performance refactoring - rebalanced stream buffers accross the device to enforce output buffering rule at compoments boudaries, finally resolved block ram R/W collusion discrepencies,

This commit is contained in:
Blaise Tine
2020-12-19 02:45:06 -08:00
parent 29cd2f5dff
commit 4bbd7bf408
76 changed files with 1313 additions and 1098 deletions

View File

@@ -66,7 +66,7 @@ make ase
./run_ase.sh build_ase_1c ../../benchmarks/opencl/vecadd/vecadd
# modify "vsim_run.tcl" to dump VCD trace
vcd file vortex.vcd
vcd file trace.vcd
vcd add -r /*/Vortex/hw/rtl/*
run -all
@@ -104,8 +104,11 @@ lsof +D build_ase_1c
make -C pipeline clean && make -C pipeline > pipeline/build.log 2>&1 &
make -C cache clean && make -C cache > cache/build.log 2>&1 &
make -C core clean && make -C core > core/build.log 2>&1 &
make -C core8 clean && make -C core8 > core8/build.log 2>&1 &
make -C vortex clean && make -C vortex > vortex/build.log 2>&1 &
make -C top clean && make -C top > top/build.log 2>&1 &
make -C top1 clean && make -C top1 > top1/build.log 2>&1 &
make -C top8 clean && make -C top8 > top8/build.log 2>&1 &
# How to calculate the maximum operating frequency?
200 Mhz -> period = 1/200x10^6 = 5ns

View File

@@ -1,6 +1,6 @@
#!/bin/bash
SCRIPT_DIR=$PWD
SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
BUILD_DIR=$1

View File

@@ -4,21 +4,21 @@
+define+QUARTUS
+define+FPU_FAST
#+define+SCOPE
+define+PERF_ENABLE
#+define+PERF_ENABLE
+define+DBG_PRINT_CORE_ICACHE
+define+DBG_PRINT_CORE_DCACHE
+define+DBG_PRINT_CACHE_BANK
+define+DBG_PRINT_CACHE_SNP
+define+DBG_PRINT_CACHE_MSRQ
+define+DBG_PRINT_CACHE_TAG
+define+DBG_PRINT_CACHE_DATA
+define+DBG_PRINT_DRAM
+define+DBG_PRINT_PIPELINE
+define+DBG_PRINT_OPAE
+define+DBG_PRINT_AVS
+define+DBG_PRINT_SCOPE
+define+DBG_CACHE_REQ_INFO
#+define+DBG_PRINT_CORE_ICACHE
#+define+DBG_PRINT_CORE_DCACHE
#+define+DBG_PRINT_CACHE_BANK
#+define+DBG_PRINT_CACHE_SNP
#+define+DBG_PRINT_CACHE_MSRQ
#+define+DBG_PRINT_CACHE_TAG
#+define+DBG_PRINT_CACHE_DATA
#+define+DBG_PRINT_DRAM
#+define+DBG_PRINT_PIPELINE
#+define+DBG_PRINT_OPAE
#+define+DBG_PRINT_AVS
#+define+DBG_PRINT_SCOPE
#+define+DBG_CACHE_REQ_INFO
vortex_afu.json
QI:vortex_afu.qsf

View File

@@ -1,5 +1,5 @@
+define+NUM_CORES=2
+define+L2_ENABLE=0
+define+SYNTHESIS
+define+QUARTUS
+define+FPU_FAST

View File

@@ -1,5 +1,5 @@
+define+NUM_CORES=4
+define+L2_ENABLE=1
+define+SYNTHESIS
+define+QUARTUS
+define+FPU_FAST

View File

@@ -1,5 +1,7 @@
# Analysis & Synthesis Assignments
set_global_assignment -name NUM_PARALLEL_PROCESSORS ALL
set_global_assignment -name VERILOG_INPUT_VERSION SYSTEMVERILOG_2009
set_global_assignment -name ADD_PASS_THROUGH_LOGIC_TO_INFERRED_RAMS ON
set_global_assignment -name VERILOG_MACRO QUARTUS
@@ -7,7 +9,14 @@ set_global_assignment -name VERILOG_MACRO SYNTHESIS
set_global_assignment -name VERILOG_MACRO NDEBUG
set_global_assignment -name MESSAGE_DISABLE 16818
set_global_assignment -name VERILOG_MACRO FPU_FAST
set_global_assignment -name TIMEQUEST_DO_REPORT_TIMING ON
set_global_assignment -name OPTIMIZATION_TECHNIQUE SPEED
set_global_assignment -name OPTIMIZATION_MODE "AGGRESSIVE PERFORMANCE"
set_global_assignment -name FITTER_EFFORT "STANDARD FIT"
set_global_assignment -name OPTIMIZE_HOLD_TIMING "ALL PATHS"
set_global_assignment -name ROUTER_TIMING_OPTIMIZATION_LEVEL MAXIMUM
set_global_assignment -name OPTIMIZE_MULTI_CORNER_TIMING ON
set_global_assignment -name MIN_CORE_JUNCTION_TEMP 0
set_global_assignment -name MAX_CORE_JUNCTION_TEMP 100
set_global_assignment -name POWER_BOARD_THERMAL_MODEL "NONE (CONSERVATIVE)"
@@ -17,10 +26,4 @@ set_global_assignment -name TIMEQUEST_DO_CCPP_REMOVAL ON
set_global_assignment -name SYNTH_TIMING_DRIVEN_SYNTHESIS ON
set_global_assignment -name TIMEQUEST_MULTICORNER_ANALYSIS ON
set_global_assignment -name POWER_USE_TA_VALUE 65
set_global_assignment -name SEED 1
set_global_assignment -name OPTIMIZE_MULTI_CORNER_TIMING ON
set_global_assignment -name FITTER_EFFORT "STANDARD FIT"
set_global_assignment -name OPTIMIZE_HOLD_TIMING "ALL PATHS"
set_global_assignment -name OPTIMIZATION_TECHNIQUE SPEED
set_global_assignment -name ROUTER_TIMING_OPTIMIZATION_LEVEL MAXIMUM
set_global_assignment -name OPTIMIZATION_MODE "AGGRESSIVE PERFORMANCE"
set_global_assignment -name SEED 1