code compilation flow
+###############################################################################
+
+# BSG_MANYCORE_KERNELS is a list of manycore executables that should
+# be built before executing.
+
+BSG_MANYCORE_KERNELS ?= kernel.riscv
+RISCV_TARGET_OBJECTS += kernel.rvo
+RISCV_INCLUDES  +=
+RISCV_CCPPFLAGS += -D__KERNEL__
+
+TILE_GROUP_DIM_X ?= 16
+TILE_GROUP_DIM_Y ?= 8
+RISCV_DEFINES += -DTILE_GROUP_DIM_X=$(TILE_GROUP_DIM_X)
+RISCV_DEFINES += -DTILE_GROUP_DIM_Y=$(TILE_GROUP_DIM_Y)
+RISCV_DEFINES += -Dbsg_tiles_X=$(TILE_GROUP_DIM_X)
+RISCV_DEFINES += -Dbsg_tiles_Y=$(TILE_GROUP_DIM_Y)
+
+include $(EXAMPLES_PATH)/cuda/riscv.mk
+
+###############################################################################
+# Execution flow
+#
+# C_ARGS: Use this to pass arguments that you want to appear in argv
+#         For SPMD tests C arguments are: <Path to RISC-V Binary> <Test Name>
+#
+# SIM_ARGS: Use this to pass arguments to the simulator
+###############################################################################
+
+C_ARGS   ?= $(BSG_MANYCORE_KERNELS) $(KERNEL_NAME)
+C_ARGS   += $(TILE_GROUP_DIM_X) $(TILE_GROUP_DIM_Y)
+
+SIM_ARGS ?=
+
+# Include platform-specific execution rules
+include $(EXAMPLES_PATH)/execution.mk
+
+###############################################################################
+# Regression Flow
+###############################################################################
+
+regression: main.exec.log
+	@grep "BSG REGRESSION TEST .*PASSED.*" $< > /dev/null
+
+###############################################################################
+# Default rules, help, and clean
+###############################################################################
+.DEFAULT_GOAL := help
+help:
+	@echo "Usage:"
+	@echo "make {clean | $(TEST_NAME).{profile,debug} | $(TEST_NAME).{profile,debug}.log}"
+	@echo "      $(TEST_NAME).profile: Build executable with profilers enabled"
+	@echo "      $(TEST_NAME).debug: Build waveform executable (if VCS)"
+	@echo "      $(TEST_NAME).{profile,debug}.log: Run specific executable"
+	@echo "      clean: Remove all subdirectory-specific outputs"
+
+
+.PHONY: clean
+
+clean:

examples/cuda/hw_barrier/kernel.c

@@ -0,0 +1,19 @@
+#include "bsg_manycore.h"
+#include "bsg_cuda_lite_barrier.h"
+#include "bsg_barrier_amoadd.h"
+
+int barrier_test()
+{
+    bsg_barrier_hw_tile_group_init();
+    for (int x = 0; x < bsg_tiles_X; x++) {
+        for (int y = 0; y < bsg_tiles_Y; y++) {
+            if (__bsg_x == x && __bsg_y == y) {
+                int *ptr = (int*)bsg_remote_ptr_io(IO_X_INDEX, 0x8888);
+                *ptr = (__bsg_x<<16)|(__bsg_y);
+            }
+            bsg_barrier_hw_tile_group_sync();
+        }
+    }
+
+    return 0;
+}

examples/cuda/hw_barrier/main.cpp

@@ -0,0 +1,113 @@
+#include "bsg_manycore_regression.h"
+#include "bsg_manycore_cuda.h"
+#include "bsg_manycore.h"
+#include "bsg_manycore_responder.h"
+#include <cstring>
+#include <vector>
+
+static hb_mc_device_t *dev = nullptr;
+static std::vector<int> pkt_data;
+
+
+//////////////////////////////////////////////////////
+// Responder to check for packets from the manycore //
+//////////////////////////////////////////////////////
+static
+hb_mc_request_packet_id_t resp_ids [] = {
+    RQST_ID(RQST_ID_ANY_X, RQST_ID_ANY_Y, RQST_ID_ADDR(0x8888)),
+    {/*sentinal*/},
+};
+
+static int resp_init(hb_mc_responder_t *resp, hb_mc_manycore_t *mc)
+{
+    return HB_MC_SUCCESS;
+}
+
+static int resp_quit(hb_mc_responder_t *resp, hb_mc_manycore_t *mc)
+{
+    return HB_MC_SUCCESS;
+}
+
+static int resp_respond(hb_mc_responder_t *resp, hb_mc_manycore_t *mc, const hb_mc_request_packet_t *rqst)
+{
+    bsg_pr_dbg("%s: received packet from (%3d,%3d)\n"
+               , __func__
+               , rqst->x_src
+               , rqst->y_src);
+
+    pkt_data.push_back(static_cast<int>(hb_mc_request_packet_get_data(rqst)));
+    return HB_MC_SUCCESS;
+}
+
+
+static
+hb_mc_responder_t resp ("barrier-test", resp_ids, resp_init, resp_quit, resp_respond);
+source_responder(resp);
+
+////////////////////////
+// Main hosst program //
+////////////////////////
+int barrier_test_main(int argc, char *argv[])
+{
+    char *rvp    = argv[1];
+    char *kname  = argv[2];
+    int tgx = atoi(argv[3]);
+    int tgy = atoi(argv[4]);
+
+    dev = (hb_mc_device_t*)malloc(sizeof(*dev));
+    if (!dev) {
+        bsg_pr_err("failed to allocate memory\n");
+        return HB_MC_NOMEM;
+    }
+
+    // initialize
+    BSG_CUDA_CALL(hb_mc_device_init(dev, "cuda hw barrier", 0));
+    BSG_CUDA_CALL(hb_mc_device_program_init(dev, rvp, "baralloc", 0));
+
+    // enque job and execute
+    hb_mc_dimension_t gd, tgd;
+    gd = hb_mc_dimension(1,1);
+    tgd = hb_mc_dimension(tgx, tgy);
+
+    BSG_CUDA_CALL(hb_mc_kernel_enqueue(dev, gd, tgd, kname, 0, nullptr));
+    BSG_CUDA_CALL(hb_mc_device_tile_groups_execute(dev));
+
+    // cleanup
+    BSG_CUDA_CALL(hb_mc_device_program_finish(dev));
+    BSG_CUDA_CALL(hb_mc_device_finish(dev));
+
+    // check that we got the right number of packets
+    if (static_cast<int>(pkt_data.size()) != tgx*tgy) {
+        bsg_pr_err("Expected %d packets from %3d X %3d group, received %d\n"
+                   , tgx*tgy
+                   , tgx
+                   , tgy
+                   , static_cast<int>(pkt_data.size()));
+        return HB_MC_FAIL;
+    }
+
+    // validate results by checking that the expected packets arrived in-order
+    int id = 0;
+    for (int x = 0; x < tgx; x++) {
+        for (int y = 0; y < tgy; y++) {
+            int data = pkt_data[id];
+            int dx = (data >> 16) & 0xffff;
+            int dy = data & 0xffff;
+
+            if (x != dx || y != dy) {
+                bsg_pr_err("Packet %d: expected from (%d,%d) but found (%d,%d)\n"
+                           , id
+                           , x
+                           , y
+                           , dx
+                           , dy);
+                return HB_MC_FAIL;
+            }
+            id++;
+        }
+    }
+    return HB_MC_SUCCESS;
+}
+
+declare_program_main("HW Barrier Test", barrier_test_main);
+

examples/cuda/riscv.mk

@@ -89,7 +89,7 @@ RISCV_LLVM_PATH  := $(RISCV_TOOLS_PATH)/llvm/llvm-install
 # RISC-V Tool Configuration
 ################################################################################
 
-RISCV_LINK_GEN := $(BSG_MANYCORE_DIR)/software/py/bsg_manycore_link_gen.py
+RISCV_LINK_GEN ?= $(BSG_MANYCORE_DIR)/software/py/bsg_manycore_link_gen.py
 
 # These flags are not supported by clang
 RISCV_GNU_FLAGS = -frerun-cse-after-loop -fweb -frename-registers -mtune=bsg_vanilla_2020
@@ -267,17 +267,25 @@ crt.rvo: $(BSG_MANYCORE_COMMON_PATH)/crt.S
 # We compile these locally so that we don't interfere with the files in
 # $(BSG_MANYCORE_LIB_PATH).
 # BSG Manycore Library Objects
-LIBBSG_MANYCORE_OBJECTS  += bsg_set_tile_x_y.rvo
-LIBBSG_MANYCORE_OBJECTS  += bsg_tile_config_vars.rvo
-LIBBSG_MANYCORE_OBJECTS  += bsg_printf.rvo
+LIBBSG_MANYCORE_OBJECTS  += bsg_set_tile_x_y.c.rvo
+LIBBSG_MANYCORE_OBJECTS  += bsg_tile_config_vars.c.rvo
+LIBBSG_MANYCORE_OBJECTS  += bsg_printf.c.rvo
+LIBBSG_MANYCORE_OBJECTS  += bsg_barrier_amoadd.S.rvo
+LIBBSG_MANYCORE_OBJECTS  += bsg_cuda_lite_barrier.c.rvo
+
+libbsg_manycore_riscv.a: $(LIBBSG_MANYCORE_OBJECTS)
+	$(RISCV_AR) rcs $@ $^
 
 # See comment above about _RISCV_GCC and _RISCV_GXX for explanation of
 # the preceding underscore.
 $(LIBBSG_MANYCORE_OBJECTS) main.rvo: RISCV_CXX = $(_RISCV_GCC)
 
-$(LIBBSG_MANYCORE_OBJECTS): %.rvo:$(BSG_MANYCORE_LIB_PATH)/%.c
+$(filter %.c.rvo,$(LIBBSG_MANYCORE_OBJECTS)): %.c.rvo:$(BSG_MANYCORE_LIB_PATH)/%.c
 	$(_RISCV_GCC) $(RISCV_CFLAGS) $(RISCV_DEFINES) $(RISCV_INCLUDES) -c $< -o $@
 
+$(filter %.S.rvo,$(LIBBSG_MANYCORE_OBJECTS)): %.S.rvo:$(BSG_MANYCORE_LIB_PATH)/%.S
+	$(_RISCV_GCC) $(RISCV_CFLAGS) $(RISCV_DEFINES) -D__ASSEMBLY__=1 $(RISCV_INCLUDES) -c $< -o $@
+
 main.rvo: $(BSG_MANYCORE_CUDALITE_MAIN_PATH)/main.c
 	$(_RISCV_GCC) $(RISCV_CFLAGS) $(RISCV_DEFINES) $(RISCV_INCLUDES) -c $< -o $@
 
@@ -379,21 +387,23 @@ RISCV_LDFLAGS += -ffast-math
 RISCV_LDFLAGS += -lc
 RISCV_LDFLAGS += -lm
 RISCV_LDFLAGS += -lgcc
+RISCV_LDFLAGS += -L.
+RISCV_LDFLAGS += -lbsg_manycore_riscv
 
 # TODO: temporary fix to solve this problem: https://stackoverflow.com/questions/56518056/risc-v-linker-throwing-sections-lma-overlap-error-despite-lmas-belonging-to-dif
 RISCV_LDFLAGS += -Wl,--no-check-sections 
 
 # This builds a .riscv binary for the current machine type and tile
 # group size. RISCV_TARGET_OBJECTS are .rvo files that will be linked
 # in the final binary.
-%.riscv: crt.rvo bsg_set_tile_x_y.rvo bsg_tile_config_vars.rvo main.rvo $(RISCV_TARGET_OBJECTS) $(RISCV_LINK_SCRIPT) 
-	$(RISCV_LD) -T $(RISCV_LINK_SCRIPT) $(RISCV_LDFLAGS) $(filter %.rvo,$^) -o $@
+%.riscv: crt.rvo libbsg_manycore_riscv.a main.rvo $(RISCV_TARGET_OBJECTS) $(RISCV_LINK_SCRIPT) 
+	$(RISCV_LD) -T $(RISCV_LINK_SCRIPT) $(filter %.rvo,$^) -o $@ $(RISCV_LDFLAGS) 
 
 %.dis: %.riscv
 	$(RISCV_OBJDUMP) -dS $<
 
 kernel.link.clean:
-	rm -rf *.riscv *.rvo.S *.rvo.ll $(RISCV_LINK_SCRIPT)
+	rm -rf *.riscv *.rvo.S *.rvo.ll $(RISCV_LINK_SCRIPT) libbsg_manycore_riscv.a
 
 
 .PRECIOUS: %.riscv

libraries/bsg_manycore_config.cpp

@@ -144,6 +144,12 @@ int hb_mc_config_init(const hb_mc_config_raw_t raw[HB_MC_CONFIG_MAX],
         CHECK_FIELD(HB_MC_CONFIG_NOC_COORD_Y_WIDTH, idx > 0 && idx < 32);
         config->noc_coord_width.y = idx;
 
+        config->noc_ruche_factor.x = 3;
+        config->noc_ruche_factor.y = 0;
+
+        config->bar_ruche_factor.x = config->noc_ruche_factor.x;
+        config->bar_ruche_factor.y = config->noc_ruche_factor.y;
+
         config->basejump = raw[HB_MC_CONFIG_REPO_BASEJUMP_HASH];
         config->manycore = raw[HB_MC_CONFIG_REPO_MANYCORE_HASH];
         config->f1 = raw[HB_MC_CONFIG_REPO_F1_HASH];

libraries/bsg_manycore_config.h

@@ -93,6 +93,8 @@ extern "C" {
                 hb_mc_dimension_t pods;      // how many vcore pods?
                 hb_mc_dimension_t pod_shape; // what is the shape of a pod?
                 hb_mc_dimension_t noc_coord_width;
+                hb_mc_dimension_t noc_ruche_factor;
+                hb_mc_dimension_t bar_ruche_factor;
                 hb_mc_dimension_t pod_coord_width;
                 hb_mc_dimension_t tile_coord_width;
                 hb_mc_coordinate_t host_interface;

libraries/bsg_manycore_config_pod.h

@@ -244,6 +244,7 @@ extern "C" {
                         return  tile.x
                                 + cfg->pod_shape.x;
                 }
+                return 0;
         }
 
         static inline hb_mc_idx_t