riscv
/
OpenFPGA
mirror of https://github.com/lnis-uofu/OpenFPGA.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

Merge pull request #469 from lnis-uofu/dsp_reg 

New test case: DSP block with input and output registers
This commit is contained in:
tangxifan 2022-01-02 21:55:37 -08:00 committed by GitHub
parent 47f9625e88 27caeb1d1f
commit 8b06c8790c
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
7 changed files with 734 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
openfpga_flow/benchmarks/micro_benchmark/mult/mult_2_pipelined/mult_2_pipelined.v Normal file
View File

@ -0,0 +1,32 @@
//-------------------------------------------------------
// Functionality: A 2-bit multiply circuit with pipelines
// Author: Xifan Tang
//-------------------------------------------------------
module mult_2_pipelined(clk, a, b, out);
parameter DATA_WIDTH = 2; /* declare a parameter. default required */
input [DATA_WIDTH - 1 : 0] a, b;
input clk;
output [DATA_WIDTH - 1 : 0] out;
reg [DATA_WIDTH - 1 : 0] a_reg;
reg [DATA_WIDTH - 1 : 0] b_reg;
reg [DATA_WIDTH - 1 : 0] out_reg;
always @(posedge clk) begin
a_reg <= a;
b_reg <= b;
out_reg <= a_reg * b_reg;
out = out_reg;
end
endmodule

4
openfpga_flow/openfpga_arch/README.md
View File

@ -10,6 +10,10 @@ Note that an OpenFPGA architecture can be applied to multiple VPR architecture f
- adder\_chain: If hard adder/carry chain is used inside CLBs
- register\_chain: If shift register chain is used inside CLBs
- scan\_chain: If scan chain testing infrastructure is used inside CLBs
- <wide>\_<frac>\_dsp<dsp\_size>reg: If Digital Signal Processor (DSP) is used or not. If used, the input size should be clarified here.
- The keyword 'wide' is to specify if the DSP spans more than 1 column.
- The keyword 'frac' is to specify if the DSP is fracturable to operate in different modes.
- The keyword 'reg' is to specify if the DSP has input and output registers. If only input or output registers are used, the keyword will be 'regin' or 'regout' respectively.
- mem<mem\_size>: If block RAM (BRAM) is used or not. If used, the memory size should be clarified here. The keyword wide is to specify if the BRAM spanns more than 1 column.
- aib: If the Advanced Interface Bus (AIB) is used in place of some I/Os.
- <bank\|cc\|frame\|standalone>: specify the type of configuration protocol used in the architecture.

205
openfpga_flow/openfpga_arch/k4_N4_40nm_dsp8reg_cc_openfpga.xml Normal file
View File

@ -0,0 +1,205 @@
<!-- Architecture annotation for OpenFPGA framework
This annotation supports the k6_N10_40nm.xml
- General purpose logic block
- K = 6, N = 10, I = 40
- Single mode
- Routing architecture
- L = 4, fc_in = 0.15, fc_out = 0.1
-->
<openfpga_architecture>
<technology_library>
<device_library>
<device_model name="logic" type="transistor">
<lib type="industry" corner="TOP_TT" ref="M" path="${OPENFPGA_PATH}/openfpga_flow/tech/PTM_45nm/45nm.pm"/>
<design vdd="0.9" pn_ratio="2"/>
<pmos name="pch" chan_length="40e-9" min_width="140e-9" variation="logic_transistor_var"/>
<nmos name="nch" chan_length="40e-9" min_width="140e-9" variation="logic_transistor_var"/>
</device_model>
<device_model name="io" type="transistor">
<lib type="academia" ref="M" path="${OPENFPGA_PATH}/openfpga_flow/tech/PTM_45nm/45nm.pm"/>
<design vdd="2.5" pn_ratio="3"/>
<pmos name="pch_25" chan_length="270e-9" min_width="320e-9" variation="io_transistor_var"/>
<nmos name="nch_25" chan_length="270e-9" min_width="320e-9" variation="io_transistor_var"/>
</device_model>
</device_library>
<variation_library>
<variation name="logic_transistor_var" abs_deviation="0.1" num_sigma="3"/>
<variation name="io_transistor_var" abs_deviation="0.1" num_sigma="3"/>
</variation_library>
</technology_library>
<circuit_library>
<circuit_model type="inv_buf" name="INVTX1" prefix="INVTX1" is_default="true">
<design_technology type="cmos" topology="inverter" size="1"/>
<device_technology device_model_name="logic"/>
<port type="input" prefix="in" size="1"/>
<port type="output" prefix="out" size="1"/>
<delay_matrix type="rise" in_port="in" out_port="out">
10e-12
</delay_matrix>
<delay_matrix type="fall" in_port="in" out_port="out">
10e-12
</delay_matrix>
</circuit_model>
<circuit_model type="inv_buf" name="buf4" prefix="buf4" is_default="false">
<design_technology type="cmos" topology="buffer" size="1" num_level="2" f_per_stage="4"/>
<device_technology device_model_name="logic"/>
<port type="input" prefix="in" size="1"/>
<port type="output" prefix="out" size="1"/>
<delay_matrix type="rise" in_port="in" out_port="out">
10e-12
</delay_matrix>
<delay_matrix type="fall" in_port="in" out_port="out">
10e-12
</delay_matrix>
</circuit_model>
<circuit_model type="inv_buf" name="tap_buf4" prefix="tap_buf4" is_default="false">
<design_technology type="cmos" topology="buffer" size="1" num_level="3" f_per_stage="4"/>
<device_technology device_model_name="logic"/>
<port type="input" prefix="in" size="1"/>
<port type="output" prefix="out" size="1"/>
<delay_matrix type="rise" in_port="in" out_port="out">
10e-12
</delay_matrix>
<delay_matrix type="fall" in_port="in" out_port="out">
10e-12
</delay_matrix>
</circuit_model>
<circuit_model type="pass_gate" name="TGATE" prefix="TGATE" is_default="true">
<design_technology type="cmos" topology="transmission_gate" nmos_size="1" pmos_size="2"/>
<device_technology device_model_name="logic"/>
<input_buffer exist="false"/>
<output_buffer exist="false"/>
<port type="input" prefix="in" size="1"/>
<port type="input" prefix="sel" size="1"/>
<port type="input" prefix="selb" size="1"/>
<port type="output" prefix="out" size="1"/>
<delay_matrix type="rise" in_port="in sel selb" out_port="out">
10e-12 5e-12 5e-12
</delay_matrix>
<delay_matrix type="fall" in_port="in sel selb" out_port="out">
10e-12 5e-12 5e-12
</delay_matrix>
</circuit_model>
<circuit_model type="chan_wire" name="chan_segment" prefix="track_seg" is_default="true">
<design_technology type="cmos"/>
<input_buffer exist="false"/>
<output_buffer exist="false"/>
<port type="input" prefix="in" size="1"/>
<port type="output" prefix="out" size="1"/>
<wire_param model_type="pi" R="101" C="22.5e-15" num_level="1"/> <!-- model_type could be T, res_val and cap_val DON'T CARE -->
</circuit_model>
<circuit_model type="wire" name="direct_interc" prefix="direct_interc" is_default="true">
<design_technology type="cmos"/>
<input_buffer exist="false"/>
<output_buffer exist="false"/>
<port type="input" prefix="in" size="1"/>
<port type="output" prefix="out" size="1"/>
<wire_param model_type="pi" R="0" C="0" num_level="1"/> <!-- model_type could be T, res_val cap_val should be defined -->
</circuit_model>
<circuit_model type="mux" name="mux_tree" prefix="mux_tree" dump_structural_verilog="true">
<design_technology type="cmos" structure="tree" add_const_input="true" const_input_val="1"/>
<input_buffer exist="true" circuit_model_name="INVTX1"/>
<output_buffer exist="true" circuit_model_name="INVTX1"/>
<pass_gate_logic circuit_model_name="TGATE"/>
<port type="input" prefix="in" size="1"/>
<port type="output" prefix="out" size="1"/>
<port type="sram" prefix="sram" size="1"/>
</circuit_model>
<circuit_model type="mux" name="mux_tree_tapbuf" prefix="mux_tree_tapbuf" is_default="true" dump_structural_verilog="true">
<design_technology type="cmos" structure="tree" add_const_input="true" const_input_val="1"/>
<input_buffer exist="true" circuit_model_name="INVTX1"/>
<output_buffer exist="true" circuit_model_name="tap_buf4"/>
<pass_gate_logic circuit_model_name="TGATE"/>
<port type="input" prefix="in" size="1"/>
<port type="output" prefix="out" size="1"/>
<port type="sram" prefix="sram" size="1"/>
</circuit_model>
<!--DFF subckt ports should be defined as <D> <Q> <CLK> <RESET> <SET> -->
<circuit_model type="ff" name="DFFSRQ" prefix="DFFSRQ" spice_netlist="${OPENFPGA_PATH}/openfpga_flow/openfpga_cell_library/spice/dff.sp" verilog_netlist="${OPENFPGA_PATH}/openfpga_flow/openfpga_cell_library/verilog/dff.v">
<design_technology type="cmos"/>
<input_buffer exist="true" circuit_model_name="INVTX1"/>
<output_buffer exist="true" circuit_model_name="INVTX1"/>
<port type="input" prefix="D" size="1"/>
<port type="input" prefix="set" lib_name="SET" size="1" is_global="true" default_val="0" is_set="true"/>
<port type="input" prefix="reset" lib_name="RST" size="1" is_global="true" default_val="0" is_reset="true"/>
<port type="output" prefix="Q" size="1"/>
<port type="clock" prefix="clk" lib_name="CK" size="1" is_global="true" default_val="0" />
</circuit_model>
<circuit_model type="lut" name="lut4" prefix="lut4" dump_structural_verilog="true">
<design_technology type="cmos"/>
<input_buffer exist="true" circuit_model_name="INVTX1"/>
<output_buffer exist="true" circuit_model_name="INVTX1"/>
<lut_input_inverter exist="true" circuit_model_name="INVTX1"/>
<lut_input_buffer exist="true" circuit_model_name="buf4"/>
<pass_gate_logic circuit_model_name="TGATE"/>
<port type="input" prefix="in" size="4"/>
<port type="output" prefix="out" size="1"/>
<port type="sram" prefix="sram" size="16"/>
</circuit_model>
<!--Scan-chain DFF subckt ports should be defined as <D> <Q> <Qb> <CLK> <RESET> <SET> -->
<circuit_model type="ccff" name="DFF" prefix="DFF" spice_netlist="${OPENFPGA_PATH}/openfpga_flow/openfpga_cell_library/spice/dff.sp" verilog_netlist="${OPENFPGA_PATH}/openfpga_flow/openfpga_cell_library/verilog/dff.v">
<design_technology type="cmos"/>
<input_buffer exist="true" circuit_model_name="INVTX1"/>
<output_buffer exist="true" circuit_model_name="INVTX1"/>
<port type="input" prefix="D" size="1"/>
<port type="output" prefix="Q" size="1"/>
<port type="output" prefix="QN" size="1"/>
<port type="clock" prefix="prog_clk" lib_name="CK" size="1" is_global="true" default_val="0" is_prog="true"/>
</circuit_model>
<circuit_model type="iopad" name="GPIO" prefix="GPIO" spice_netlist="${OPENFPGA_PATH}/openfpga_flow/openfpga_cell_library/spice/gpio.sp" verilog_netlist="${OPENFPGA_PATH}/openfpga_flow/openfpga_cell_library/verilog/gpio.v">
<design_technology type="cmos"/>
<input_buffer exist="true" circuit_model_name="INVTX1"/>
<output_buffer exist="true" circuit_model_name="INVTX1"/>
<port type="inout" prefix="PAD" size="1" is_global="true" is_io="true" is_data_io="true"/>
<port type="sram" prefix="DIR" size="1" mode_select="true" circuit_model_name="DFF" default_val="1"/>
<port type="input" prefix="outpad" lib_name="A" size="1"/>
<port type="output" prefix="inpad" lib_name="Y" size="1"/>
</circuit_model>
<circuit_model type="hard_logic" name="mult_8x8" prefix="mult_8x8" is_default="true" spice_netlist="${OPENFPGA_PATH}/openfpga_flow/openfpga_cell_library/spice/mult_8x8.sp" verilog_netlist="${OPENFPGA_PATH}/openfpga_flow/openfpga_cell_library/verilog/mult_8x8.v">
<design_technology type="cmos"/>
<input_buffer exist="true" circuit_model_name="INVTX1"/>
<output_buffer exist="true" circuit_model_name="INVTX1"/>
<port type="input" prefix="A" lib_name="A" size="8"/>
<port type="input" prefix="B" lib_name="B" size="8"/>
<port type="output" prefix="Y" lib_name="Y" size="16"/>
</circuit_model>
</circuit_library>
<configuration_protocol>
<organization type="scan_chain" circuit_model_name="DFF"/>
</configuration_protocol>
<connection_block>
<switch name="ipin_cblock" circuit_model_name="mux_tree_tapbuf"/>
</connection_block>
<switch_block>
<switch name="0" circuit_model_name="mux_tree_tapbuf"/>
</switch_block>
<routing_segment>
<segment name="L4" circuit_model_name="chan_segment"/>
</routing_segment>
<pb_type_annotations>
<!-- physical pb_type binding in complex block IO -->
<pb_type name="io" physical_mode_name="physical" idle_mode_name="inpad"/>
<pb_type name="io[physical].iopad" circuit_model_name="GPIO" mode_bits="1"/>
<pb_type name="io[inpad].inpad" physical_pb_type_name="io[physical].iopad" mode_bits="1"/>
<pb_type name="io[outpad].outpad" physical_pb_type_name="io[physical].iopad" mode_bits="0"/>
<!-- End physical pb_type binding in complex block IO -->
<!-- physical pb_type binding in complex block CLB -->
<!-- physical mode will be the default mode if not specified -->
<pb_type name="clb">
<!-- Binding interconnect to circuit models as their physical implementation, if not defined, we use the default model -->
<interconnect name="crossbar" circuit_model_name="mux_tree"/>
</pb_type>
<pb_type name="clb.fle[n1_lut4].ble4.lut4" circuit_model_name="lut4"/>
<pb_type name="clb.fle[n1_lut4].ble4.ff" circuit_model_name="DFFSRQ"/>
<!-- End physical pb_type binding in complex block IO -->
<!-- physical pb_type binding in complex block dsp -->
<pb_type name="mult_8" physical_mode_name="mult_8x8" idle_mode_name="mult_8x8"/>
<!-- Bind the primitive pb_type in the physical mode to a circuit model -->
<pb_type name="mult_8[mult_8x8].mult_8x8_slice.mult_8x8" circuit_model_name="mult_8x8"/>
<pb_type name="mult_8[mult_8x8].mult_8x8_slice.ff_A" circuit_model_name="DFFSRQ"/>
<pb_type name="mult_8[mult_8x8].mult_8x8_slice.ff_B" circuit_model_name="DFFSRQ"/>
<pb_type name="mult_8[mult_8x8].mult_8x8_slice.ff_Y" circuit_model_name="DFFSRQ"/>
</pb_type_annotations>
</openfpga_architecture>

3
openfpga_flow/regression_test_scripts/fpga_verilog_reg_test.sh
View File

@ -53,6 +53,9 @@ run-task fpga_verilog/dsp/multi_mode_mult_16x16 --debug --show_thread_logs
echo -e "Testing Verilog generation with heterogeneous fabric using multi-width 16-bit multi-mode multipliers ";
run-task fpga_verilog/dsp/wide_multi_mode_mult_16x16 --debug --show_thread_logs
echo -e "Testing Verilog generation with heterogeneous fabric using 8-bit single-mode registerable multipliers ";
run-task fpga_verilog/dsp/single_mode_mult_8x8_reg --debug --show_thread_logs
echo -e "Testing Verilog generation with different I/O capacities on each side of an FPGA ";
run-task fpga_verilog/io/multi_io_capacity --debug --show_thread_logs

43
openfpga_flow/tasks/fpga_verilog/dsp/single_mode_mult_8x8_reg/config/task.conf Normal file
View File

@ -0,0 +1,43 @@
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# Configuration file for running experiments
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# timeout_each_job : FPGA Task script splits fpga flow into multiple jobs
# Each job execute fpga_flow script on combination of architecture & benchmark
# timeout_each_job is timeout for each job
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
[GENERAL]
run_engine=openfpga_shell
power_tech_file = ${PATH:OPENFPGA_PATH}/openfpga_flow/tech/PTM_45nm/45nm.xml
power_analysis = false
spice_output=false
verilog_output=true
timeout_each_job = 20*60
fpga_flow=yosys_vpr
[OpenFPGA_SHELL]
openfpga_shell_template=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_shell_scripts/fix_heterogeneous_device_example_script.openfpga
openfpga_arch_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_arch/k4_N4_40nm_dsp8reg_cc_openfpga.xml
openfpga_sim_setting_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_simulation_settings/fixed_sim_openfpga.xml
# Yosys script parameters
yosys_cell_sim_verilog=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_yosys_techlib/k4_frac_N8_tileable_reset_softadder_register_scan_chain_dsp8_nonLR_caravel_io_skywater130nm_cell_sim.v
yosys_dsp_map_verilog=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_yosys_techlib/k4_frac_N8_tileable_reset_softadder_register_scan_chain_dsp8_nonLR_caravel_io_skywater130nm_dsp_map.v
yosys_dsp_map_parameters=-D DSP_A_MAXWIDTH=8 -D DSP_B_MAXWIDTH=8 -D DSP_A_MINWIDTH=2 -D DSP_B_MINWIDTH=2 -D DSP_NAME=mult_8x8
# VPR parameter
openfpga_vpr_device_layout=3x2
[ARCHITECTURES]
arch0=${PATH:OPENFPGA_PATH}/openfpga_flow/vpr_arch/k4_N4_tileable_dsp8reg_40nm.xml
[BENCHMARKS]
bench0=${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/mult/mult_2_pipelined/mult_2_pipelined.v
[SYNTHESIS_PARAM]
bench_yosys_common=${PATH:OPENFPGA_PATH}/openfpga_flow/misc/ys_tmpl_yosys_vpr_dsp_flow.ys
bench_yosys_rewrite_common=${PATH:OPENFPGA_PATH}/openfpga_flow/misc/ys_tmpl_yosys_vpr_flow_with_rewrite.ys;${PATH:OPENFPGA_PATH}/openfpga_flow/misc/ys_tmpl_rewrite_flow.ys
bench0_top = mult_2_pipelined
[SCRIPT_PARAM_MIN_ROUTE_CHAN_WIDTH]
end_flow_with_test=
vpr_fpga_verilog_formal_verification_top_netlist=

6
openfpga_flow/vpr_arch/README.md
View File

@ -12,7 +12,11 @@ Please reveal the following architecture features in the names to help quickly s
- register\_chain: If shift register chain is used inside CLBs
- scan\_chain: If scan chain testing infrastructure is used inside CLBs
- <wide>\_<frac>\_mem<mem\_size>: If block RAM (BRAM) is used or not. If used, the memory size should be clarified here. The keyword 'wide' is to specify if the BRAM spans more than 1 column. The keyword 'frac' is to specify if the BRAM is fracturable to operate in different modes.
- <wide>\_<frac>\_dsp<dsp\_size>: If Digital Signal Processor (DSP) is used or not. If used, the input size should be clarified here. The keyword 'wide' is to specify if the DSP spans more than 1 column. The keyword 'frac' is to specify if the DSP is fracturable to operate in different modes.
- <wide>\_<frac>\_dsp<dsp\_size>reg: If Digital Signal Processor (DSP) is used or not. If used, the input size should be clarified here.
- The keyword 'wide' is to specify if the DSP spans more than 1 column.
- The keyword 'frac' is to specify if the DSP is fracturable to operate in different modes.
- The keyword 'reg' is to specify if the DSP has input and output registers. If only input or output registers are used, the keyword will be 'regin' or 'regout' respectively.
- mem<mem\_size>: If block RAM (BRAM) is used or not. If used, the memory size should be clarified here. The keyword wide is to specify if the BRAM spanns more than 1 column.
- aib: If the Advanced Interface Bus (AIB) is used in place of some I/Os.
- multi\_io\_capacity: If I/O capacity is different on each side of FPGAs.
- reduced\_io: If I/Os only appear a certain or multiple sides of FPGAs

442
openfpga_flow/vpr_arch/k4_N4_tileable_dsp8reg_40nm.xml Normal file
View File

@ -0,0 +1,442 @@
<!--
Architecture with no fracturable LUTs
- 40 nm technology
- General purpose logic block:
K = 4, N = 4
- Routing architecture: L = 4, fc_in = 0.15, Fc_out = 0.1
- DSP block
- 8-bit multiplier
- input and output registers
Details on Modelling:
Based on flagship k6_frac_N10_mem32K_40nm.xml architecture. This architecture has no fracturable LUTs nor any heterogeneous blocks.
Authors: Jason Luu, Jeff Goeders, Vaughn Betz
-->
<architecture>
<!--
ODIN II specific config begins
Describes the types of user-specified netlist blocks (in blif, this corresponds to
".model [type_of_block]") that this architecture supports.
Note: Basic LUTs, I/Os, and flip-flops are not included here as there are
already special structures in blif (.names, .input, .output, and .latch)
that describe them.
-->
<models>
<!-- A virtual model for I/O to be used in the physical mode of io block -->
<model name="io">
<input_ports>
<port name="outpad"/>
</input_ports>
<output_ports>
<port name="inpad"/>
</output_ports>
</model>
<model name="mult_8">
<input_ports>
<port name="A" combinational_sink_ports="Y"/>
<port name="B" combinational_sink_ports="Y"/>
</input_ports>
<output_ports>
<port name="Y"/>
</output_ports>
</model>
</models>
<tiles>
<tile name="io" capacity="8" area="0">
<equivalent_sites>
<site pb_type="io"/>
</equivalent_sites>
<input name="outpad" num_pins="1"/>
<output name="inpad" num_pins="1"/>
<fc in_type="frac" in_val="0.15" out_type="frac" out_val="0.10"/>
<pinlocations pattern="custom">
<loc side="left">io.outpad io.inpad</loc>
<loc side="top">io.outpad io.inpad</loc>
<loc side="right">io.outpad io.inpad</loc>
<loc side="bottom">io.outpad io.inpad</loc>
</pinlocations>
</tile>
<tile name="clb" area="53894">
<equivalent_sites>
<site pb_type="clb"/>
</equivalent_sites>
<input name="I" num_pins="10" equivalent="full"/>
<output name="O" num_pins="4" equivalent="none"/>
<clock name="clk" num_pins="1"/>
<fc in_type="frac" in_val="0.15" out_type="frac" out_val="0.10"/>
<pinlocations pattern="spread"/>
</tile>
<tile name="mult_8" height="2" area="396000">
<equivalent_sites>
<site pb_type="mult_8" pin_mapping="direct"/>
</equivalent_sites>
<input name="a" num_pins="8"/>
<input name="b" num_pins="8"/>
<output name="out" num_pins="16"/>
<clock name="clk" num_pins="1"/>
<fc in_type="frac" in_val="0.15" out_type="frac" out_val="0.10"/>
<!-- Highly recommand to customize pin location when direct connection is used!!! -->
<!-- pinlocations are designed to spread pin on 4 sides evenly -->
<pinlocations pattern="custom">
<loc side="left">mult_8.a[0:3] mult_8.b[0:3] mult_8.out[0:7]</loc>
<loc side="top">mult_8.clk</loc>
<loc side="right">mult_8.a[4:7] mult_8.b[4:7] mult_8.out[8:15]</loc>
<loc side="bottom"></loc>
</pinlocations>
</tile>
</tiles>
<!-- ODIN II specific config ends -->
<!-- Physical descriptions begin -->
<layout tileable="true">
<auto_layout aspect_ratio="1.0">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<perimeter type="io" priority="100"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</auto_layout>
<fixed_layout name="2x2" width="4" height="4">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<perimeter type="io" priority="100"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</fixed_layout>
<fixed_layout name="3x2" width="5" height="4">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<perimeter type="io" priority="100"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
<!--Column of 'memory' with 'EMPTY' blocks wherever a 'memory' does not fit. Vertical offset by 1 for perimeter.-->
<col type="mult_8" startx="2" starty="1" repeatx="8" priority="20"/>
<col type="EMPTY" startx="2" repeatx="8" starty="1" priority="19"/>
</fixed_layout>
</layout>
<device>
<!-- VB & JL: Using Ian Kuon's transistor sizing and drive strength data for routing, at 40 nm. Ian used BPTM
models. We are modifying the delay values however, to include metal C and R, which allows more architecture
experimentation. We are also modifying the relative resistance of PMOS to be 1.8x that of NMOS
(vs. Ian's 3x) as 1.8x lines up with Jeff G's data from a 45 nm process (and is more typical of
45 nm in general). I'm upping the Rmin_nmos from Ian's just over 6k to nearly 9k, and dropping
RminW_pmos from 18k to 16k to hit this 1.8x ratio, while keeping the delays of buffers approximately
lined up with Stratix IV.
We are using Jeff G.'s capacitance data for 45 nm (in tech/ptm_45nm).
Jeff's tables list C in for transistors with widths in multiples of the minimum feature size (45 nm).
The minimum contactable transistor is 2.5 * 45 nm, so I need to multiply drive strength sizes in this file
by 2.5x when looking up in Jeff's tables.
The delay values are lined up with Stratix IV, which has an architecture similar to this
proposed FPGA, and which is also 40 nm
C_ipin_cblock: input capacitance of a track buffer, which VPR assumes is a single-stage
4x minimum drive strength buffer. -->
<sizing R_minW_nmos="8926" R_minW_pmos="16067"/>
<!-- The grid_logic_tile_area below will be used for all blocks that do not explicitly set their own (non-routing)
area; set to 0 since we explicitly set the area of all blocks currently in this architecture file.
-->
<area grid_logic_tile_area="0"/>
<chan_width_distr>
<x distr="uniform" peak="1.000000"/>
<y distr="uniform" peak="1.000000"/>
</chan_width_distr>
<switch_block type="wilton" fs="3"/>
<connection_block input_switch_name="ipin_cblock"/>
</device>
<switchlist>
<!-- VB: the mux_trans_size and buf_size data below is in minimum width transistor *areas*, assuming the purple
book area formula. This means the mux transistors are about 5x minimum drive strength.
We assume the first stage of the buffer is 3x min drive strength to be reasonable given the large
mux transistors, and this gives a reasonable stage ratio of a bit over 5x to the second stage. We assume
the n and p transistors in the first stage are equal-sized to lower the buffer trip point, since it's fed
by a pass transistor mux. We can then reverse engineer the buffer second stage to hit the specified
buf_size (really buffer area) - 16.2x minimum drive nmos and 1.8*16.2 = 29.2x minimum drive.
I then took the data from Jeff G.'s PTM modeling of 45 nm to get the Cin (gate of first stage) and Cout
(diff of second stage) listed below. Jeff's models are in tech/ptm_45nm, and are in min feature multiples.
The minimum contactable transistor is 2.5 * 45 nm, so I need to multiply the drive strength sizes above by
2.5x when looking up in Jeff's tables.
Finally, we choose a switch delay (58 ps) that leads to length 4 wires having a delay equal to that of SIV of 126 ps.
This also leads to the switch being 46% of the total wire delay, which is reasonable. -->
<switch type="mux" name="0" R="551" Cin=".77e-15" Cout="4e-15" Tdel="58e-12" mux_trans_size="2.630740" buf_size="27.645901"/>
<!--switch ipin_cblock resistance set to yeild for 4x minimum drive strength buffer-->
<switch type="mux" name="ipin_cblock" R="2231.5" Cout="0." Cin="1.47e-15" Tdel="7.247000e-11" mux_trans_size="1.222260" buf_size="auto"/>
</switchlist>
<segmentlist>
<!--- VB & JL: using ITRS metal stack data, 96 nm half pitch wires, which are intermediate metal width/space.
With the 96 nm half pitch, such wires would take 60 um of height, vs. a 90 nm high (approximated as square) Stratix IV tile so this seems
reasonable. Using a tile length of 90 nm, corresponding to the length of a Stratix IV tile if it were square. -->
<segment name="L4" freq="1.000000" length="4" type="unidir" Rmetal="101" Cmetal="22.5e-15">
<mux name="0"/>
<sb type="pattern">1 1 1 1 1</sb>
<cb type="pattern">1 1 1 1</cb>
</segment>
</segmentlist>
<complexblocklist>
<!-- Define I/O pads begin -->
<!-- Capacity is a unique property of I/Os, it is the maximum number of I/Os that can be placed at the same (X,Y) location on the FPGA -->
<!-- Not sure of the area of an I/O (varies widely), and it's not relevant to the design of the FPGA core, so we're setting it to 0. -->
<pb_type name="io">
<input name="outpad" num_pins="1"/>
<output name="inpad" num_pins="1"/>
<!-- A mode denotes the physical implementation of an I/O
This mode will be not packable but is mainly used for fabric verilog generation
-->
<mode name="physical" disable_packing="true">
<pb_type name="iopad" blif_model=".subckt io" num_pb="1">
<input name="outpad" num_pins="1"/>
<output name="inpad" num_pins="1"/>
</pb_type>
<interconnect>
<direct name="outpad" input="io.outpad" output="iopad.outpad">
<delay_constant max="1.394e-11" in_port="io.outpad" out_port="iopad.outpad"/>
</direct>
<direct name="inpad" input="iopad.inpad" output="io.inpad">
<delay_constant max="4.243e-11" in_port="iopad.inpad" out_port="io.inpad"/>
</direct>
</interconnect>
</mode>
<!-- IOs can operate as either inputs or outputs.
Delays below come from Ian Kuon. They are small, so they should be interpreted as
the delays to and from registers in the I/O (and generally I/Os are registered
today and that is when you timing analyze them.
-->
<mode name="inpad">
<pb_type name="inpad" blif_model=".input" num_pb="1">
<output name="inpad" num_pins="1"/>
</pb_type>
<interconnect>
<direct name="inpad" input="inpad.inpad" output="io.inpad">
<delay_constant max="4.243e-11" in_port="inpad.inpad" out_port="io.inpad"/>
</direct>
</interconnect>
</mode>
<mode name="outpad">
<pb_type name="outpad" blif_model=".output" num_pb="1">
<input name="outpad" num_pins="1"/>
</pb_type>
<interconnect>
<direct name="outpad" input="io.outpad" output="outpad.outpad">
<delay_constant max="1.394e-11" in_port="io.outpad" out_port="outpad.outpad"/>
</direct>
</interconnect>
</mode>
<!-- Every input pin is driven by 15% of the tracks in a channel, every output pin is driven by 10% of the tracks in a channel -->
<!-- IOs go on the periphery of the FPGA, for consistency,
make it physically equivalent on all sides so that only one definition of I/Os is needed.
If I do not make a physically equivalent definition, then I need to define 4 different I/Os, one for each side of the FPGA
-->
<!-- Place I/Os on the sides of the FPGA -->
<power method="ignore"/>
</pb_type>
<!-- Define I/O pads ends -->
<!-- Define general purpose logic block (CLB) begin -->
<!--- Area calculation: Total Stratix IV tile area is about 8100 um^2, and a minimum width transistor
area is 60 L^2 yields a tile area of 84375 MWTAs.
Routing at W=300 is 30481 MWTAs, leaving us with a total of 53000 MWTAs for logic block area
This means that only 37% of our area is in the general routing, and 63% is inside the logic
block. Note that the crossbar / local interconnect is considered part of the logic block
area in this analysis. That is a lower proportion of of routing area than most academics
assume, but note that the total routing area really includes the crossbar, which would push
routing area up significantly, we estimate into the ~70% range.
-->
<pb_type name="clb">
<input name="I" num_pins="10" equivalent="full"/>
<output name="O" num_pins="4" equivalent="none"/>
<clock name="clk" num_pins="1"/>
<!-- Describe basic logic element.
Each basic logic element has a 4-LUT that can be optionally registered
-->
<pb_type name="fle" num_pb="4">
<input name="in" num_pins="4"/>
<output name="out" num_pins="1"/>
<clock name="clk" num_pins="1"/>
<!-- 4-LUT mode definition begin -->
<mode name="n1_lut4">
<!-- Define 4-LUT mode -->
<pb_type name="ble4" num_pb="1">
<input name="in" num_pins="4"/>
<output name="out" num_pins="1"/>
<clock name="clk" num_pins="1"/>
<!-- Define LUT -->
<pb_type name="lut4" blif_model=".names" num_pb="1" class="lut">
<input name="in" num_pins="4" port_class="lut_in"/>
<output name="out" num_pins="1" port_class="lut_out"/>
<!-- LUT timing using delay matrix -->
<delay_matrix type="max" in_port="lut4.in" out_port="lut4.out">
261e-12
261e-12
261e-12
261e-12
</delay_matrix>
</pb_type>
<!-- Define flip-flop -->
<pb_type name="ff" blif_model=".latch" num_pb="1" class="flipflop">
<input name="D" num_pins="1" port_class="D"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="clk" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="ff.D" clock="clk"/>
<T_clock_to_Q max="124e-12" port="ff.Q" clock="clk"/>
</pb_type>
<interconnect>
<direct name="direct1" input="ble4.in" output="lut4[0:0].in"/>
<direct name="direct2" input="lut4.out" output="ff.D">
<!-- Advanced user option that tells CAD tool to find LUT+FF pairs in netlist -->
<pack_pattern name="ble4" in_port="lut4.out" out_port="ff.D"/>
</direct>
<direct name="direct3" input="ble4.clk" output="ff.clk"/>
<mux name="mux1" input="ff.Q lut4.out" output="ble4.out">
<!-- LUT to output is faster than FF to output on a Stratix IV -->
<delay_constant max="25e-12" in_port="lut4.out" out_port="ble4.out"/>
<delay_constant max="45e-12" in_port="ff.Q" out_port="ble4.out"/>
</mux>
</interconnect>
</pb_type>
<interconnect>
<direct name="direct1" input="fle.in" output="ble4.in"/>
<direct name="direct2" input="ble4.out" output="fle.out[0:0]"/>
<direct name="direct3" input="fle.clk" output="ble4.clk"/>
</interconnect>
</mode>
<!-- 6-LUT mode definition end -->
</pb_type>
<interconnect>
<!-- We use a full crossbar to get logical equivalence at inputs of CLB
The delays below come from Stratix IV. the delay through a connection block
input mux + the crossbar in Stratix IV is 167 ps. We already have a 72 ps
delay on the connection block input mux (modeled by Ian Kuon), so the remaining
delay within the crossbar is 95 ps.
The delays of cluster feedbacks in Stratix IV is 100 ps, when driven by a LUT.
Since all our outputs LUT outputs go to a BLE output, and have a delay of
25 ps to do so, we subtract 25 ps from the 100 ps delay of a feedback
to get the part that should be marked on the crossbar. -->
<complete name="crossbar" input="clb.I fle[3:0].out" output="fle[3:0].in">
<delay_constant max="95e-12" in_port="clb.I" out_port="fle[3:0].in"/>
<delay_constant max="75e-12" in_port="fle[3:0].out" out_port="fle[3:0].in"/>
</complete>
<complete name="clks" input="clb.clk" output="fle[3:0].clk">
</complete>
<!-- This way of specifying direct connection to clb outputs is important because this architecture uses automatic spreading of opins.
By grouping to output pins in this fashion, if a logic block is completely filled by 6-LUTs,
then the outputs those 6-LUTs take get evenly distributed across all four sides of the CLB instead of clumped on two sides (which is what happens with a more
naive specification).
-->
<direct name="clbouts1" input="fle[3:0].out" output="clb.O"/>
</interconnect>
<!-- Every input pin is driven by 15% of the tracks in a channel, every output pin is driven by 10% of the tracks in a channel -->
<!-- Place this general purpose logic block in any unspecified column -->
</pb_type>
<!-- Define general purpose logic block (CLB) ends -->
<!-- Define 8-bit multiplier with input and output registers begin -->
<pb_type name="mult_8">
<input name="a" num_pins="8"/>
<input name="b" num_pins="8"/>
<clock name="clk" num_pins="1"/>
<output name="out" num_pins="16"/>
<mode name="mult_8x8">
<pb_type name="mult_8x8_slice" num_pb="1">
<input name="A_cfg" num_pins="8"/>
<input name="B_cfg" num_pins="8"/>
<output name="OUT_cfg" num_pins="16"/>
<clock name="clk" num_pins="1"/>
<pb_type name="mult_8x8" blif_model=".subckt mult_8" num_pb="1">
<input name="A" num_pins="8"/>
<input name="B" num_pins="8"/>
<output name="Y" num_pins="16"/>
<delay_constant max="1.523e-9" min="0.776e-9" in_port="mult_8x8.A" out_port="mult_8x8.Y"/>
<delay_constant max="1.523e-9" min="0.776e-9" in_port="mult_8x8.B" out_port="mult_8x8.Y"/>
</pb_type>
<pb_type name="ff_A" blif_model=".latch" num_pb="8" class="flipflop">
<input name="D" num_pins="1" port_class="D"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="clk" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="ff_A.D" clock="clk"/>
<T_clock_to_Q max="124e-12" port="ff_A.Q" clock="clk"/>
</pb_type>
<pb_type name="ff_B" blif_model=".latch" num_pb="8" class="flipflop">
<input name="D" num_pins="1" port_class="D"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="clk" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="ff_B.D" clock="clk"/>
<T_clock_to_Q max="124e-12" port="ff_B.Q" clock="clk"/>
</pb_type>
<pb_type name="ff_Y" blif_model=".latch" num_pb="16" class="flipflop">
<input name="D" num_pins="1" port_class="D"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="clk" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="ff_Y.D" clock="clk"/>
<T_clock_to_Q max="124e-12" port="ff_Y.Q" clock="clk"/>
</pb_type>
<interconnect>
<mux name="a2a0" input="mult_8x8_slice.A_cfg[0] ff_A[0].Q" output="mult_8x8.A[0]"/>
<mux name="a2a1" input="mult_8x8_slice.A_cfg[1] ff_A[1].Q" output="mult_8x8.A[1]"/>
<mux name="a2a2" input="mult_8x8_slice.A_cfg[2] ff_A[2].Q" output="mult_8x8.A[2]"/>
<mux name="a2a3" input="mult_8x8_slice.A_cfg[3] ff_A[3].Q" output="mult_8x8.A[3]"/>
<mux name="a2a4" input="mult_8x8_slice.A_cfg[4] ff_A[4].Q" output="mult_8x8.A[4]"/>
<mux name="a2a5" input="mult_8x8_slice.A_cfg[5] ff_A[5].Q" output="mult_8x8.A[5]"/>
<mux name="a2a6" input="mult_8x8_slice.A_cfg[6] ff_A[6].Q" output="mult_8x8.A[6]"/>
<mux name="a2a7" input="mult_8x8_slice.A_cfg[7] ff_A[7].Q" output="mult_8x8.A[7]"/>
<direct name="a2ff" input="mult_8x8_slice.A_cfg[7:0]" output="ff_A[7:0].D"/>
<mux name="b2b0" input="mult_8x8_slice.B_cfg[0] ff_B[0].Q" output="mult_8x8.B[0]"/>
<mux name="b2b1" input="mult_8x8_slice.B_cfg[1] ff_B[1].Q" output="mult_8x8.B[1]"/>
<mux name="b2b2" input="mult_8x8_slice.B_cfg[2] ff_B[2].Q" output="mult_8x8.B[2]"/>
<mux name="b2b3" input="mult_8x8_slice.B_cfg[3] ff_B[3].Q" output="mult_8x8.B[3]"/>
<mux name="b2b4" input="mult_8x8_slice.B_cfg[4] ff_B[4].Q" output="mult_8x8.B[4]"/>
<mux name="b2b5" input="mult_8x8_slice.B_cfg[5] ff_B[5].Q" output="mult_8x8.B[5]"/>
<mux name="b2b6" input="mult_8x8_slice.B_cfg[6] ff_B[6].Q" output="mult_8x8.B[6]"/>
<mux name="b2b7" input="mult_8x8_slice.B_cfg[7] ff_B[7].Q" output="mult_8x8.B[7]"/>
<direct name="b2ff" input="mult_8x8_slice.B_cfg[7:0]" output="ff_B[7:0].D"/>
<mux name="out2out0" input="mult_8x8.Y[0] ff_Y[0].Q" output="mult_8x8_slice.OUT_cfg[0]"/>
<mux name="out2out1" input="mult_8x8.Y[1] ff_Y[1].Q" output="mult_8x8_slice.OUT_cfg[1]"/>
<mux name="out2out2" input="mult_8x8.Y[2] ff_Y[2].Q" output="mult_8x8_slice.OUT_cfg[2]"/>
<mux name="out2out3" input="mult_8x8.Y[3] ff_Y[3].Q" output="mult_8x8_slice.OUT_cfg[3]"/>
<mux name="out2out4" input="mult_8x8.Y[4] ff_Y[4].Q" output="mult_8x8_slice.OUT_cfg[4]"/>
<mux name="out2out5" input="mult_8x8.Y[5] ff_Y[5].Q" output="mult_8x8_slice.OUT_cfg[5]"/>
<mux name="out2out6" input="mult_8x8.Y[6] ff_Y[6].Q" output="mult_8x8_slice.OUT_cfg[6]"/>
<mux name="out2out7" input="mult_8x8.Y[7] ff_Y[7].Q" output="mult_8x8_slice.OUT_cfg[7]"/>
<mux name="out2out8" input="mult_8x8.Y[8] ff_Y[8].Q" output="mult_8x8_slice.OUT_cfg[8]"/>
<mux name="out2out9" input="mult_8x8.Y[9] ff_Y[9].Q" output="mult_8x8_slice.OUT_cfg[9]"/>
<mux name="out2out10" input="mult_8x8.Y[10] ff_Y[10].Q" output="mult_8x8_slice.OUT_cfg[10]"/>
<mux name="out2out11" input="mult_8x8.Y[11] ff_Y[11].Q" output="mult_8x8_slice.OUT_cfg[11]"/>
<mux name="out2out12" input="mult_8x8.Y[12] ff_Y[12].Q" output="mult_8x8_slice.OUT_cfg[12]"/>
<mux name="out2out13" input="mult_8x8.Y[13] ff_Y[13].Q" output="mult_8x8_slice.OUT_cfg[13]"/>
<mux name="out2out14" input="mult_8x8.Y[14] ff_Y[14].Q" output="mult_8x8_slice.OUT_cfg[14]"/>
<mux name="out2out15" input="mult_8x8.Y[15] ff_Y[15].Q" output="mult_8x8_slice.OUT_cfg[15]"/>
<direct name="out2ff" input="mult_8x8.Y[15:0]" output="ff_Y[15:0].D"/>
<complete name="clk_ff_A" input="mult_8x8_slice.clk" output="ff_A.clk"/>
<complete name="clk_ff_B" input="mult_8x8_slice.clk" output="ff_B.clk"/>
<complete name="clk_ff_Y" input="mult_8x8_slice.clk" output="ff_Y.clk"/>
</interconnect>
<power method="pin-toggle">
<port name="A_cfg" energy_per_toggle="2.13e-12"/>
<port name="B_cfg" energy_per_toggle="2.13e-12"/>
<static_power power_per_instance="0.0"/>
</power>
</pb_type>
<interconnect>
<!-- Stratix IV input delay of 207ps is conservative for this architecture because this architecture does not have an input crossbar in the multiplier.
Subtract 72.5 ps delay, which is already in the connection block input mux, leading
to a 134 ps delay.
The interconnect difference for DSP blocks is 0.5523, which leads to a minimum delay of 74 ps
-->
<direct name="a2a" input="mult_8.a" output="mult_8x8_slice.A_cfg">
<delay_constant max="134e-12" min="74e-12" in_port="mult_8.a" out_port="mult_8x8_slice.A_cfg"/>
</direct>
<direct name="b2b" input="mult_8.b" output="mult_8x8_slice.B_cfg">
<delay_constant max="134e-12" min="74e-12" in_port="mult_8.b" out_port="mult_8x8_slice.B_cfg"/>
</direct>
<direct name="out2out" input="mult_8x8_slice.OUT_cfg" output="mult_8.out">
<delay_constant max="1.93e-9" min="74e-12" in_port="mult_8x8_slice.OUT_cfg" out_port="mult_8.out"/>
</direct>
<complete name="clk" input="mult_8.clk" output="mult_8x8_slice.clk"/>
</interconnect>
</mode>
<!-- Place this multiplier block every 8 columns from (and including) the sixth column -->
<power method="sum-of-children"/>
</pb_type>
<!-- Define fracturable multiplier end -->
</complexblocklist>
</architecture>