OpenXiangShan · YJMSTR · Jul 2, 2025 · Jul 2, 2025 · Jul 3, 2025 · Jul 3, 2025
diff --git a/.gitignore b/.gitignore
@@ -16,3 +16,5 @@ doc/html
 *.data*
 *.txt
 emu/obj*
+*.fst
+*.fst.hier
diff --git a/.gitmodules b/.gitmodules
@@ -1,3 +1,3 @@
 [submodule "ready-to-run"]
 	path = ready-to-run
-	url = git@github.com:jaypiper/gsim-ready-to-run.git
+	url = http://github.com/jaypiper/gsim-ready-to-run.git
diff --git a/Makefile b/Makefile
@@ -24,22 +24,22 @@ else ifeq ($(dutName),large-boom)
 	NAME ?= TestHarness
 	TEST_FILE = $(NAME)-LargeBoom
 	GSIM_FLAGS += --supernode-max-size=65 --cpp-max-size-KB=4096
-	VERI_THREADS = --threads 5
+	VERI_THREADS = --threads 4
 else ifeq ($(dutName),small-boom)
 	NAME ?= TestHarness
 	TEST_FILE = $(NAME)-SmallBoom
 	GSIM_FLAGS += --supernode-max-size=90 --cpp-max-size-KB=4096
-	VERI_THREADS = --threads 5
+	VERI_THREADS = --threads 4
 else ifeq ($(dutName),minimal-xiangshan)
 	NAME ?= SimTop
 	TEST_FILE = $(NAME)-xiangshan-minimal
 	GSIM_FLAGS += --supernode-max-size=60 --cpp-max-size-KB=8192
-	VERI_THREADS = --threads 16
+	VERI_THREADS = --threads 4
 else ifeq ($(dutName),default-xiangshan)
 	NAME ?= SimTop
 	TEST_FILE = $(NAME)-xiangshan-default
 	GSIM_FLAGS += --supernode-max-size=2 --cpp-max-size-KB=8192
-	VERI_THREADS = --threads 16
+	VERI_THREADS = --threads 4
 endif
 
 ##############################################
@@ -49,13 +49,18 @@ endif
 BUILD_DIR ?= build
 WORK_DIR = $(BUILD_DIR)/$(dutName)
 CXX = clang++
+CC = clang
 
 SHELL := /bin/bash
 TIME = /usr/bin/time
 LOG_FILE = $(WORK_DIR)/$(dutName).log
 
 CFLAGS_DUT = -DDUT_NAME=S$(NAME) -DDUT_HEADER=\"$(NAME).h\" -D__DUT_$(shell echo $(dutName) | tr - _)__
 
+# FST_CFLAGS = -DFST_WAVE -I$(abspath include/libfst)
+FST_CFLAGS = -I$(abspath include/libfst)
+FST_LDFLAGS = -L$(abspath $(BUILD_DIR)/gsim/libfst) -lfst -lz
+
 # $(1): object file
 # $(2): source file
 # $(3): compile flags (for linker flags, use -Xlinker --xxx instead of -Wl,--xxx)
@@ -129,9 +134,24 @@ PARSER_GEN_SRCS = $(foreach x, $(LEXICAL_NAME) $(SYNTAX_NAME), $(PARSER_BUILD_DI
 PARSER_GEN_HEADER = $(PARSER_BUILD_DIR)/$(SYNTAX_NAME).hh
 GSIM_SRCS = $(foreach x, src $(PARSER_DIR), $(wildcard $(x)/*.cpp))
 
-GSIM_INC_DIR = include $(PARSER_DIR)/include $(PARSER_BUILD_DIR)
+GSIM_INC_DIR = include $(PARSER_DIR)/include $(PARSER_BUILD_DIR) include/libfst
 CXXFLAGS += -ggdb -O3 -MMD $(addprefix -I,$(GSIM_INC_DIR)) -Wall -Werror --std=c++17
 
+LDLIBS = -lz
+
+LIBFST_DIR = include/libfst
+CFLAGS_LIBFST = -ggdb -O3 -MMD $(addprefix -I,$(GSIM_INC_DIR)) -Wall
+LIBFST_SRCS = $(wildcard $(LIBFST_DIR)/*.c)
+LIBFST_OBJS = $(patsubst %.c,$(GSIM_BUILD_DIR)/libfst/%.o,$(notdir $(LIBFST_SRCS)))
+LIBFST_A = $(GSIM_BUILD_DIR)/libfst/libfst.a
+
+$(GSIM_BUILD_DIR)/libfst/%.o: $(LIBFST_DIR)/%.c
+	@mkdir -p $(@D)
+	$(CC) $(CFLAGS_LIBFST) -c $< -o $@
+
+$(LIBFST_A): $(LIBFST_OBJS)
+	ar rcs $@ $^
+
 ifeq ($(DEBUG),1)
 	CXXFLAGS += -DDEBUG
 endif
@@ -150,9 +170,9 @@ $(foreach x, $(PARSER_GEN_SRCS), $(eval \
 	$(call CXX_TEMPLATE, $(PARSER_BUILD_DIR)/$(basename $(notdir $(x))).o, $(x), $(CXXFLAGS), GSIM_OBJS, $(PARSER_GEN_HEADER))))
 
 $(foreach x, $(GSIM_SRCS), $(eval \
-	$(call CXX_TEMPLATE, $(GSIM_BUILD_DIR)/$(basename $(x)).o, $(x), $(CXXFLAGS), GSIM_OBJS, $(PARSER_GEN_HEADER))))
+	$(call CXX_TEMPLATE, $(GSIM_BUILD_DIR)/$(basename $(x)).o, $(x), $(CXXFLAGS) $(FST_CFLAGS), GSIM_OBJS, $(PARSER_GEN_HEADER))))
 
-$(eval $(call LD_TEMPLATE, $(GSIM_BIN), $(GSIM_OBJS), $(CXXFLAGS) -lgmp))
+$(eval $(call LD_TEMPLATE, $(GSIM_BIN), $(GSIM_OBJS) $(LIBFST_A), $(CXXFLAGS) -lgmp $(FST_LDFLAGS)))
 
 build-gsim: $(GSIM_BIN)
 
@@ -172,7 +192,6 @@ $(GEN_CPP_DIR)/$(NAME)0.cpp: $(GSIM_BIN) $(FIRRTL_FILE)
 	@mkdir -p $(@D)
 	set -o pipefail && $(TIME) $(GSIM_BIN) $(GSIM_FLAGS) --dir $(@D) \
 		$(GSIM_FLAGS_EXTRA) $(FIRRTL_FILE) | tee $(BUILD_DIR)/gsim.log
-	$(SIG_COMMAND)
 
 compile: $(GEN_CPP_DIR)/$(NAME)0.cpp
 
@@ -194,19 +213,23 @@ EMU_GEN_SRCS = $(shell find $(GEN_CPP_DIR) -name "*.cpp" 2> /dev/null)
 EMU_SRCS = $(EMU_MAIN_SRCS) $(EMU_GEN_SRCS)
 
 EMU_CFLAGS := -O1 -MMD $(addprefix -I, $(abspath $(GEN_CPP_DIR))) $(EMU_CFLAGS) # allow to overwrite optimization level
-EMU_CFLAGS += $(MODE_FLAGS) $(CFLAGS_DUT) -Wno-parentheses-equality
+EMU_CFLAGS += $(MODE_FLAGS) $(CFLAGS_DUT) -Wno-parentheses-equality -Wno-unknown-warning-option
 EMU_CFLAGS += -fbracket-depth=2048
+EMU_CFLAGS += $(FST_CFLAGS)
 #EMU_CFLAGS += -fsanitize=address -fsanitize-address-use-after-scope
 #EMU_CFLAGS += -fsanitize=undefined -fsanitize=pointer-compare -fsanitize=pointer-subtract
 #EMU_CFLAGS += -pg -ggdb
 ifeq ($(SIMPOINT),1)
 EMU_LDFLAGS += -lz -lzstd
 endif
+EMU_LDFLAGS += $(FST_LDFLAGS)
 
 $(foreach x, $(EMU_SRCS), $(eval \
 	$(call CXX_TEMPLATE, $(EMU_BUILD_DIR)/$(basename $(notdir $(x))).o, $(x), $(EMU_CFLAGS), EMU_OBJS,)))
 
-$(eval $(call LD_TEMPLATE, $(EMU_BIN), $(EMU_OBJS), $(EMU_CFLAGS) $(EMU_LDFLAGS)))
+$(EMU_BIN): $(EMU_OBJS) $(LIBFST_A)
+	@mkdir -p $(@D) && echo + LD $@
+	@$(CXX) $^ $(EMU_CFLAGS) $(EMU_LDFLAGS) -o $@
 
 build-emu: $(EMU_BIN)
 
@@ -240,7 +263,8 @@ VERI_GEN_MK = $(VERI_BUILD_DIR)/V$(NAME).mk
 
 VERI_CFLAGS = $(call escape_quote,$(EMU_CFLAGS) $(CFLAGS_REF))
 VERI_LDFLAGS = -O1
-VERI_VFLAGS = --top $(NAME) -Wno-lint -j 8 --cc --exe +define+RANDOMIZE_GARBAGE_ASSIGN --max-num-width 1048576 --compiler clang
+VERI_LDFLAGS += $(FST_LDFLAGS)
+VERI_VFLAGS = --top $(NAME) -Wno-lint -O0 -j 8 --cc --exe +define+RANDOMIZE_GARBAGE_ASSIGN --max-num-width 1048576 --compiler clang
 ifeq ($(SIMPOINT),1)
 VERI_LDFLAGS += -lz -lzstd
 endif
@@ -251,15 +275,18 @@ VERI_VFLAGS += $(VERI_THREADS)
 VERI_VSRCS = ready-to-run/difftest/$(TEST_FILE).sv
 VERI_CSRCS-2 = $(EMU_GEN_SRCS)
 
-$(VERI_GEN_MK): $(VERI_VSRCS) $(VERI_CSRCS-$(MODE)) | $(EMU_MAIN_SRCS)
+$(VERI_GEN_MK): $(VERI_VSRCS) $(VERI_CSRCS-$(MODE)) $(LIBFST_A) | $(EMU_MAIN_SRCS)
 	@mkdir -p $(@D)
-	verilator $(VERI_VFLAGS) $(abspath $^ $|)
+	verilator $(VERI_VFLAGS) $(abspath $(VERI_VSRCS) $(VERI_CSRCS-$(MODE)) $|)
 
 $(VERI_BIN): | $(VERI_GEN_MK)
 	$(TIME) $(MAKE) OPT_FAST="-O1" CXX=clang++ -s -C $(VERI_BUILD_DIR) -f $(abspath $|)
 	ln -sf $(abspath $(VERI_BUILD_DIR)/V$(NAME)) $@
 
 compile-veri: $(VERI_GEN_MK)
+ifeq ($(MODE), 2)
+	$(SIG_COMMAND)
+endif
 
 run-veri-simpoint: $(VERI_BIN)
 	@echo 'Please run "$^ <gcpt> <checkpoint>" manually'
@@ -352,8 +379,8 @@ run-internal:
 	$(MAKE) MODE=0 difftest
 
 diff-internal:
-	$(MAKE) MODE=2 compile-veri
 	$(MAKE) MODE=2 compile
+	$(MAKE) MODE=2 compile-veri
 	$(MAKE) MODE=2 difftest
 
 run:
@@ -388,3 +415,11 @@ format-obj:
 	@clang-format -i --style=file obj/$(NAME).cpp
 
 .PHONY: clean run-internal diff-internal run diff init perf count gendoc format-obj
+
+$(EMU_EXE): $(EMU_OBJS) $(WORK_OBJS) $(LIBFST_A)
+	@mkdir -p $(@D)
+	$(CXX) $(LDFLAGS) $^ -o $@ $(LDLIBS)
+
+# LDFLAGS
+LDFLAGS ?=
+LDFLAGS += -lz
diff --git a/README.md b/README.md
@@ -6,7 +6,7 @@ GSIM accepts chirrtl, and compiles it to C++
 
 + Install [GMP](https://gmplib.org/), [clang 19(+)](https://clang.llvm.org/).
 
-## Quike Start
+## Quick Start
 
 + GSIM provides 4 RISC-V cores ready for simulation: [ysyx3](https://ysyx.oscc.cc/), [Rocket](https://github.com/chipsalliance/rocket-chip), [BOOM](https://github.com/riscv-boom/riscv-boom), [XiangShan](https://github.com/OpenXiangShan/XiangShan).
 
@@ -23,3 +23,19 @@ GSIM accepts chirrtl, and compiles it to C++
 + Run `build/gsim/gsim $(chirrtl-file)` to compile chirrtl to C++
 + Refer to `build/gsim/gsim --help` for more information
 + See [C++ harness example](https://github.com/jaypiper/simulator/blob/master/emu/emu.cpp) to know how it interacts with the emitted C++ code.
+
+## FST Waveform Generation
+
+GSIM supports FST waveform generation for debugging and analysis. You can control this feature via Makefile:
+
++ **Enable FST waveform**: Add `-DFST_WAVE` to `FST_CFLAGS` in Makefile
+  ```
+  FST_CFLAGS = -DFST_WAVE -I$(abspath include/libfst)
+  ```
+
++ **Disable FST waveform**: Remove `-DFST_WAVE` from `FST_CFLAGS`
+  ```
+  FST_CFLAGS = -I$(abspath include/libfst)
+  ```
+
+When enabled, the generated C++ model will produce `.fst` waveform files that can be viewed with tools like GTKWave.
diff --git a/emu/emu.cpp b/emu/emu.cpp
@@ -14,33 +14,62 @@
 
 #define CYCLE_STEP_PERCENT 1
 
+
+
+template<typename T>
+void* getDutMemory(T* dut) {
+  #if defined(__DUT_ysyx3__)
+    return (void*)&dut->mem_ram;
+  #elif defined(__DUT_NutShell__)
+    return (void*)&dut->mem_rdata_mem_mem;
+  #elif defined(__DUT_rocket__)
+    return (void*)&dut->mem_srams_mem;
+  #elif defined(__DUT_large_boom__) || defined(__DUT_small_boom__)
+    return (void*)&dut->mem_srams_mem;
+  #elif defined(__DUT_minimal_xiangshan__) || defined(__DUT_default_xiangshan__)
+    return (void*)&dut->memory_ram_rdata_mem_mem;
+  #else
+    return nullptr;
+  #endif
+}
+
+template<typename T>
+void* getRefMemory(T* ref) {
+  #if defined(__DUT_ysyx3__)
+    return (void*)&ref->rootp->__PVT__newtop->__PVT__mem->__PVT__ram_ext->__PVT__Memory;
+  #elif defined(__DUT_NutShell__)
+    return (void*)&ref->rootp->__PVT__SimTop->__PVT__mem->__PVT__rdata_mem->__PVT__mem_ext->__PVT__Memory;
+  #elif defined(__DUT_rocket__)
+    return (void*)&ref->rootp->__PVT__TestHarness->__PVT__mem->__PVT__srams->__PVT__mem_ext->__PVT__Memory;
+  #elif defined(__DUT_large_boom__) || defined(__DUT_small_boom__)
+    return (void*)&ref->rootp->__PVT__TestHarness->__PVT__mem->__PVT__srams->__PVT__mem_ext->__PVT__Memory;
+  #elif defined(__DUT_minimal_xiangshan__) || defined(__DUT_default_xiangshan__)
+    return (void*)&ref->rootp->__PVT__SimTop->__PVT__memory->__PVT__ram->__PVT__rdata_mem->__PVT__mem_ext->__PVT__Memory;
+  #else
+    return nullptr;
+  #endif
+}
+
+#define GET_DUT_MEMORY(dut, mem) getDutMemory(dut)
+#define GET_REF_MEMORY(ref, path) getRefMemory(ref)
+
 #if defined(__DUT_ysyx3__)
-#define DUT_MEMORY mem$ram
-#define REF_MEMORY newtop__DOT__mem__DOT__ram_ext__DOT__Memory
 #define CYCLE_MAX_PERF 5000000
 #define CYCLE_MAX_SIM  11000000
 #elif defined(__DUT_NutShell__)
-#define DUT_MEMORY mem$rdata_mem$mem
-#define REF_MEMORY SimTop__DOT__mem__DOT__rdata_mem__DOT__mem_ext__DOT__Memory
 #define CYCLE_MAX_PERF 50000000
 #define CYCLE_MAX_SIM  250000000
 #elif defined(__DUT_rocket__)
-#define DUT_MEMORY mem$srams$mem
-#define REF_MEMORY TestHarness__DOT__mem__DOT__srams__DOT__mem_ext__DOT__Memory
 #define CYCLE_MAX_PERF 2000000
 #define CYCLE_MAX_SIM  4200000
 #elif defined(__DUT_large_boom__) || defined(__DUT_small_boom__)
-#define DUT_MEMORY mem$srams$mem
-#define REF_MEMORY TestHarness__DOT__mem__DOT__srams__DOT__mem_ext__DOT__Memory
 #define CYCLE_MAX_PERF 1000000
 #ifdef __DUT_large_boom__
 #define CYCLE_MAX_SIM  3900000
 #else
 #define CYCLE_MAX_SIM  5400000
 #endif
 #elif defined(__DUT_minimal_xiangshan__) || defined(__DUT_default_xiangshan__)
-#define DUT_MEMORY memory$ram$rdata_mem$mem
-#define REF_MEMORY SimTop__DOT__memory__DOT__ram__DOT__rdata_mem__DOT__mem_ext__DOT__Memory
 #define CYCLE_MAX_PERF 500000
 #ifdef __DUT_default_xiangshan__
 #define CYCLE_MAX_SIM  1900000
@@ -98,28 +127,26 @@ static DUT_NAME* dut;
 
 void dut_init(DUT_NAME *dut) {
 #if defined(__DUT_NutShell__)
-  dut->set_difftest$$logCtrl$$begin(0);
-  dut->set_difftest$$logCtrl$$end(0);
-  dut->set_difftest$$uart$$in$$ch(-1);
+  dut->set_difftest__uart__in__ch(-1);
 #elif defined(__DUT_minimal_xiangshan__) || defined(__DUT_default_xiangshan__)
-  // dut->set_difftest$$perfCtrl$$clean(0);
-  // dut->set_difftest$$perfCtrl$$dump(0);
-  // dut->set_difftest$$logCtrl$$begin(0);
-  // dut->set_difftest$$logCtrl$$end(0);
-  // dut->set_difftest$$logCtrl$$level(0);
-  dut->set_difftest$$uart$$in$$ch(-1);
+  // dut->set_difftest__perfCtrl__clean(0);
+  // dut->set_difftest__perfCtrl__dump(0);
+  // dut->set_difftest__logCtrl__begin(0);
+  // dut->set_difftest__logCtrl__end(0);
+  // dut->set_difftest__logCtrl__level(0);
+  dut->set_difftest__uart__in__ch(-1);
 #endif
 }
 
 void dut_hook(DUT_NAME *dut) {
 #if defined(__DUT_NutShell__)
-  if (dut->get_difftest$$uart$$out$$valid()) {
-    printf("%c", dut->get_difftest$$uart$$out$$ch());
+  if (dut->get_difftest__uart__out__valid()) {
+    printf("%c", dut->get_difftest__uart__out__ch());
     fflush(stdout);
   }
 #elif defined(__DUT_minimal_xiangshan__) || defined(__DUT_default_xiangshan__)
-  if (dut->get_difftest$$uart$$out$$valid()) {
-    printf("%c", dut->get_difftest$$uart$$out$$ch());
+  if (dut->get_difftest__uart__out__valid()) {
+    printf("%c", dut->get_difftest__uart__out__ch());
     fflush(stdout);
   }
 #endif
@@ -133,8 +160,9 @@ static REF_NAME* ref;
 
 void ref_init(REF_NAME *ref) {
 #if defined(__DUT_NutShell__)
-  ref->rootp->difftest_logCtrl_begin = ref->rootp->difftest_logCtrl_begin = 0;
-  ref->rootp->difftest_uart_in_valid = -1;
+  ref->difftest_perfCtrl_clean = ref->difftest_perfCtrl_dump = 0;
+  ref->difftest_uart_in_ch = -1;
+  ref->difftest_uart_in_valid = 0;
 #elif defined(__DUT_minimal_xiangshan__) || defined(__DUT_default_xiangshan__)
   ref->difftest_perfCtrl_clean = ref->difftest_perfCtrl_dump = 0;
   ref->difftest_uart_in_ch = -1;
@@ -144,8 +172,8 @@ void ref_init(REF_NAME *ref) {
 
 void ref_hook(REF_NAME *ref) {
 #if defined(__DUT_NutShell__)
-  if (ref->rootp->difftest_uart_out_valid) {
-    printf("%c", ref->rootp->difftest_uart_out_ch);
+  if (ref->difftest_uart_out_valid) {
+    printf("%c", ref->difftest_uart_out_ch);
     fflush(stdout);
   }
 #elif defined(__DUT_minimal_xiangshan__) || defined(__DUT_default_xiangshan__)
@@ -232,19 +260,19 @@ int main(int argc, char** argv) {
   load_program(argv[1]);
 #ifdef GSIM
   dut = new DUT_NAME();
-  memcpy(&dut->DUT_MEMORY, program, program_sz);
+  memcpy(GET_DUT_MEMORY(dut, DUT_MEMORY), program, program_sz);
   dut_init(dut);
   dut_reset();
 #endif
 #ifdef VERILATOR
   ref = new REF_NAME();
-  memcpy(&ref->rootp->REF_MEMORY, program, program_sz);
+  memcpy(GET_REF_MEMORY(ref, REF_MEMORY_PATH), program, program_sz);
   ref_init(ref);
   ref_reset();
 #endif
 #ifdef GSIM_DIFF
   ref = new REF_NAME();
-  memcpy(&ref->DUT_MEMORY, program, program_sz);
+  memcpy(GET_DUT_MEMORY(ref, DUT_MEMORY), program, program_sz);
   ref_reset();
 #endif
   close_program();
@@ -284,8 +312,9 @@ int main(int argc, char** argv) {
     if (cycles % (CYCLE_MAX_SIM / (CYCLE_STEP_PERCENT * 100)) == 0 && cycles <= CYCLE_MAX_SIM) {
       auto dur = std::chrono::system_clock::now() - start;
       auto msec = std::chrono::duration_cast<std::chrono::milliseconds>(dur);
-      fprintf(stderr, "cycles %ld (%ld ms, %ld per sec) simulation process %.2lf%% \n",
-          cycles, msec.count(), cycles * 1000 / msec.count(), (double)cycles * 100 / CYCLE_MAX_SIM);
+      long long ips = msec.count() == 0 ? 0 : cycles * 1000 / msec.count();
+      fprintf(stderr, "cycles %ld (%ld ms, %lld per sec) simulation process %.2lf%% \n",
+          cycles, msec.count(), ips, (double)cycles * 100 / CYCLE_MAX_SIM);
 #ifdef PERF
       size_t totalActives = 0;
       size_t validActives = 0;
@@ -315,9 +344,15 @@ int main(int argc, char** argv) {
       }
 #endif
 #if defined(PERF) || defined(PERF_CYCLE)
-      if (cycles >= CYCLE_MAX_PERF) return 0;
+      if (cycles >= CYCLE_MAX_PERF) break;
 #endif
-      if (cycles == CYCLE_MAX_SIM) return 0;
+      if (cycles == CYCLE_MAX_SIM) break;
     }
   }
+#ifdef GSIM
+  delete dut;
+#endif
+#if (defined(VERILATOR) || defined(GSIM_DIFF)) && defined(GSIM)
+  delete ref;
+#endif
 }
-Original file line number
+Diff line change
@@ Expand Up / @@ -16,3 +16,5 @@ doc/html @@
     *.data*
     *.txt
     emu/obj*
+    *.fst
+    *.fst.hier