Add Blackwell tensor core support to Chipyard

- Update RadianceConfigs.scala with Blackwell configurations
- Update Verilator Makefile with optimized build flags
- Update submodules: radiance (Blackwell implementation), gemmini (params update)
- Update build flags and gitignore

sims/.gitignore

@@ -1,2 +1,2 @@
 *.bin
-*.bin.*
+*.bin.*

sims/common-sim-flags.mk

@@ -1,7 +1,7 @@
 #----------------------------------------------------------------------------------------
 # common gcc configuration/optimization
 #----------------------------------------------------------------------------------------
-SIM_OPT_CXXFLAGS := -O3
+SIM_OPT_CXXFLAGS := -O0
 LRISCV=-lriscv
 
 export USE_CHISEL6=1

sims/verilator/Makefile

@@ -91,6 +91,7 @@ RUNTIME_PROFILING_VFLAGS := $(if $(filter $(VERILATOR_PROFILE),all),\
 
 VERILATOR_THREADS ?= 1
 RUNTIME_THREADS := --threads $(VERILATOR_THREADS) --threads-dpi all
+VERILATOR_MAKE_JOBS ?= 4
 
 USE_FST ?= 0
 TRACING_OPTS := $(if $(filter $(USE_FST),0),\
@@ -103,12 +104,16 @@ get_waveform_flag = +vcdfile=$(1).$(if $(filter $(USE_FST),0),vcd,fst)
 #----------------------------------------------------------------------------------------
 # we initially had --noassert for performance, but several modules use
 # assertions, including dramsim, so we enable --assert by default
+VERILATOR_OUTPUT_SPLIT ?= 10000
+VERILATOR_OUTPUT_SPLIT_CFUNCS ?= 100
+
 VERILATOR_OPT_FLAGS ?= \
-	-O3 \
+	-O0 \
 	--x-assign fast \
 	--x-initial fast \
-	--output-split 10000 \
-	--output-split-cfuncs 100
+	--output-split $(VERILATOR_OUTPUT_SPLIT) \
+	--output-split-cfuncs $(VERILATOR_OUTPUT_SPLIT_CFUNCS) \
+	-DPRINTF_COND_=1
 
 # default flags added for external IP (cva6/NVDLA)
 VERILOG_IP_VERILATOR_FLAGS := \
@@ -164,6 +169,7 @@ VERILATOR_NONCC_OPTS = \
 VERILATOR_CXXFLAGS = \
 	$(SIM_CXXFLAGS) \
 	$(RUNTIME_PROFILING_CFLAGS) \
+	-fno-inline \
 	-DVERILATOR
 
 VERILATOR_LDFLAGS = $(SIM_LDFLAGS)
@@ -208,10 +214,10 @@ $(model_mk_debug): $(sim_common_files) $(EXTRA_SIM_REQS)
 # invoke make to make verilator sim rules
 #########################################################################################
 $(sim): $(model_mk) $(dramsim_lib)
-	$(MAKE) VM_PARALLEL_BUILDS=1 -C $(model_dir) -f V$(TB).mk
+	$(MAKE) -j $(VERILATOR_MAKE_JOBS) VM_PARALLEL_BUILDS=1 -C $(model_dir) -f V$(TB).mk
 
 $(sim_debug): $(model_mk_debug) $(dramsim_lib)
-	$(MAKE) VM_PARALLEL_BUILDS=1 -C $(model_dir_debug) -f V$(TB).mk
+	$(MAKE) -j $(VERILATOR_MAKE_JOBS) VM_PARALLEL_BUILDS=1 -C $(model_dir_debug) -f V$(TB).mk
 
 #########################################################################################
 # create a verilator vpd rule