script1.tcl

set log [open log.log w]
set now [exec date]

set TIME_start [clock clicks -milliseconds]

puts $log "$now: starting run..."

# this will mark all assets as dont_touch
source ../../../assets.tcl

# this helps cts to succeed. otherwise it complains that there are no inv/buffs available in the library.
set_db cts_buffer_cells {CLKBUF_X1 CLKBUF_X2 CLKBUF_X3}
#set_db cts_inverter_cells {BUF_X1 BUF_X2 BUF_X4 BUF_X8 BUF_X16 BUF_X32 CLKBUF_X1 CLKBUF_X2 CLKBUF_X3}

connect_global_net VDD -type pg_pin -pin_base_name VDD -inst_base_name * -override
connect_global_net VSS -type pg_pin -pin_base_name VSS -inst_base_name * 

# std cell height is 1.400
# std cell pitch is  0.190

# this helps a lot with our super dense designs. innovus default is 0.95 (95% density)
set_db opt_max_density 1.5

switch $DESIGN {
	"present" {
		set param_rows 30
		set param_cols 213
		#30/214 gives good results except for power. 30/213 is better.
	} 
	"camellia" {
		set param_rows 72
		set param_cols 480
	}
	"cast" {
		set param_rows 91
		set param_cols 617
	}
	"misty" {
		set param_rows 74
		set param_cols 578
	}
	"openmsp430_1" {
		set param_rows 69
		set param_cols 492
	}
	"openmsp430_2" {
		set param_rows 70
		set param_cols 550
	}
	"openmsp430_array" {
		set param_rows 560
		set param_cols 4400
	}
	"seed" {
		set param_rows 87
		set param_cols 647
	}
	"sparx" { 
		set param_rows 99
		set param_cols 714
		# 711 is good for fspfi. 712 is good for power. 714 is better for power
		# paper used 99 / 714
	}
	"tdea" {
		set param_rows 43
		set param_cols 325
	}
	default {
		# all AESs are named the same, so these lines have to be adjusted manually
		set param_rows 135
		set param_cols 991
	}
}
# AES1 136 984 used to work, but doesnt pass timing. 986 works but power values are bad. will use it for now
# AES2 135 991
# AES3 135 991 


set height [expr $param_rows * 1.4]
set width [expr $param_cols * 0.19]

delete_all_floorplan_objs

if {$DESIGN=="AES"} {
	create_floorplan -core_size $width $height 1.7 1.7 1.7 1.7 -no_snap_to_grid
} else {
	create_floorplan -core_size $width $height 2.4 2.4 2.4 2.4 -no_snap_to_grid
}

deselect_routes
select_routes -nets VDD
select_routes -nets VSS
delete_selected_from_floorplan

if {$DESIGN=="AES"} {
	add_rings -nets {VDD VSS} -width 0.6 -spacing 0.5 -offset 0 -layer {top 7 bottom 7 left 6 right 6}
	route_special
	add_stripes -nets {VDD VSS} -width 0.4 -spacing 0.5 -start_offset 2.32 -set_to_set_distance 5 -layer 6 -direction vertical
	add_stripes -nets {VDD VSS} -width 0.4 -spacing 0.5 -start_offset 5.06 -set_to_set_distance 5 -layer 7 -direction horizontal
} else {
	add_rings -nets {VDD VSS} -width 0.8 -spacing 0.8 -offset 0 -layer {top 5 bottom 5 left 6 right 6}
	route_special
	add_stripes -nets {VDD VSS} -width 0.4 -spacing 0.4 -start_offset 5.00 -set_to_set_distance 15 -layer 4 -direction vertical
	add_stripes -nets {VDD VSS} -width 0.4 -spacing 0.4 -start_offset 4.00 -set_to_set_distance 14 -layer 5 -direction horizontal 
}


# this keeps the same pin strategy from the original design
read_floorplan ../originals/floorplan.fp -sections {pin}

# commands for moving the clk pin to perform additional analysis on overheads. this violates the contest rules 
# and was not used for comparison with other works
#edit_pin -pin_width 0.07 -pin_depth 0.07 -fix_overlap 1 -side left -layer 3  -pin clk -assign 0 5 -fixed_pin 1
#edit_pin -pin_width 0.07 -pin_depth 0.07 -fix_overlap 1 -side right -layer 3  -pin clk -assign 0 5 -fixed_pin 1
#edit_pin -pin_width 0.07 -pin_depth 0.07 -fix_overlap 1 -side top -layer 3  -pin clk -assign 5 0 -fixed_pin 1
#edit_pin -pin_width 0.07 -pin_depth 0.07 -fix_overlap 1 -side bot -layer 3  -pin clk -assign 5 0 -fixed_pin 1
legalize_pins

#suspend

# initial settings
set setup_target 0.000
set_db opt_setup_target_slack $setup_target
set_db opt_hold_target_slack -1

# useless
#set_db opt_leakage_to_dynamic_ratio 0.0
# useless because it is the default?
#setOptMode -usefulSkew true

# this bloated up the design considerably!
#set_db place_global_timing_effort high

# these don't do much at all, decreased quality a bit
set_db place_global_activity_power_driven true
#set_db place_global_activity_power_driven_effort high
#set_db place_global_clock_power_driven_effort high

# cts stuff, largely inconsequential
#set_db opt_useful_skew_eco_route true
#set_db cts_merge_clock_gates true
#set_db cts_merge_clock_logic true

# this one helps with long nets and improves the score... but design quality goes down very quickly
#set_db opt_max_length 100

#this did nothing
#set_db route_design_with_timing_driven true

# this does something, but messes with convergence. has to be studied further.
set_db design_flow_effort extreme

#power as a priority insted of insertion delay? seems to do nothing
#eval_legacy "set_ccopt_mode -cts_opt_priority power"
#set_db ccopt_auto_limit_insertion_delay_factor 0.5

#does nothing
#eval_legacy  "set_ccopt_property effort high"

# this does something but it is not a clear win without convergence
#set_db opt_post_route_area_reclaim setup_aware

# this increases runtime considerably, but indeed saves some power. it also messes with timing a bit. might be useful to try this out, turn on, then turn off, halfway?
#set_db opt_power_effort high
#set_db opt_leakage_to_dynamic_ratio 0.01

#lowest cell count but causes power inconsistency
#set_db opt_drv false

# this has a negative effect, apparently. might have to check against ccopt skew target
#set_db opt_skew_ccopt extreme

# these two do nothing
#set_db place_global_activity_power_driven true
#set_db place_global_activity_power_driven_effort high

#decreases quality
#set_db place_global_clock_power_driven_effort high

# does almost nothing, small bloating
#set_db place_global_cong_effort high

# does almost nothing
#set_db place_detail_activity_power_driven true

# does nothing
#set_db place_detail_swap_eeq_cells true

# does nothing
#set_db place_detail_wire_length_opt_effort high

#basic flow is like this:
#place_opt_design
#opt_design -pre_cts 
#ccopt_design
#opt_design -post_cts
#route_design
#opt_design -post_route

# advanced flow is below

set now [exec date]
puts $log "$now: starting place_opt..."
place_opt_design

set tp [report_timing -collection]
set slack [get_db $tp .slack]
set round 1

while {$slack < 0} {
	set setup_target [expr $setup_target + 0.001]
	set_db opt_setup_target_slack $setup_target
	set now [exec date]
	puts $log "$now: starting pre_cts opt round#$round ..."
	flush $log
	opt_design -pre_cts
	set tp [report_timing -collection ]
	set slack [get_db $tp .slack]
	set round [expr $round + 1]
}

if {$round == 1} {
	# this means the slack was positive so pre cts opt was never called in the loop above
	opt_design -pre_cts
}

set now [exec date]
puts $log "$now: starting ccopt..."
# these commands configure the target skew, but then it gets recomputed internally. there must be a better/guaranteed way to set this.
#create_clock_tree -name mytree -source clk -no_skew_group
#create_skew_group -name skew -sources clk -auto_sinks -target_skew 0.100

# this should tell the cts engine to not aim for fast transitions and use whatever the liberty file allows. should save power? not sure it is working
#set cts_target_max_transition_time ignore

#set_db cts_use_inverters false
#set_db cts_inverter_cells INV*
ccopt_design

set now [exec date]
puts $log "$now: starting post_cts opt..."
opt_design -post_cts

#route_opt_design -setup
#suspend

set now [exec date]
puts $log "$now: starting route_design..."
route_design

if {[get_db designs .markers] != ""} {
	# meaning routing created drcs
	route_design
	#set_db route_design_detail_auto_stop false
}

set now [exec date]
puts $log "$now: starting post_rout opt..."
# by default, area recovery is disabled during post_route opt. but our designs are very dense... this can help?
set_db opt_post_route_area_reclaim setup_aware

set tp [report_timing -collection]
set slack [get_db $tp .slack]

if {$slack > 0} { 
	opt_power -post_route -force
	# only if there is positive slack
}

while {1} {
	set tp [report_timing -collection]
	set old_slack [get_db $tp .slack]
	opt_design -post_route
	set tp [report_timing -collection]
	set new_slack [get_db $tp .slack]
	if {$new_slack == $old_slack} {
		break
		# got stuck, giving up
	}
	if {$new_slack > 0} {
		break
		# done :)
	}
}

# deflate the design?
# this disturbs the existing solution considerably by bringing the target slack back to 0. it would be wise to call the eval script here and decide whether to move on or not. but that is hard to automate...
#set_db opt_setup_target_slack 0.0
#opt_design -post_route
#opt_design incremental has an odd behavior, it seems to do more harm than good. no longer using it.

close $log

set TIME_end [clock clicks -milliseconds]
set TIME_taken [expr ($TIME_end - $TIME_start)/1000]

source ../scripts/eval.stylus.tcl

echo "time taken for implementation: $TIME_taken"

get_db current_design .bbox.area