mirror of
https://github.com/openhwgroup/cvw.git
synced 2025-04-22 21:08:08 -04:00
Merge branch 'main' of https://github.com/davidharrishmc/riscv-wally
This commit is contained in:
Madeleine Masser-Frye
commit
fcf1cb794e
18 changed files with 1797 additions and 1741 deletions
|
|
@ -2,39 +2,56 @@
|
|||
# Expanded and developed by dtorres@hmc.edu
|
||||
# Compile Embench for Wally
|
||||
|
||||
all: build sim
|
||||
all: sim size
|
||||
|
||||
allClean: clean all
|
||||
|
||||
build: buildspeed buildsize
|
||||
|
||||
# uses the build_all.py python file to build the tests in addins/embench-iot/bd_speed/ optimized for speed
|
||||
buildspeed:
|
||||
../../addins/embench-iot/build_all.py --builddir=bd_speed --arch riscv32 --chip generic --board rv32wallyverilog --ldflags="-nostartfiles ../../../config/riscv32/boards/rv32wallyverilog/startup/crt0.S" --cflags="-nostartfiles"
|
||||
../../addins/embench-iot/build_all.py --builddir=bd_speed --arch riscv32 --chip generic --board rv32wallyverilog --ldflags="-nostartfiles ../../../config/riscv32/boards/rv32wallyverilog/startup/crt0.S" --cflags="-O2 -nostartfiles"
|
||||
find ../../addins/embench-iot/bd_speed/ -type f ! -name "*.*" | while read f; do cp "$$f" "$$f.elf"; done
|
||||
|
||||
# uses the build_all.py python file to build the tests in addins/embench-iot/bd_speed/ optimized for size
|
||||
buildsize:
|
||||
../../addins/embench-iot/build_all.py --builddir=bd_size --arch riscv32 --chip generic --board rv32wallyverilog --ldflags="-nostdlib -nostartfiles ../../../config/riscv32/boards/rv32wallyverilog/startup/dummy.S" --cflags="-msave-restore" --dummy-libs="libgcc libm libc crt0"
|
||||
../../addins/embench-iot/build_all.py --builddir=bd_size --arch riscv32 --chip generic --board rv32wallyverilog --ldflags="-nostdlib -nostartfiles ../../../config/riscv32/boards/rv32wallyverilog/startup/dummy.S" --cflags="-Os -msave-restore" --dummy-libs="libgcc libm libc crt0"
|
||||
|
||||
sim: modelSimBuild speed
|
||||
# builds dependencies, then launches modelsim and finally runs python wrapper script to present results
|
||||
sim: modelsim_build_memfile modelsim_run speed
|
||||
|
||||
# vsim:
|
||||
# cd ../../pipelined/regression/
|
||||
# vsim -c -do "do wally-pipelined-batch.do rv32gc embench"
|
||||
# cd ../../benchmarks/embench/
|
||||
|
||||
modelSimBuild: buildspeed objdump
|
||||
find ../../addins/embench-iot/bd_speed/ -type f -name "*.elf" | while read f; do riscv64-unknown-elf-elf2hex --bit-width 32 --input "$$f" --output "$$f.memfile"; done
|
||||
find ../../addins/embench-iot/bd_speed/ -type f -name "*.elf.objdump" | while read f; do extractFunctionRadix.sh $$f; done
|
||||
|
||||
size:
|
||||
../../addins/embench-iot/benchmark_size.py --builddir=bd_size
|
||||
|
||||
speed:
|
||||
../../addins/embench-iot/benchmark_speed.py --builddir=bd_speed --target-module run_wally --cpu-mhz=1
|
||||
# launches modelsim to simulate tests on wally
|
||||
modelsim_run:
|
||||
(cd ../../pipelined/regression/ && vsim -c -do "do wally-pipelined-batch.do rv32gc embench")
|
||||
cd ../../benchmarks/embench/
|
||||
|
||||
# builds the objdump based on the compiled c elf files
|
||||
objdump: buildspeed
|
||||
find ../../addins/embench-iot/bd_speed/ -type f -name "*.elf" | while read f; do riscv64-unknown-elf-objdump -S -D "$$f" > "$$f.objdump"; done
|
||||
|
||||
# build memfiles, objdump.lab and objdump.addr files
|
||||
modelsim_build_memfile: objdump
|
||||
find ../../addins/embench-iot/bd_speed/ -type f -name "*.elf" | while read f; do riscv64-unknown-elf-elf2hex --bit-width 32 --input "$$f" --output "$$f.memfile"; done
|
||||
find ../../addins/embench-iot/bd_speed/ -type f -name "*.elf.objdump" | while read f; do extractFunctionRadix.sh $$f; done
|
||||
|
||||
# builds the tests for speed, runs them on spike and then launches python script to present results
|
||||
# note that the speed python script benchmark_speed.py can get confused if there's both a .output file created from spike and modelsim
|
||||
# you'll need to manually remove one of the two .output files, or run make clean
|
||||
spike: buildspeed spikecmd speed
|
||||
|
||||
# command to run spike on all of the benchmarks
|
||||
spike_run: buildspeed
|
||||
find ../../addins/embench-iot/bd_speed/ -type f -name "*.elf" | while read f; do spike --isa=rv32imac +signature=$$f.spike.output +signature-granularity=4 $$f; done
|
||||
|
||||
# python wrapper to present results of embench size benchmark
|
||||
size: buildsize
|
||||
../../addins/embench-iot/benchmark_size.py --builddir=bd_size
|
||||
|
||||
# python wrapper to present results of embench speed benchmark
|
||||
speed:
|
||||
../../addins/embench-iot/benchmark_speed.py --builddir=bd_speed --target-module run_wally --cpu-mhz=1
|
||||
|
||||
# deletes all files
|
||||
clean:
|
||||
rm -rf ../../addins/embench-iot/bd_speed/
|
||||
rm -rf ../../addins/embench-iot/bd_size/
|
||||
|
|
@ -42,13 +59,4 @@ clean:
|
|||
allclean: clean
|
||||
rm -rf ../../addins/embench-iot/logs/
|
||||
|
||||
# std:
|
||||
# ../../addins/embench-iot/build_all.py --builddir=bd_std --arch riscv32 --chip generic --board rv32wallyverilog --cc riscv64-unknown-elf-gcc --cflags="-v -c -O2 -ffunction-sections -march=rv32imac -mabi=ilp32" --ldflags="-Wl,-gc-sections -v -march=rv32imac -mabi=ilp32 ../../../../../benchmarks/embench/tohost.S -T../../../config/riscv32/boards/rv32wallyverilog/link.ld" --user-libs="-lm"
|
||||
# riscv64-unknown-elf-objdump -D ../../addins/embench-iot/bd_std/src/aha-mont64/aha-mont64 > ../../addins/embench-iot/bd_std/src/aha-mont64/aha-mont64.objdump
|
||||
# --dummy-libs="libgcc libm libc"
|
||||
# --cflags "-O2 -g -nostartfiles"
|
||||
# ../../addins/embench-iot/build_all.py --arch riscv32 --chip generic --board rv32wallyverilog --cc riscv64-unknown-elf-gcc --cflags="-c -Os -ffunction-sections -nostdlib -march=rv32imac -mabi=ilp32" --ldflags="-Wl,-gc-sections -nostdlib -march=rv32imac -mabi=ilp32 -T../../../config/riscv32/boards/rv32wallyverilog/link.ld" --dummy-libs="libgcc libm libc"
|
||||
# --user-libs="-lm"
|
||||
# riscv64-unknown-elf-gcc -O2 -g -nostartfiles -I/home/harris/riscv-wally/addins/embench-iot/support -I/home/harris/riscv-wally/addins/embench-iot/config/riscv32/boards/ri5cyverilator -I/home/harris/riscv-wally/addins/embench-iot/config/riscv32/chips/generic -I/home/harris/riscv-wally/addins/embench-iot/config/riscv32 -DCPU_MHZ=1 -DWARMUP_HEAT=1 -o main.o /home/harris/riscv-wally/addins/embench-iot/support/main.c
|
||||
|
||||
# find ../../addins/embench-iot/bd_speed/ -type f -name "*.elf.objdump.lab" | while read f; do grep -n "begin_signature" $f | cut -f1 -d:
|
||||
# riscv64-unknown-elf-gcc -O2 -g -nostartfiles -I/home/harris/riscv-wally/addins/embench-iot/support -I/home/harris/riscv-wally/addins/embench-iot/config/riscv32/boards/ri5cyverilator -I/home/harris/riscv-wally/addins/embench-iot/config/riscv32/chips/generic -I/home/harris/riscv-wally/addins/embench-iot/config/riscv32 -DCPU_MHZ=1 -DWARMUP_HEAT=1 -o main.o /home/harris/riscv-wally/addins/embench-iot/support/main.c
|
||||
|
|
@ -327,7 +327,7 @@ connect_debug_port u_ila_0/probe72 [get_nets [list wallypipelinedsoc/core/hzu/BP
|
|||
create_debug_port u_ila_0 probe
|
||||
set_property port_width 1 [get_debug_ports u_ila_0/probe73]
|
||||
set_property PROBE_TYPE DATA_AND_TRIGGER [get_debug_ports u_ila_0/probe73]
|
||||
connect_debug_port u_ila_0/probe73 [get_nets [list wallypipelinedsoc/core/hzu/CSRWritePendingDEM ]]
|
||||
connect_debug_port u_ila_0/probe73 [get_nets [list wallypipelinedsoc/core/hzu/CSRWriteFencePendingDEM ]]
|
||||
|
||||
create_debug_port u_ila_0 probe
|
||||
set_property port_width 1 [get_debug_ports u_ila_0/probe74]
|
||||
|
|
@ -402,7 +402,7 @@ connect_debug_port u_ila_0/probe87 [get_nets [list wallypipelinedsoc/core/hzu/Br
|
|||
create_debug_port u_ila_0 probe
|
||||
set_property port_width 1 [get_debug_ports u_ila_0/probe88]
|
||||
set_property PROBE_TYPE DATA_AND_TRIGGER [get_debug_ports u_ila_0/probe88]
|
||||
connect_debug_port u_ila_0/probe88 [get_nets [list wallypipelinedsoc/core/hzu/InvalidateICacheM ]]
|
||||
connect_debug_port u_ila_0/probe88 [get_nets [list {wallypipelinedsoc/uncore/uart.uart/u/RXerrIP} ]]
|
||||
|
||||
create_debug_port u_ila_0 probe
|
||||
set_property port_width 1 [get_debug_ports u_ila_0/probe89]
|
||||
|
|
@ -433,7 +433,8 @@ connect_debug_port u_ila_0/probe93 [get_nets [list wallypipelinedsoc/core/hzu/St
|
|||
create_debug_port u_ila_0 probe
|
||||
set_property port_width 1 [get_debug_ports u_ila_0/probe94]
|
||||
set_property PROBE_TYPE DATA_AND_TRIGGER [get_debug_ports u_ila_0/probe94]
|
||||
connect_debug_port u_ila_0/probe94 [get_nets [list wallypipelinedsoc/core/hzu/FlushF ]]
|
||||
connect_debug_port u_ila_0/probe94 [get_nets [list {wallypipelinedsoc/uncore/uart.uart/u/RXerrIP} ]]
|
||||
|
||||
|
||||
create_debug_port u_ila_0 probe
|
||||
set_property port_width 1 [get_debug_ports u_ila_0/probe95]
|
||||
|
|
@ -835,8 +836,4 @@ set_property port_width 4 [get_debug_ports u_ila_0/probe171]
|
|||
set_property PROBE_TYPE DATA_AND_TRIGGER [get_debug_ports u_ila_0/probe171]
|
||||
connect_debug_port u_ila_0/probe171 [get_nets [list {wallypipelinedsoc/uncore/uart.uart/u/rxfifotail[0]} {wallypipelinedsoc/uncore/uart.uart/u/rxfifotail[1]} {wallypipelinedsoc/uncore/uart.uart/u/rxfifotail[2]} {wallypipelinedsoc/uncore/uart.uart/u/rxfifotail[3]} ]]
|
||||
|
||||
create_debug_port u_ila_0 probe
|
||||
set_property port_width 1 [get_debug_ports u_ila_0/probe172]
|
||||
set_property PROBE_TYPE DATA_AND_TRIGGER [get_debug_ports u_ila_0/probe172]
|
||||
connect_debug_port u_ila_0/probe172 [get_nets [list {wallypipelinedsoc/uncore/uart.uart/u/RXerrIP} ]]
|
||||
|
||||
|
|
|
|||
|
|
@ -38,7 +38,8 @@
|
|||
`define IEEE754 0
|
||||
|
||||
// MISA RISC-V configuration per specification
|
||||
`define MISA (32'h00000104 | 1 << 5 | 1 << 3 | 1 << 18 | 1 << 20 | 1 << 12 | 1 << 0 )
|
||||
// ZYXWVUTSRQPONMLKJIHGFEDCBA
|
||||
`define MISA 32'b0000000000101000001000100101101
|
||||
`define ZICSR_SUPPORTED 1
|
||||
`define ZIFENCEI_SUPPORTED 1
|
||||
`define COUNTERS 32
|
||||
|
|
|
|||
|
|
@ -51,43 +51,45 @@
|
|||
`define PMPCFG_ENTRIES (`PMP_ENTRIES/8)
|
||||
|
||||
// Floating point constants for Quad, Double, Single, and Half precisions
|
||||
`define Q_LEN 128
|
||||
`define Q_NE 15
|
||||
`define Q_NF 112
|
||||
`define Q_BIAS 16383
|
||||
`define D_LEN 64
|
||||
`define D_NE 11
|
||||
`define D_NF 52
|
||||
`define D_BIAS 1023
|
||||
`define S_LEN 32
|
||||
`define S_NE 8
|
||||
`define S_NF 23
|
||||
`define S_BIAS 127
|
||||
`define H_LEN 16
|
||||
`define H_NE 5
|
||||
`define H_NF 10
|
||||
`define H_BIAS 15
|
||||
`define Q_LEN 32'd128
|
||||
`define Q_NE 32'd15
|
||||
`define Q_NF 32'd112
|
||||
`define Q_BIAS 32'd16383
|
||||
`define D_LEN 32'd64
|
||||
`define D_NE 32'd11
|
||||
`define D_NF 32'd52
|
||||
`define D_BIAS 32'd1023
|
||||
`define D_FMT 32'd1
|
||||
`define S_LEN 32'd32
|
||||
`define S_NE 32'd8
|
||||
`define S_NF 32'd23
|
||||
`define S_BIAS 32'd127
|
||||
`define S_FMT 32'd1
|
||||
`define H_LEN 32'd16
|
||||
`define H_NE 32'd5
|
||||
`define H_NF 32'd10
|
||||
`define H_BIAS 32'd15
|
||||
|
||||
// Floating point length FLEN and number of exponent (NE) and fraction (NF) bits
|
||||
`define FLEN (`Q_SUPPORTED ? `Q_LEN : `D_SUPPORTED ? `D_LEN : `F_SUPPORTED ? `S_LEN : `H_LEN)
|
||||
`define NE (`Q_SUPPORTED ? `Q_NE : `D_SUPPORTED ? `D_NE : `F_SUPPORTED ? `S_NE : `H_NE)
|
||||
`define NF (`Q_SUPPORTED ? `Q_NF : `D_SUPPORTED ? `D_NF : `F_SUPPORTED ? `S_NF : `H_NF)
|
||||
`define FMT (`Q_SUPPORTED ? 3 : `D_SUPPORTED ? 1 : `F_SUPPORTED ? 0 : 2)
|
||||
`define FMT (`Q_SUPPORTED ? 2'd3 : `D_SUPPORTED ? 2'd1 : `F_SUPPORTED ? 2'd0 : 2'd2)
|
||||
`define BIAS (`Q_SUPPORTED ? `Q_BIAS : `D_SUPPORTED ? `D_BIAS : `F_SUPPORTED ? `S_BIAS : `H_BIAS)
|
||||
|
||||
// Floating point constants needed for FPU paramerterization
|
||||
`define FPSIZES ((3)'(`Q_SUPPORTED)+(3)'(`D_SUPPORTED)+(3)'(`F_SUPPORTED)+(3)'(`ZFH_SUPPORTED))
|
||||
`define FMTBITS (((`FPSIZES==3'b011)|(`FPSIZES==3'b100)) ? 2 : 1)
|
||||
`define LEN1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? `D_LEN : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? `S_LEN : `H_LEN)
|
||||
`define NE1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? `D_NE : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? `S_NE : `H_NE)
|
||||
`define NF1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? `D_NF : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? `S_NF : `H_NF)
|
||||
`define FMT1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? 1 : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? 0 : 2)
|
||||
`define BIAS1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? `D_BIAS : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? `S_BIAS : `H_BIAS)
|
||||
`define LEN2 ((`F_SUPPORTED & (`LEN1 != `S_LEN)) ? `S_LEN : `H_LEN)
|
||||
`define FPSIZES ((32)'(`Q_SUPPORTED)+(32)'(`D_SUPPORTED)+(32)'(`F_SUPPORTED)+(32)'(`ZFH_SUPPORTED))
|
||||
`define FMTBITS ((`FPSIZES>=3)+1)
|
||||
`define LEN1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? `D_LEN : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? `S_LEN : `H_LEN)
|
||||
`define NE1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? `D_NE : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? `S_NE : `H_NE)
|
||||
`define NF1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? `D_NF : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? `S_NF : `H_NF)
|
||||
`define FMT1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? 2'd1 : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? 2'd0 : 2'd2)
|
||||
`define BIAS1 ((`D_SUPPORTED & (`FLEN != `D_LEN)) ? `D_BIAS : (`F_SUPPORTED & (`FLEN != `S_LEN)) ? `S_BIAS : `H_BIAS)
|
||||
`define LEN2 ((`F_SUPPORTED & (`LEN1 != `S_LEN)) ? `S_LEN : `H_LEN)
|
||||
`define NE2 ((`F_SUPPORTED & (`LEN1 != `S_LEN)) ? `S_NE : `H_NE)
|
||||
`define NF2 ((`F_SUPPORTED & (`LEN1 != `S_LEN)) ? `S_NF : `H_NF)
|
||||
`define FMT2 ((`F_SUPPORTED & (`LEN1 != `S_LEN)) ? 0 : 2)
|
||||
`define BIAS2 ((`F_SUPPORTED & (`LEN1 != `S_LEN)) ? `S_BIAS : `H_BIAS)
|
||||
`define NF2 ((`F_SUPPORTED & (`LEN1 != `S_LEN)) ? `S_NF : `H_NF)
|
||||
`define FMT2 ((`F_SUPPORTED & (`LEN1 != `S_LEN)) ? 2'd0 : 2'd2)
|
||||
`define BIAS2 ((`F_SUPPORTED & (`LEN1 != `S_LEN)) ? `S_BIAS : `H_BIAS)
|
||||
|
||||
// Disable spurious Verilator warnings
|
||||
|
||||
|
|
|
|||
|
|
@ -5,7 +5,7 @@ export PATH=$PATH:/usr/local/bin/
|
|||
verilator=`which verilator`
|
||||
|
||||
basepath=$(dirname $0)/..
|
||||
for config in rv64fp rv32e rv64gc rv32gc rv32ic; do
|
||||
for config in rv64fp rv64fpquad rv32e rv64gc rv32gc rv32ic; do
|
||||
echo "$config linting..."
|
||||
if !($verilator --lint-only "$@" --top-module wallypipelinedsoc "-I$basepath/config/shared" "-I$basepath/config/$config" $basepath/src/*/*.sv $basepath/src/*/*/*.sv --relative-includes); then
|
||||
echo "Exiting after $config lint due to errors or warnings"
|
||||
|
|
|
|||
|
|
@ -1,2 +1,2 @@
|
|||
vsim -do "do wally-pipelined.do rv32e imperas64d"
|
||||
vsim -do "do wally-pipelined.do rv32gc arch32f"
|
||||
|
||||
|
|
|
|||
|
|
@ -122,7 +122,7 @@ module fctrl (
|
|||
else if (`FPSIZES == 2)begin
|
||||
logic [1:0] FmtTmp;
|
||||
assign FmtTmp = (FResultSelD == 2'b00) ? {~Funct3D[1], ~(Funct3D[1]^Funct3D[0])} : ((Funct7D[6:3] == 4'b0100)&OpD[4]) ? Rs2D[1:0] : Funct7D[1:0];
|
||||
assign FmtD = `FMT == FmtTmp;
|
||||
assign FmtD = (`FMT == FmtTmp);
|
||||
end
|
||||
else if (`FPSIZES == 3|`FPSIZES == 4)
|
||||
assign FmtD = (FResultSelD == 2'b00) ? {~Funct3D[1], ~(Funct3D[1]^Funct3D[0])} : ((Funct7D[6:3] == 4'b0100)&OpD[4]) ? Rs2D[1:0] : Funct7D[1:0];
|
||||
|
|
|
|||
|
|
@ -2,6 +2,7 @@
|
|||
`include "wally-config.vh"
|
||||
// largest length in IEU/FPU
|
||||
`define LGLEN ((`NF<`XLEN) ? `XLEN : `NF)
|
||||
`define LOGLGLEN $unsigned($clog2(`LGLEN+1))
|
||||
|
||||
module fcvt (
|
||||
input logic XSgnE, // input's sign
|
||||
|
|
@ -42,7 +43,7 @@ module fcvt (
|
|||
logic [`XLEN-1:0] TrimInt; // integer trimmed to the correct size
|
||||
logic [`LGLEN-1:0] LzcIn; // input to the Leading Zero Counter (priority encoder)
|
||||
logic [`NE:0] CalcExp; // the calculated expoent
|
||||
logic [$clog2(`LGLEN+1)-1:0] ShiftAmt; // how much to shift by
|
||||
logic [`LOGLGLEN-1:0] ShiftAmt; // how much to shift by
|
||||
logic [`LGLEN+`NF:0] ShiftIn; // number to be shifted
|
||||
logic ResDenormUf;// does the result underflow or is denormalized
|
||||
logic ResUf; // does the result underflow
|
||||
|
|
@ -72,7 +73,7 @@ module fcvt (
|
|||
logic Int64; // is the integer 64 bits?
|
||||
logic IntToFp; // is the opperation an int->fp conversion?
|
||||
logic ToInt; // is the opperation an fp->int conversion?
|
||||
logic [$clog2(`LGLEN+1)-1:0] ZeroCnt; // output from the LZC
|
||||
logic [`LOGLGLEN-1:0] ZeroCnt; // output from the LZC
|
||||
|
||||
|
||||
// seperate OpCtrl for code readability
|
||||
|
|
@ -143,9 +144,9 @@ module fcvt (
|
|||
// - only shift fp -> fp if the intital value is denormalized
|
||||
// - this is a problem because the input to the lzc was the fraction rather than the mantissa
|
||||
// - rather have a few and-gates than an extra bit in the priority encoder??? *** is this true?
|
||||
assign ShiftAmt = ToInt ? CalcExp[$clog2(`LGLEN+1)-1:0]&{$clog2(`LGLEN+1){~CalcExp[`NE]}} :
|
||||
ResDenormUf&~IntToFp ? ($clog2(`LGLEN+1))'(`NF-1)+CalcExp[$clog2(`LGLEN+1)-1:0] :
|
||||
(ZeroCnt+1)&{$clog2(`LGLEN+1){XDenormE|IntToFp}};
|
||||
assign ShiftAmt = ToInt ? CalcExp[`LOGLGLEN-1:0]&{`LOGLGLEN{~CalcExp[`NE]}} :
|
||||
ResDenormUf&~IntToFp ? (`LOGLGLEN)'(`NF-1)+CalcExp[`LOGLGLEN-1:0] :
|
||||
(ZeroCnt+1)&{`LOGLGLEN{XDenormE|IntToFp}};
|
||||
|
||||
// shift
|
||||
// fp -> int: | `XLEN zeros | Mantissa | 0's if nessisary | << CalcExp
|
||||
|
|
@ -261,34 +262,34 @@ module fcvt (
|
|||
// - shift left to normilize (-1-ZeroCnt)
|
||||
// - newBias to make the biased exponent
|
||||
//
|
||||
assign CalcExp = {1'b0, OldExp} - (`NE+1)'(`BIAS) + {2'b0, NewBias} - {{`NE{1'b0}}, XDenormE|IntToFp} - {{`NE-$clog2(`LGLEN+1)+1{1'b0}}, (ZeroCnt&{$clog2(`LGLEN+1){XDenormE|IntToFp}})};
|
||||
assign CalcExp = {1'b0, OldExp} - (`NE+1)'(`BIAS) + {2'b0, NewBias} - {{`NE{1'b0}}, XDenormE|IntToFp} - {{`NE-`LOGLGLEN+1{1'b0}}, (ZeroCnt&{`LOGLGLEN{XDenormE|IntToFp}})};
|
||||
// find if the result is dnormal or underflows
|
||||
// - if Calculated expoenent is 0 or negitive (and the input/result is not exactaly 0)
|
||||
// - can't underflow an integer to Fp conversion
|
||||
assign ResDenormUf = (~|CalcExp | CalcExp[`NE])&~XZeroE&~IntToFp;
|
||||
// choose the negative of the fraction size
|
||||
if (`FPSIZES == 1) begin
|
||||
assign ResNegNF = -`NF;
|
||||
assign ResNegNF = -($clog2(`NF)+1)'(`NF);
|
||||
|
||||
end else if (`FPSIZES == 2) begin
|
||||
assign ResNegNF = OutFmt ? -`NF : -`NF1;
|
||||
assign ResNegNF = OutFmt ? -($clog2(`NF)+1)'(`NF) : -($clog2(`NF)+1)'(`NF1);
|
||||
|
||||
end else if (`FPSIZES == 3) begin
|
||||
always_comb
|
||||
case (OutFmt)
|
||||
`FMT: ResNegNF = -`NF;
|
||||
`FMT1: ResNegNF = -`NF1;
|
||||
`FMT2: ResNegNF = -`NF2;
|
||||
`FMT: ResNegNF = -($clog2(`NF)+1)'(`NF);
|
||||
`FMT1: ResNegNF = -($clog2(`NF)+1)'(`NF1);
|
||||
`FMT2: ResNegNF = -($clog2(`NF)+1)'(`NF2);
|
||||
default: ResNegNF = 1'bx;
|
||||
endcase
|
||||
|
||||
end else if (`FPSIZES == 4) begin
|
||||
always_comb
|
||||
case (OutFmt)
|
||||
2'h3: ResNegNF = -`Q_NF;
|
||||
2'h1: ResNegNF = -`D_NF;
|
||||
2'h0: ResNegNF = -`S_NF;
|
||||
2'h2: ResNegNF = -`H_NF;
|
||||
2'h3: ResNegNF = -($clog2(`NF)+1)'(`Q_NF);
|
||||
2'h1: ResNegNF = -($clog2(`NF)+1)'(`D_NF);
|
||||
2'h0: ResNegNF = -($clog2(`NF)+1)'(`S_NF);
|
||||
2'h2: ResNegNF = -($clog2(`NF)+1)'(`H_NF);
|
||||
endcase
|
||||
end
|
||||
// determine if the result underflows ??? -> fp
|
||||
|
|
@ -453,10 +454,10 @@ module fcvt (
|
|||
|
||||
// find the maximum exponent (the exponent and larger overflows)
|
||||
if (`FPSIZES == 1) begin
|
||||
assign MaxExp = ToInt ? Int64 ? 65 : 33 : {`NE{1'b1}};
|
||||
assign MaxExp = ToInt ? Int64 ? (`NE)'(65) : (`NE)'(33) : {`NE{1'b1}};
|
||||
|
||||
end else if (`FPSIZES == 2) begin
|
||||
assign MaxExp = ToInt ? Int64 ? 65 : 33 :
|
||||
assign MaxExp = ToInt ? Int64 ? (`NE)'($unsigned(65)) : (`NE)'($unsigned(33)) :
|
||||
OutFmt ? {`NE{1'b1}} : {{`NE-`NE1{1'b0}}, {`NE1{1'b1}}};
|
||||
|
||||
end else if (`FPSIZES == 3) begin
|
||||
|
|
@ -476,7 +477,7 @@ module fcvt (
|
|||
MaxExpFp = 1'bx;
|
||||
end
|
||||
endcase
|
||||
assign MaxExp = ToInt ? Int64 ? 65 : 33 : MaxExpFp;
|
||||
assign MaxExp = ToInt ? Int64 ? (`NE)'(65) : (`NE)'(33) : MaxExpFp;
|
||||
|
||||
end else if (`FPSIZES == 4) begin
|
||||
logic [`NE-1:0] MaxExpFp;
|
||||
|
|
@ -495,7 +496,7 @@ module fcvt (
|
|||
MaxExpFp = {{`Q_NE-`H_NE{1'b0}}, {`H_NE{1'b1}}};
|
||||
end
|
||||
endcase
|
||||
assign MaxExp = ToInt ? Int64 ? 65 : 33 : MaxExpFp;
|
||||
assign MaxExp = ToInt ? Int64 ? (`NE)'(65) : (`NE)'(33) : MaxExpFp;
|
||||
end
|
||||
|
||||
// if the result exponent is larger then the maximum possible exponent
|
||||
|
|
|
|||
|
|
@ -81,7 +81,7 @@ module fma(
|
|||
|
||||
// E/M pipeline registers
|
||||
flopenrc #(3*`NF+6) EMRegFma2(clk, reset, FlushM, ~StallM, SumE, SumM);
|
||||
flopenrc #(13) EMRegFma3(clk, reset, FlushM, ~StallM, ProdExpE, ProdExpM);
|
||||
flopenrc #(`NE+2) EMRegFma3(clk, reset, FlushM, ~StallM, ProdExpE, ProdExpM);
|
||||
flopenrc #($clog2(3*`NF+7)+8) EMRegFma4(clk, reset, FlushM, ~StallM,
|
||||
{AddendStickyE, KillProdE, InvZE, NormCntE, NegSumE, ZSgnEffE, PSgnE, FOpCtrlE[2]&~FOpCtrlE[1]&~FOpCtrlE[0], ZDenormE},
|
||||
{AddendStickyM, KillProdM, InvZM, NormCntM, NegSumM, ZSgnEffM, PSgnM, Mult, ZDenormM});
|
||||
|
|
@ -237,7 +237,7 @@ module align(
|
|||
// - positive means the product is larger, so shift Z right
|
||||
// *** can we use ProdExpE instead of XExp/YExp to save an adder? DH 5/12/22
|
||||
// KP- yes we used ProdExpE originally but we did this for timing
|
||||
assign AlignCnt = XZeroE|YZeroE ? -1 : {2'b0, XExpE} + {2'b0, YExpE} - {2'b0, (`NE)'(`BIAS)} + `NF+3 - {2'b0, ZExpE};
|
||||
assign AlignCnt = XZeroE|YZeroE ? -(`NE+2)'($unsigned(1)) : {2'b0, XExpE} + {2'b0, YExpE} - {2'b0, (`NE)'(`BIAS)} + (`NE+2)'(`NF)+3 - {2'b0, ZExpE};
|
||||
|
||||
// Defualt Addition without shifting
|
||||
// | 54'b0 | 106'b(product) | 2'b0 |
|
||||
|
|
@ -320,7 +320,7 @@ module add(
|
|||
|
||||
// Do the addition
|
||||
// - calculate a positive and negitive sum in parallel
|
||||
assign PreSum = AlignedAddendInv + {55'b0, ProdManKilled, 2'b0} + {{3*`NF+6{1'b0}}, InvZE};
|
||||
assign PreSum = AlignedAddendInv + {{`NF+3{1'b0}}, ProdManKilled, 2'b0} + {{3*`NF+6{1'b0}}, InvZE};
|
||||
assign NegPreSum = XZeroE|YZeroE|KillProdE ? {1'b0, AlignedAddendE} : {1'b0, AlignedAddendE} + {{`NF+3{1'b1}}, ~ProdManKilled, 2'b0} + {(3*`NF+7)'(4)};
|
||||
|
||||
// Is the sum negitive
|
||||
|
|
@ -543,7 +543,7 @@ module normalize(
|
|||
assign SumZero = ~(|SumM);
|
||||
|
||||
// calculate the sum's exponent
|
||||
assign SumExpTmpTmp = KillProdM ? {2'b0, ZExpM[`NE-1:1], ZExpM[0]&~ZDenormM} : ProdExpM + -({4'b0, NormCntM} + 1 - (`NF+4));
|
||||
assign SumExpTmpTmp = KillProdM ? {2'b0, ZExpM[`NE-1:1], ZExpM[0]&~ZDenormM} : ProdExpM + -({{`NE+2-$unsigned($clog2(3*`NF+7)){1'b0}}, NormCntM} + 1 - (`NE+2)'(`NF+4));
|
||||
|
||||
//convert the sum's exponent into the propper percision
|
||||
if (`FPSIZES == 1) begin
|
||||
|
|
@ -556,8 +556,8 @@ module normalize(
|
|||
always_comb begin
|
||||
case (FmtM)
|
||||
`FMT: SumExpTmp = SumExpTmpTmp;
|
||||
`FMT1: SumExpTmp = (SumExpTmpTmp-`BIAS+`BIAS1)&{`NE+2{|SumExpTmpTmp}};
|
||||
`FMT2: SumExpTmp = (SumExpTmpTmp-`BIAS+`BIAS2)&{`NE+2{|SumExpTmpTmp}};
|
||||
`FMT1: SumExpTmp = (SumExpTmpTmp-(`NE+2)'(`BIAS)+(`NE+2)'(`BIAS1))&{`NE+2{|SumExpTmpTmp}};
|
||||
`FMT2: SumExpTmp = (SumExpTmpTmp-(`NE+2)'(`BIAS)+(`NE+2)'(`BIAS2))&{`NE+2{|SumExpTmpTmp}};
|
||||
default: SumExpTmp = `NE+2'bx;
|
||||
endcase
|
||||
end
|
||||
|
|
@ -566,9 +566,9 @@ module normalize(
|
|||
always_comb begin
|
||||
case (FmtM)
|
||||
2'h3: SumExpTmp = SumExpTmpTmp;
|
||||
2'h1: SumExpTmp = (SumExpTmpTmp-`BIAS+`D_BIAS)&{`NE+2{|SumExpTmpTmp}};
|
||||
2'h0: SumExpTmp = (SumExpTmpTmp-`BIAS+`S_BIAS)&{`NE+2{|SumExpTmpTmp}};
|
||||
2'h2: SumExpTmp = (SumExpTmpTmp-`BIAS+`H_BIAS)&{`NE+2{|SumExpTmpTmp}};
|
||||
2'h1: SumExpTmp = (SumExpTmpTmp-(`NE+2)'(`BIAS)+(`NE+2)'(`D_BIAS))&{`NE+2{|SumExpTmpTmp}};
|
||||
2'h0: SumExpTmp = (SumExpTmpTmp-(`NE+2)'(`BIAS)+(`NE+2)'(`S_BIAS))&{`NE+2{|SumExpTmpTmp}};
|
||||
2'h2: SumExpTmp = (SumExpTmpTmp-(`NE+2)'(`BIAS)+(`NE+2)'(`H_BIAS))&{`NE+2{|SumExpTmpTmp}};
|
||||
endcase
|
||||
end
|
||||
|
||||
|
|
@ -674,7 +674,7 @@ module normalize(
|
|||
|
||||
// Determine sum's exponent
|
||||
// if plus1 If plus2 if said denorm but norm plus 1 if said denorm but norm plus 2
|
||||
assign SumExp = (SumExpTmp+{12'b0, LZAPlus1&~KillProdM}+{11'b0, LZAPlus2&~KillProdM, 1'b0}+{12'b0, ~ResultDenorm&PreResultDenorm&~KillProdM}+{12'b0, &SumExpTmp&SumShifted[3*`NF+6]&~KillProdM}) & {`NE+2{~(SumZero|ResultDenorm)}};
|
||||
assign SumExp = (SumExpTmp+{{`NE+1{1'b0}}, LZAPlus1&~KillProdM}+{{`NE{1'b0}}, LZAPlus2&~KillProdM, 1'b0}+{{`NE+1{1'b0}}, ~ResultDenorm&PreResultDenorm&~KillProdM}+{{`NE+1{1'b0}}, &SumExpTmp&SumShifted[3*`NF+6]&~KillProdM}) & {`NE+2{~(SumZero|ResultDenorm)}};
|
||||
// recalculate if the result is denormalized
|
||||
assign ResultDenorm = PreResultDenorm&~SumShifted[3*`NF+6]&~SumShifted[3*`NF+7];
|
||||
|
||||
|
|
|
|||
|
|
@ -107,7 +107,7 @@ module fpu (
|
|||
logic FOpCtrlQ;
|
||||
|
||||
// result and flag signals
|
||||
logic [`FLEN-1:0] FDivResM, FDivResW; // divide/squareroot result
|
||||
logic [63:0] FDivResM, FDivResW; // divide/squareroot result
|
||||
logic [4:0] FDivFlgM; // divide/squareroot flags
|
||||
logic [`FLEN-1:0] FMAResM, FMAResW; // FMA/multiply result
|
||||
logic [4:0] FMAFlgM; // FMA/multiply result
|
||||
|
|
@ -125,7 +125,7 @@ module fpu (
|
|||
logic [`FLEN-1:0] FPUResultW; // final FP result being written to the FP register
|
||||
// other signals
|
||||
logic FDivSqrtDoneE; // is divide done
|
||||
logic [`FLEN-1:0] DivInput1E, DivInput2E; // inputs to divide/squareroot unit
|
||||
logic [63:0] DivInput1E, DivInput2E; // inputs to divide/squareroot unit
|
||||
logic load_preload; // enable for FF on fpdivsqrt
|
||||
logic [`FLEN-1:0] AlignedSrcAE; // align SrcA to the floating point format
|
||||
logic [`FLEN-1:0] BoxedZeroE; // Zero value for Z for multiplication, with NaN boxing if needed
|
||||
|
|
@ -184,11 +184,11 @@ module fpu (
|
|||
generate
|
||||
if(`FPSIZES == 1) assign BoxedZeroE = 0;
|
||||
else if(`FPSIZES == 2)
|
||||
mux2 #(`FLEN) fmulzeromux ({{`FLEN-`LEN1{1'b1}}, {`FLEN-`LEN1{1'b0}}}, (`FLEN)'(0), FmtE, BoxedZeroE); // NaN boxing zeroes
|
||||
mux2 #(`FLEN) fmulzeromux ({{`FLEN-`LEN1{1'b1}}, {`LEN1{1'b0}}}, (`FLEN)'(0), FmtE, BoxedZeroE); // NaN boxing zeroes
|
||||
else if(`FPSIZES == 3 | `FPSIZES == 4)
|
||||
mux4 #(`FLEN) fmulzeromux ({{`FLEN-`S_LEN{1'b1}}, (`FLEN-`S_LEN)'(0)},
|
||||
{{`FLEN-`D_LEN{1'b1}}, (`FLEN-`D_LEN)'(0)},
|
||||
{{`FLEN-`H_LEN{1'b1}}, (`FLEN-`H_LEN)'(0)},
|
||||
mux4 #(`FLEN) fmulzeromux ({{`FLEN-`S_LEN{1'b1}}, {`S_LEN{1'b0}}},
|
||||
{{`FLEN-`D_LEN{1'b1}}, {`D_LEN{1'b0}}},
|
||||
{{`FLEN-`H_LEN{1'b1}}, {`H_LEN{1'b0}}},
|
||||
(`FLEN)'(0), FmtE, BoxedZeroE); // NaN boxing zeroes
|
||||
endgenerate
|
||||
|
||||
|
|
@ -218,17 +218,27 @@ module fpu (
|
|||
.FMAFlgM, .FMAResM);
|
||||
|
||||
// fpdivsqrt using Goldschmidt's iteration
|
||||
flopenrc #(`FLEN) reg_input1 (.d({XSgnE, XExpE, XManE[51:0]}), .q(DivInput1E),
|
||||
if(`FLEN == 64) begin
|
||||
flopenrc #(64) reg_input1 (.d({FSrcXE[63:0]}), .q(DivInput1E),
|
||||
.clear(FDivSqrtDoneE), .en(load_preload),
|
||||
.reset(reset), .clk(clk));
|
||||
flopenrc #(`FLEN) reg_input2 (.d({YSgnE, YExpE, YManE[51:0]}), .q(DivInput2E),
|
||||
flopenrc #(64) reg_input2 (.d({FSrcYE[63:0]}), .q(DivInput2E),
|
||||
.clear(FDivSqrtDoneE), .en(load_preload),
|
||||
.reset(reset), .clk(clk));
|
||||
flopenrc #(8+int'(`FMTBITS-1)) reg_input3 (.d({XNaNE, YNaNE, XInfE, YInfE, XZeroE, YZeroE, FmtE, FOpCtrlE[0]}),
|
||||
end
|
||||
else if (`FLEN == 32) begin
|
||||
flopenrc #(64) reg_input1 (.d({32'b0, FSrcXE[31:0]}), .q(DivInput1E),
|
||||
.clear(FDivSqrtDoneE), .en(load_preload),
|
||||
.reset(reset), .clk(clk));
|
||||
flopenrc #(64) reg_input2 (.d({32'b0, FSrcYE[31:0]}), .q(DivInput2E),
|
||||
.clear(FDivSqrtDoneE), .en(load_preload),
|
||||
.reset(reset), .clk(clk));
|
||||
end
|
||||
flopenrc #(8) reg_input3 (.d({XNaNE, YNaNE, XInfE, YInfE, XZeroE, YZeroE, FmtE[0], FOpCtrlE[0]}),
|
||||
.q({XNaNQ, YNaNQ, XInfQ, YInfQ, XZeroQ, YZeroQ, FmtQ, FOpCtrlQ}),
|
||||
.clear(FDivSqrtDoneE), .en(load_preload),
|
||||
.reset(reset), .clk(clk));
|
||||
fpdiv_pipe fdivsqrt (.op1(DivInput1E), .op2(DivInput2E), .rm(FrmE[1:0]), .op_type(FOpCtrlQ),
|
||||
fpdiv_pipe fdivsqrt (.op1(DivInput1E[63:0]), .op2(DivInput2E[63:0]), .rm(FrmE[1:0]), .op_type(FOpCtrlQ),
|
||||
.reset, .clk(clk), .start(FDivStartE), .P(~FmtQ), .OvEn(1'b1), .UnEn(1'b1),
|
||||
.XNaNQ, .YNaNQ, .XInfQ, .YInfQ, .XZeroQ, .YZeroQ, .load_preload,
|
||||
.FDivBusyE, .done(FDivSqrtDoneE), .AS_Result(FDivResM), .Flags(FDivFlgM));
|
||||
|
|
@ -244,7 +254,8 @@ module fpu (
|
|||
// data to be stored in memory - to IEU
|
||||
// - FP uses NaN-blocking format
|
||||
// - if there are any unsused bits the most significant bits are filled with 1s
|
||||
assign FWriteDataE = FSrcYE[`XLEN-1:0];
|
||||
if (`FLEN>`XLEN) assign FWriteDataE = FSrcYE[`XLEN-1:0];
|
||||
else assign FWriteDataE = {{`XLEN-`FLEN{FSrcYE[`FLEN-1]}}, FSrcYE};
|
||||
|
||||
// NaN Block SrcA
|
||||
generate
|
||||
|
|
@ -262,8 +273,12 @@ module fpu (
|
|||
mux4 #(5) FFlgMux(5'b0, 5'b0, {CmpNVE, 4'b0}, CvtFlgE, FResSelE, FFlgE);
|
||||
|
||||
// select the result that may be written to the integer register - to IEU
|
||||
mux4 #(`XLEN) IntResMux(CmpResE[`XLEN-1:0], FSrcXE[`XLEN-1:0], ClassResE,
|
||||
CvtIntResE, FIntResSelE, FIntResE);
|
||||
if (`FLEN>`XLEN)
|
||||
mux4 #(`XLEN) IntResMux(CmpResE[`XLEN-1:0], FSrcXE[`XLEN-1:0], ClassResE,
|
||||
CvtIntResE, FIntResSelE, FIntResE);
|
||||
else
|
||||
mux4 #(`XLEN) IntResMux({{`XLEN-`FLEN{CmpResE[`FLEN-1:0]}}, CmpResE}, {{`XLEN-`FLEN{FSrcXE[`FLEN-1:0]}}, FSrcXE}, ClassResE,
|
||||
CvtIntResE, FIntResSelE, FIntResE);
|
||||
// *** DH 5/25/22: CvtRes will move to mem stage. Premux in execute to save area, then make sure stalls are ok
|
||||
// *** make sure the fpu matches the chapter diagram
|
||||
|
||||
|
|
@ -290,7 +305,7 @@ module fpu (
|
|||
|
||||
// M/W pipe registers
|
||||
flopenrc #(`FLEN) MWRegFma(clk, reset, FlushW, ~StallW, FMAResM, FMAResW);
|
||||
flopenrc #(`FLEN) MWRegDiv(clk, reset, FlushW, ~StallW, FDivResM, FDivResW);
|
||||
flopenrc #(64) MWRegDiv(clk, reset, FlushW, ~StallW, FDivResM, FDivResW);
|
||||
flopenrc #(`FLEN) MWRegClass(clk, reset, FlushW, ~StallW, FResM, FResW);
|
||||
flopenrc #(4+int'(`FMTBITS-1)) MWCtrlReg(clk, reset, FlushW, ~StallW,
|
||||
{FRegWriteM, FResultSelM, FmtM},
|
||||
|
|
@ -313,5 +328,6 @@ module fpu (
|
|||
endgenerate
|
||||
|
||||
// select the result to be written to the FP register
|
||||
mux4 #(`FLEN) FPUResultMux (ReadResW, FMAResW, FDivResW, FResW, FResultSelW, FPUResultW);
|
||||
if(`FLEN>=64)
|
||||
mux4 #(`FLEN) FPUResultMux (ReadResW, FMAResW, {{`FLEN-64{1'b0}},FDivResW}, FResW, FResultSelW, FPUResultW);
|
||||
endmodule // fpu
|
||||
|
|
|
|||
|
|
@ -1,11 +1,11 @@
|
|||
`include "wally-config.vh"
|
||||
|
||||
module unpack (
|
||||
input logic [$signed(`FLEN)-$signed(1):0] X, Y, Z, // inputs from register file
|
||||
input logic [$signed(`FMTBITS)-$signed(1):0] FmtE, // format signal 00 - single 01 - double 11 - quad 10 - half
|
||||
input logic [`FLEN-1:0] X, Y, Z, // inputs from register file
|
||||
input logic [`FMTBITS-1:0] FmtE, // format signal 00 - single 01 - double 11 - quad 10 - half
|
||||
output logic XSgnE, YSgnE, ZSgnE, // sign bits of XYZ
|
||||
output logic [$signed(`NE)-$signed(1):0] XExpE, YExpE, ZExpE, // exponents of XYZ (converted to largest supported precision)
|
||||
output logic [$signed(`NF):0] XManE, YManE, ZManE, // mantissas of XYZ (converted to largest supported precision)
|
||||
output logic [`NE-1:0] XExpE, YExpE, ZExpE, // exponents of XYZ (converted to largest supported precision)
|
||||
output logic [`NF:0] XManE, YManE, ZManE, // mantissas of XYZ (converted to largest supported precision)
|
||||
output logic XNaNE, YNaNE, ZNaNE, // is XYZ a NaN
|
||||
output logic XSNaNE, YSNaNE, ZSNaNE, // is XYZ a signaling NaN
|
||||
output logic XDenormE, ZDenormE, // is XYZ denormalized
|
||||
|
|
@ -14,7 +14,7 @@ module unpack (
|
|||
output logic XExpMaxE // does X have the maximum exponent (NaN or Inf)
|
||||
);
|
||||
|
||||
logic [$signed(`NF)-$signed(1):0] XFracE, YFracE, ZFracE; //Fraction of XYZ
|
||||
logic [`NF-1:0] XFracE, YFracE, ZFracE; //Fraction of XYZ
|
||||
logic XExpNonZero, YExpNonZero, ZExpNonZero; // is the exponent of XYZ non-zero
|
||||
logic XFracZero, YFracZero, ZFracZero; // is the fraction zero
|
||||
logic YExpMaxE, ZExpMaxE; // is the exponent all 1s
|
||||
|
|
|
|||
|
|
@ -17,7 +17,7 @@ if [file exists work] {
|
|||
}
|
||||
vlib work
|
||||
|
||||
vlog +incdir+../config/rv64gc +incdir+../config/shared srt.sv testbench.sv ../src/generic/flop/flop*.sv ../src/generic/mux.sv
|
||||
vlog +incdir+../config/rv64gc +incdir+../config/shared srt.sv testbench.sv ../src/generic/flop/flop*.sv ../src/generic/mux.sv ../src/generic/lzc.sv
|
||||
vopt +acc work.testbench -o workopt
|
||||
vsim workopt
|
||||
|
||||
|
|
|
|||
|
|
@ -30,7 +30,11 @@
|
|||
|
||||
`include "wally-config.vh"
|
||||
|
||||
module srt #(parameter Nf=52) (
|
||||
`define DIVLEN ((`NF<(`XLEN+1)) ? (`XLEN + 1) : `NF)
|
||||
`define EXTRAFRACBITS ((`NF<(`XLEN+1)) ? (`XLEN - `NF + 1) : 0)
|
||||
`define EXTRAINTBITS ((`NF<(`XLEN+1)) ? 0 : (`NF - `XLEN))
|
||||
|
||||
module srt (
|
||||
input logic clk,
|
||||
input logic Start,
|
||||
input logic Stall, // *** multiple pipe stages
|
||||
|
|
@ -39,7 +43,7 @@ module srt #(parameter Nf=52) (
|
|||
// later add exponents, signs, special cases
|
||||
input logic XSign, YSign,
|
||||
input logic [`NE-1:0] XExp, YExp,
|
||||
input logic [Nf-1:0] SrcXFrac, SrcYFrac,
|
||||
input logic [`NF-1:0] SrcXFrac, SrcYFrac,
|
||||
input logic [`XLEN-1:0] SrcA, SrcB,
|
||||
input logic [1:0] Fmt, // Floats: 00 = 16 bit, 01 = 32 bit, 10 = 64 bit, 11 = 128 bit
|
||||
input logic W64, // 32-bit ints on XLEN=64
|
||||
|
|
@ -47,7 +51,7 @@ module srt #(parameter Nf=52) (
|
|||
input logic Int, // Choose integer inputs
|
||||
input logic Sqrt, // perform square root, not divide
|
||||
output logic rsign,
|
||||
output logic [Nf-1:0] Quot, Rem, QuotOTFC, // *** later handle integers
|
||||
output logic [`DIVLEN-1:0] Quot, Rem, QuotOTFC, // *** later handle integers
|
||||
output logic [`NE-1:0] rExp,
|
||||
output logic [3:0] Flags
|
||||
);
|
||||
|
|
@ -55,38 +59,40 @@ module srt #(parameter Nf=52) (
|
|||
logic qp, qz, qm; // quotient is +1, 0, or -1
|
||||
logic [`NE-1:0] calcExp;
|
||||
logic calcSign;
|
||||
logic [Nf-1:0] X, Dpreproc;
|
||||
logic [Nf+3:0] WS, WSA, WSN, WC, WCA, WCN, D, Db, Dsel;
|
||||
logic [Nf+2:0] rp, rm;
|
||||
logic [`DIVLEN-1:0] X, Dpreproc;
|
||||
logic [`DIVLEN+3:0] WS, WSA, WSN, WC, WCA, WCN, D, Db, Dsel;
|
||||
logic [`DIVLEN+2:0] rp, rm;
|
||||
logic [$clog2(`XLEN+1)-1:0] intExp;
|
||||
logic intSign;
|
||||
|
||||
srtpreproc #(Nf) preproc(SrcA, SrcB, SrcXFrac, SrcYFrac, Fmt, W64, Signed, Int, Sqrt, X, Dpreproc);
|
||||
srtpreproc preproc(SrcA, SrcB, SrcXFrac, SrcYFrac, Fmt, W64, Signed, Int, Sqrt, X, Dpreproc, intExp, intSign);
|
||||
|
||||
// Top Muxes and Registers
|
||||
// When start is asserted, the inputs are loaded into the divider.
|
||||
// Otherwise, the divisor is retained and the partial remainder
|
||||
// is fed back for the next iteration.
|
||||
mux2 #(Nf+4) wsmux({WSA[54:0], 1'b0}, {4'b0001, X}, Start, WSN);
|
||||
flop #(Nf+4) wsflop(clk, WSN, WS);
|
||||
mux2 #(Nf+4) wcmux({WCA[54:0], 1'b0}, 56'b0, Start, WCN);
|
||||
flop #(Nf+4) wcflop(clk, WCN, WC);
|
||||
flopen #(Nf+4) dflop(clk, Start, {4'b0001, Dpreproc}, D);
|
||||
mux2 #(`DIVLEN+4) wsmux({WSA[`DIVLEN+2:0], 1'b0}, {4'b0001, X}, Start, WSN);
|
||||
flop #(`DIVLEN+4) wsflop(clk, WSN, WS);
|
||||
mux2 #(`DIVLEN+4) wcmux({WCA[`DIVLEN+2:0], 1'b0}, {(`DIVLEN+4){1'b0}}, Start, WCN);
|
||||
flop #(`DIVLEN+4) wcflop(clk, WCN, WC);
|
||||
flopen #(`DIVLEN+4) dflop(clk, Start, {4'b0001, Dpreproc}, D);
|
||||
|
||||
// Quotient Selection logic
|
||||
// Given partial remainder, select quotient of +1, 0, or -1 (qp, qz, pm)
|
||||
// Accumulate quotient digits in a shift register
|
||||
qsel #(Nf) qsel(WS[55:52], WC[55:52], qp, qz, qm);
|
||||
qacc #(Nf+3) qacc(clk, Start, qp, qz, qm, rp, rm);
|
||||
qsel2 qsel2(WS[`DIVLEN+3:`DIVLEN], WC[`DIVLEN+3:`DIVLEN], qp, qz, qm);
|
||||
// Accumulate quotient digits in a shift register (now done in OTFC)
|
||||
qacc #(`DIVLEN+3) qacc(clk, Start, qp, qz, qm, rp, rm);
|
||||
flopen #(`NE) expflop(clk, Start, calcExp, rExp);
|
||||
flopen #(1) signflop(clk, Start, calcSign, rsign);
|
||||
|
||||
// Divisor Selection logic
|
||||
inv dinv(D, Db);
|
||||
mux3onehot divisorsel(Db, 56'b0, D, qp, qz, qm, Dsel);
|
||||
mux3onehot #(`DIVLEN) divisorsel(Db, {(`DIVLEN+4){1'b0}}, D, qp, qz, qm, Dsel);
|
||||
|
||||
// Partial Product Generation
|
||||
csa csa(WS, WC, Dsel, qp, WSA, WCA);
|
||||
csa #(`DIVLEN+4) csa(WS, WC, Dsel, qp, WSA, WCA);
|
||||
|
||||
otfc2 otfc2(clk, Start, qp, qz, qm, QuotOTFC);
|
||||
otfc2 #(`DIVLEN) otfc2(clk, Start, qp, qz, qm, QuotOTFC);
|
||||
|
||||
expcalc expcalc(.XExp, .YExp, .calcExp);
|
||||
|
||||
|
|
@ -95,70 +101,60 @@ module srt #(parameter Nf=52) (
|
|||
srtpostproc postproc(rp, rm, Quot);
|
||||
endmodule
|
||||
|
||||
module srtpostproc #(parameter N=52) (
|
||||
input [N+2:0] rp, rm,
|
||||
output [N-1:0] Quot
|
||||
);
|
||||
////////////////
|
||||
// Submodules //
|
||||
////////////////
|
||||
|
||||
//assign Quot = rp - rm;
|
||||
finaladd finaladd(rp, rm, Quot);
|
||||
endmodule
|
||||
|
||||
module srtpreproc #(parameter Nf=52) (
|
||||
///////////////////
|
||||
// Preprocessing //
|
||||
///////////////////
|
||||
module srtpreproc (
|
||||
input logic [`XLEN-1:0] SrcA, SrcB,
|
||||
input logic [Nf-1:0] SrcXFrac, SrcYFrac,
|
||||
input logic [`NF-1:0] SrcXFrac, SrcYFrac,
|
||||
input logic [1:0] Fmt, // Floats: 00 = 16 bit, 01 = 32 bit, 10 = 64 bit, 11 = 128 bit
|
||||
input logic W64, // 32-bit ints on XLEN=64
|
||||
input logic Signed, // Interpret integers as signed 2's complement
|
||||
input logic Int, // Choose integer inputss
|
||||
input logic Int, // Choose integer inputs
|
||||
input logic Sqrt, // perform square root, not divide
|
||||
output logic [Nf-1:0] X, D
|
||||
output logic [`DIVLEN-1:0] X, D,
|
||||
output logic [$clog2(`XLEN+1)-1:0] intExp, // Quotient integer exponent
|
||||
output logic intSign // Quotient integer sign
|
||||
);
|
||||
|
||||
// Initial: just pass X and Y through for simple fp division
|
||||
assign X = SrcXFrac;
|
||||
assign D = SrcYFrac;
|
||||
logic [$clog2(`XLEN+1)-1:0] zeroCntA, zeroCntB;
|
||||
logic [`XLEN-1:0] PosA, PosB;
|
||||
logic [`DIVLEN-1:0] ExtraA, ExtraB, PreprocA, PreprocB, PreprocX, PreprocY;
|
||||
|
||||
assign PosA = (Signed & SrcA[`XLEN - 1]) ? -SrcA : SrcA;
|
||||
assign PosB = (Signed & SrcB[`XLEN - 1]) ? -SrcB : SrcB;
|
||||
|
||||
lzc #(`XLEN) lzcA (PosA, zeroCntA);
|
||||
lzc #(`XLEN) lzcB (PosB, zeroCntB);
|
||||
|
||||
assign ExtraA = {1'b0, PosA, {`EXTRAINTBITS{1'b0}}};
|
||||
assign ExtraB = {1'b0, PosB, {`EXTRAINTBITS{1'b0}}};
|
||||
|
||||
assign PreprocA = ExtraA << zeroCntA;
|
||||
assign PreprocB = ExtraB << (zeroCntB + 1);
|
||||
assign PreprocX = {SrcXFrac, {`EXTRAFRACBITS{1'b0}}};
|
||||
assign PreprocY = {SrcYFrac, {`EXTRAFRACBITS{1'b0}}};
|
||||
|
||||
|
||||
assign X = Int ? PreprocA : PreprocX;
|
||||
assign D = Int ? PreprocB : PreprocY;
|
||||
assign intExp = zeroCntB - zeroCntA + 1;
|
||||
assign intSign = Signed & (SrcA[`XLEN - 1] ^ SrcB[`XLEN - 1]);
|
||||
endmodule
|
||||
|
||||
/*
|
||||
|
||||
//////////
|
||||
// mux2 //
|
||||
//////////
|
||||
module mux2(input logic [55:0] in0, in1,
|
||||
input logic sel,
|
||||
output logic [55:0] out);
|
||||
|
||||
assign #1 out = sel ? in1 : in0;
|
||||
endmodule
|
||||
|
||||
//////////
|
||||
// flop //
|
||||
//////////
|
||||
module flop(clk, in, out);
|
||||
input clk;
|
||||
input [55:0] in;
|
||||
output [55:0] out;
|
||||
|
||||
logic [55:0] state;
|
||||
|
||||
always @(posedge clk)
|
||||
state <= #1 in;
|
||||
|
||||
assign #1 out = state;
|
||||
endmodule
|
||||
|
||||
*/
|
||||
|
||||
//////////
|
||||
// qsel //
|
||||
//////////
|
||||
module qsel #(parameter Nf=52) ( // *** eventually just change to 4 bits
|
||||
input logic [Nf+3:Nf] ps, pc,
|
||||
/////////////////////////////////
|
||||
// Quotient Selection, Radix 2 //
|
||||
/////////////////////////////////
|
||||
module qsel2 ( // *** eventually just change to 4 bits
|
||||
input logic [`DIVLEN+3:`DIVLEN] ps, pc,
|
||||
output logic qp, qz, qm
|
||||
);
|
||||
|
||||
logic [Nf+3:Nf] p, g;
|
||||
logic [`DIVLEN+3:`DIVLEN] p, g;
|
||||
logic magnitude, sign, cout;
|
||||
|
||||
// The quotient selection logic is presented for simplicity, not
|
||||
|
|
@ -169,9 +165,9 @@ module qsel #(parameter Nf=52) ( // *** eventually just change to 4 bits
|
|||
assign p = ps ^ pc;
|
||||
assign g = ps & pc;
|
||||
|
||||
assign #1 magnitude = ~(&p[54:52]);
|
||||
assign #1 cout = g[54] | (p[54] & (g[53] | p[53] & g[52]));
|
||||
assign #1 sign = p[55] ^ cout;
|
||||
assign #1 magnitude = ~(&p[`DIVLEN+2:`DIVLEN]);
|
||||
assign #1 cout = g[`DIVLEN+2] | (p[`DIVLEN+2] & (g[`DIVLEN+1] | p[`DIVLEN+1] & g[`DIVLEN]));
|
||||
assign #1 sign = p[`DIVLEN+3] ^ cout;
|
||||
/* assign #1 magnitude = ~((ps[54]^pc[54]) & (ps[53]^pc[53]) &
|
||||
(ps[52]^pc[52]));
|
||||
assign #1 sign = (ps[55]^pc[55])^
|
||||
|
|
@ -188,15 +184,16 @@ endmodule
|
|||
//////////
|
||||
// qacc //
|
||||
//////////
|
||||
module qacc #(parameter N=55) (
|
||||
// To be replaced by OTFC
|
||||
module qacc #(parameter N=68) (
|
||||
input logic clk,
|
||||
input logic req,
|
||||
input logic qp, qz, qm,
|
||||
output logic [N-1:0] rp, rm
|
||||
);
|
||||
|
||||
flopr #(N) rmreg(clk, req, {rm[53:0], qm}, rm);
|
||||
flopr #(N) rpreg(clk, req, {rp[53:0], qp}, rp);
|
||||
flopr #(N) rmreg(clk, req, {rm[N-2:0], qm}, rm);
|
||||
flopr #(N) rpreg(clk, req, {rp[N-2:0], qp}, rp);
|
||||
/* always @(posedge clk)
|
||||
begin
|
||||
if (req)
|
||||
|
|
@ -212,11 +209,10 @@ module qacc #(parameter N=55) (
|
|||
end */
|
||||
endmodule
|
||||
|
||||
//////////
|
||||
// otfc //
|
||||
//////////
|
||||
|
||||
module otfc2 #(parameter N=52) (
|
||||
///////////////////////////////////
|
||||
// On-The-Fly Converter, Radix 2 //
|
||||
///////////////////////////////////
|
||||
module otfc2 #(parameter N=65) (
|
||||
input logic clk,
|
||||
input logic Start,
|
||||
input logic qp, qz, qm,
|
||||
|
|
@ -255,16 +251,15 @@ module otfc2 #(parameter N=52) (
|
|||
QMNext = {QMR, 1'b0};
|
||||
end
|
||||
end
|
||||
assign r = Q[54] ? Q[53:2] : Q[52:1];
|
||||
assign r = Q[N+2] ? Q[N+1:2] : Q[N:1];
|
||||
|
||||
endmodule
|
||||
|
||||
/////////
|
||||
// inv //
|
||||
/////////
|
||||
|
||||
module inv(input logic [55:0] in,
|
||||
output logic [55:0] out);
|
||||
module inv(input logic [`DIVLEN+3:0] in,
|
||||
output logic [`DIVLEN+3:0] out);
|
||||
|
||||
assign #1 out = ~in;
|
||||
endmodule
|
||||
|
|
@ -272,14 +267,11 @@ endmodule
|
|||
//////////
|
||||
// mux3 //
|
||||
//////////
|
||||
module mux3onehot(in0, in1, in2, sel0, sel1, sel2, out);
|
||||
input [55:0] in0;
|
||||
input [55:0] in1;
|
||||
input [55:0] in2;
|
||||
input sel0;
|
||||
input sel1;
|
||||
input sel2;
|
||||
output [55:0] out;
|
||||
module mux3onehot #(parameter N=65) (
|
||||
input logic [N+3:0] in0, in1, in2,
|
||||
input logic sel0, sel1, sel2,
|
||||
output logic [N+3:0] out
|
||||
);
|
||||
|
||||
// lazy inspection of the selects
|
||||
// really we should make sure selects are mutually exclusive
|
||||
|
|
@ -290,7 +282,7 @@ endmodule
|
|||
/////////
|
||||
// csa //
|
||||
/////////
|
||||
module csa #(parameter N=56) (
|
||||
module csa #(parameter N=69) (
|
||||
input logic [N-1:0] in1, in2, in3,
|
||||
input logic cin,
|
||||
output logic [N-1:0] out1, out2
|
||||
|
|
@ -305,28 +297,26 @@ module csa #(parameter N=56) (
|
|||
// insert cin.
|
||||
|
||||
assign #1 out1 = in1 ^ in2 ^ in3;
|
||||
assign #1 out2 = {in1[54:0] & (in2[54:0] | in3[54:0]) |
|
||||
(in2[54:0] & in3[54:0]), cin};
|
||||
assign #1 out2 = {in1[N-2:0] & (in2[N-2:0] | in3[N-2:0]) |
|
||||
(in2[N-2:0] & in3[N-2:0]), cin};
|
||||
endmodule
|
||||
|
||||
|
||||
//////////////
|
||||
// expcalc //
|
||||
//////////////
|
||||
|
||||
module expcalc(
|
||||
input logic [`NE-1:0] XExp, YExp,
|
||||
output logic [`NE-1:0] calcExp
|
||||
);
|
||||
|
||||
assign calcExp = XExp - YExp + 11'b01111111111;
|
||||
assign calcExp = XExp - YExp + (`NE)'(`BIAS);
|
||||
|
||||
endmodule
|
||||
|
||||
//////////////
|
||||
// signcalc //
|
||||
//////////////
|
||||
|
||||
module signcalc(
|
||||
input logic XSign, YSign,
|
||||
output logic calcSign
|
||||
|
|
@ -336,15 +326,27 @@ module signcalc(
|
|||
|
||||
endmodule
|
||||
|
||||
////////////////////
|
||||
// Postprocessing //
|
||||
////////////////////
|
||||
module srtpostproc (
|
||||
input [`DIVLEN+2:0] rp, rm,
|
||||
output [`DIVLEN-1:0] Quot
|
||||
);
|
||||
|
||||
//assign Quot = rp - rm;
|
||||
finaladd #(`DIVLEN+3) finaladd(rp, rm, Quot);
|
||||
endmodule
|
||||
|
||||
//////////////
|
||||
// finaladd //
|
||||
//////////////
|
||||
module finaladd(
|
||||
input logic [54:0] rp, rm,
|
||||
output logic [51:0] r
|
||||
module finaladd #(parameter N=68) (
|
||||
input logic [N-1:0] rp, rm,
|
||||
output logic [N-4:0] r
|
||||
);
|
||||
|
||||
logic [54:0] diff;
|
||||
logic [N-1:0] diff;
|
||||
|
||||
// this magic block performs the final addition for you
|
||||
// to convert the positive and negative quotient digits
|
||||
|
|
@ -359,6 +361,6 @@ module finaladd(
|
|||
// The checker ignores such an error.
|
||||
|
||||
assign #1 diff = rp - rm;
|
||||
assign #1 r = diff[54] ? diff[53:2] : diff[52:1];
|
||||
assign #1 r = diff[N-1] ? diff[N-2:2] : diff[N-3:1];
|
||||
endmodule
|
||||
|
||||
|
|
|
|||
|
|
@ -1,3 +1,5 @@
|
|||
`define DIVLEN 65
|
||||
|
||||
/////////////
|
||||
// counter //
|
||||
/////////////
|
||||
|
|
@ -37,15 +39,16 @@ endmodule
|
|||
// testbench //
|
||||
//////////
|
||||
module testbench;
|
||||
logic clk;
|
||||
logic req;
|
||||
logic done;
|
||||
logic [63:0] a, b;
|
||||
logic [51:0] afrac, bfrac;
|
||||
logic [10:0] aExp, bExp;
|
||||
logic asign, bsign;
|
||||
logic [51:0] r, rOTFC;
|
||||
logic [54:0] rp, rm; // positive quotient digits
|
||||
logic clk;
|
||||
logic req;
|
||||
logic done;
|
||||
logic [63:0] a, b;
|
||||
logic [51:0] afrac, bfrac;
|
||||
logic [10:0] aExp, bExp;
|
||||
logic asign, bsign;
|
||||
logic [51:0] r, rOTFC;
|
||||
logic [`DIVLEN-1:0] Quot, QuotOTFC;
|
||||
logic [54:0] rp, rm; // positive quotient digits
|
||||
|
||||
// Test parameters
|
||||
parameter MEM_SIZE = 40000;
|
||||
|
|
@ -65,14 +68,14 @@ module testbench;
|
|||
integer testnum, errors;
|
||||
|
||||
// Divider
|
||||
srt #(52) srt(.clk, .Start(req),
|
||||
srt srt(.clk, .Start(req),
|
||||
.Stall(1'b0), .Flush(1'b0),
|
||||
.XExp(aExp), .YExp(bExp), .rExp,
|
||||
.XSign(asign), .YSign(bsign), .rsign,
|
||||
.SrcXFrac(afrac), .SrcYFrac(bfrac),
|
||||
.SrcA('0), .SrcB('0), .Fmt(2'b00),
|
||||
.W64(1'b0), .Signed(1'b0), .Int(1'b0), .Sqrt(1'b0),
|
||||
.Quot(r), .QuotOTFC(rOTFC), .Rem(), .Flags());
|
||||
.Quot, .QuotOTFC, .Rem(), .Flags());
|
||||
|
||||
// Counter
|
||||
counter counter(clk, req, done);
|
||||
|
|
@ -98,6 +101,8 @@ module testbench;
|
|||
b = Vec[`memb];
|
||||
{bsign, bExp, bfrac} = b;
|
||||
nextr = Vec[`memr];
|
||||
r = Quot[`DIVLEN:`DIVLEN - 52];
|
||||
rOTFC = QuotOTFC[`DIVLEN:`DIVLEN - 52];
|
||||
req <= #5 1;
|
||||
end
|
||||
|
||||
|
|
|
|||
|
|
@ -261,26 +261,26 @@ module testbenchfp;
|
|||
Fmt = {Fmt, 2'b11};
|
||||
end
|
||||
end
|
||||
if (TEST === "div" | TEST === "all") begin // if division is being tested
|
||||
// add the divide tests/op-ctrls/unit/fmt
|
||||
Tests = {Tests, f128div};
|
||||
OpCtrl = {OpCtrl, `DIV_OPCTRL};
|
||||
WriteInt = {WriteInt, 1'b0};
|
||||
for(int i = 0; i<5; i++) begin
|
||||
Unit = {Unit, `DIVUNIT};
|
||||
Fmt = {Fmt, 2'b11};
|
||||
end
|
||||
end
|
||||
if (TEST === "sqrt" | TEST === "all") begin // if square-root is being tested
|
||||
// add the square-root tests/op-ctrls/unit/fmt
|
||||
Tests = {Tests, f128sqrt};
|
||||
OpCtrl = {OpCtrl, `SQRT_OPCTRL};
|
||||
WriteInt = {WriteInt, 1'b0};
|
||||
for(int i = 0; i<5; i++) begin
|
||||
Unit = {Unit, `DIVUNIT};
|
||||
Fmt = {Fmt, 2'b11};
|
||||
end
|
||||
end
|
||||
// if (TEST === "div" | TEST === "all") begin // if division is being tested
|
||||
// // add the divide tests/op-ctrls/unit/fmt
|
||||
// Tests = {Tests, f128div};
|
||||
// OpCtrl = {OpCtrl, `DIV_OPCTRL};
|
||||
// WriteInt = {WriteInt, 1'b0};
|
||||
// for(int i = 0; i<5; i++) begin
|
||||
// Unit = {Unit, `DIVUNIT};
|
||||
// Fmt = {Fmt, 2'b11};
|
||||
// end
|
||||
// end
|
||||
// if (TEST === "sqrt" | TEST === "all") begin // if square-root is being tested
|
||||
// // add the square-root tests/op-ctrls/unit/fmt
|
||||
// Tests = {Tests, f128sqrt};
|
||||
// OpCtrl = {OpCtrl, `SQRT_OPCTRL};
|
||||
// WriteInt = {WriteInt, 1'b0};
|
||||
// for(int i = 0; i<5; i++) begin
|
||||
// Unit = {Unit, `DIVUNIT};
|
||||
// Fmt = {Fmt, 2'b11};
|
||||
// end
|
||||
// end
|
||||
if (TEST === "fma" | TEST === "all") begin // if fused-mutliply-add is being tested
|
||||
// add each rounding mode to it's own list of tests
|
||||
// - fma tests are very long, so run all rounding modes in parallel
|
||||
|
|
@ -390,26 +390,26 @@ module testbenchfp;
|
|||
Fmt = {Fmt, 2'b01};
|
||||
end
|
||||
end
|
||||
if (TEST === "div" | TEST === "all") begin // if division is being tested
|
||||
// add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
Tests = {Tests, f64div};
|
||||
OpCtrl = {OpCtrl, `DIV_OPCTRL};
|
||||
WriteInt = {WriteInt, 1'b0};
|
||||
for(int i = 0; i<5; i++) begin
|
||||
Unit = {Unit, `DIVUNIT};
|
||||
Fmt = {Fmt, 2'b01};
|
||||
end
|
||||
end
|
||||
if (TEST === "sqrt" | TEST === "all") begin // if square-root is being tessted
|
||||
// add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
Tests = {Tests, f64sqrt};
|
||||
OpCtrl = {OpCtrl, `SQRT_OPCTRL};
|
||||
WriteInt = {WriteInt, 1'b0};
|
||||
for(int i = 0; i<5; i++) begin
|
||||
Unit = {Unit, `DIVUNIT};
|
||||
Fmt = {Fmt, 2'b01};
|
||||
end
|
||||
end
|
||||
// if (TEST === "div" | TEST === "all") begin // if division is being tested
|
||||
// // add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
// Tests = {Tests, f64div};
|
||||
// OpCtrl = {OpCtrl, `DIV_OPCTRL};
|
||||
// WriteInt = {WriteInt, 1'b0};
|
||||
// for(int i = 0; i<5; i++) begin
|
||||
// Unit = {Unit, `DIVUNIT};
|
||||
// Fmt = {Fmt, 2'b01};
|
||||
// end
|
||||
// end
|
||||
// if (TEST === "sqrt" | TEST === "all") begin // if square-root is being tessted
|
||||
// // add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
// Tests = {Tests, f64sqrt};
|
||||
// OpCtrl = {OpCtrl, `SQRT_OPCTRL};
|
||||
// WriteInt = {WriteInt, 1'b0};
|
||||
// for(int i = 0; i<5; i++) begin
|
||||
// Unit = {Unit, `DIVUNIT};
|
||||
// Fmt = {Fmt, 2'b01};
|
||||
// end
|
||||
// end
|
||||
if (TEST === "fma" | TEST === "all") begin // if the fused multiply add is being tested
|
||||
// add each rounding mode to it's own list of tests
|
||||
// - fma tests are very long, so run all rounding modes in parallel
|
||||
|
|
@ -502,26 +502,26 @@ module testbenchfp;
|
|||
Fmt = {Fmt, 2'b00};
|
||||
end
|
||||
end
|
||||
if (TEST === "div" | TEST === "all") begin // if division is being tested
|
||||
// add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
Tests = {Tests, f32div};
|
||||
OpCtrl = {OpCtrl, `DIV_OPCTRL};
|
||||
WriteInt = {WriteInt, 1'b0};
|
||||
for(int i = 0; i<5; i++) begin
|
||||
Unit = {Unit, `DIVUNIT};
|
||||
Fmt = {Fmt, 2'b00};
|
||||
end
|
||||
end
|
||||
if (TEST === "sqrt" | TEST === "all") begin // if sqrt is being tested
|
||||
// add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
Tests = {Tests, f32sqrt};
|
||||
OpCtrl = {OpCtrl, `SQRT_OPCTRL};
|
||||
WriteInt = {WriteInt, 1'b0};
|
||||
for(int i = 0; i<5; i++) begin
|
||||
Unit = {Unit, `DIVUNIT};
|
||||
Fmt = {Fmt, 2'b00};
|
||||
end
|
||||
end
|
||||
// if (TEST === "div" | TEST === "all") begin // if division is being tested
|
||||
// // add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
// Tests = {Tests, f32div};
|
||||
// OpCtrl = {OpCtrl, `DIV_OPCTRL};
|
||||
// WriteInt = {WriteInt, 1'b0};
|
||||
// for(int i = 0; i<5; i++) begin
|
||||
// Unit = {Unit, `DIVUNIT};
|
||||
// Fmt = {Fmt, 2'b00};
|
||||
// end
|
||||
// end
|
||||
// if (TEST === "sqrt" | TEST === "all") begin // if sqrt is being tested
|
||||
// // add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
// Tests = {Tests, f32sqrt};
|
||||
// OpCtrl = {OpCtrl, `SQRT_OPCTRL};
|
||||
// WriteInt = {WriteInt, 1'b0};
|
||||
// for(int i = 0; i<5; i++) begin
|
||||
// Unit = {Unit, `DIVUNIT};
|
||||
// Fmt = {Fmt, 2'b00};
|
||||
// end
|
||||
// end
|
||||
if (TEST === "fma" | TEST === "all") begin // if fma is being tested
|
||||
// add each rounding mode to it's own list of tests
|
||||
// - fma tests are very long, so run all rounding modes in parallel
|
||||
|
|
@ -596,26 +596,26 @@ module testbenchfp;
|
|||
Fmt = {Fmt, 2'b10};
|
||||
end
|
||||
end
|
||||
if (TEST === "div" | TEST === "all") begin // if division is being tested
|
||||
// add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
Tests = {Tests, f16div};
|
||||
OpCtrl = {OpCtrl, `DIV_OPCTRL};
|
||||
WriteInt = {WriteInt, 1'b0};
|
||||
for(int i = 0; i<5; i++) begin
|
||||
Unit = {Unit, `DIVUNIT};
|
||||
Fmt = {Fmt, 2'b10};
|
||||
end
|
||||
end
|
||||
if (TEST === "sqrt" | TEST === "all") begin // if sqrt is being tested
|
||||
// add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
Tests = {Tests, f16sqrt};
|
||||
OpCtrl = {OpCtrl, `SQRT_OPCTRL};
|
||||
WriteInt = {WriteInt, 1'b0};
|
||||
for(int i = 0; i<5; i++) begin
|
||||
Unit = {Unit, `DIVUNIT};
|
||||
Fmt = {Fmt, 2'b10};
|
||||
end
|
||||
end
|
||||
// if (TEST === "div" | TEST === "all") begin // if division is being tested
|
||||
// // add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
// Tests = {Tests, f16div};
|
||||
// OpCtrl = {OpCtrl, `DIV_OPCTRL};
|
||||
// WriteInt = {WriteInt, 1'b0};
|
||||
// for(int i = 0; i<5; i++) begin
|
||||
// Unit = {Unit, `DIVUNIT};
|
||||
// Fmt = {Fmt, 2'b10};
|
||||
// end
|
||||
// end
|
||||
// if (TEST === "sqrt" | TEST === "all") begin // if sqrt is being tested
|
||||
// // add the correct tests/op-ctrls/unit/fmt to their lists
|
||||
// Tests = {Tests, f16sqrt};
|
||||
// OpCtrl = {OpCtrl, `SQRT_OPCTRL};
|
||||
// WriteInt = {WriteInt, 1'b0};
|
||||
// for(int i = 0; i<5; i++) begin
|
||||
// Unit = {Unit, `DIVUNIT};
|
||||
// Fmt = {Fmt, 2'b10};
|
||||
// end
|
||||
// end
|
||||
if (TEST === "fma" | TEST === "all") begin // if fma is being tested
|
||||
// add each rounding mode to it's own list of tests
|
||||
// - fma tests are very long, so run all rounding modes in parallel
|
||||
|
|
@ -673,10 +673,10 @@ module testbenchfp;
|
|||
// - 1 for the larger precision
|
||||
// - 0 for the smaller precision
|
||||
always_comb begin
|
||||
if(`FMTBITS == 2) ModFmt = FmtVal;
|
||||
else ModFmt = FmtVal === `FMT;
|
||||
if(`FMTBITS == 2) FmaModFmt = FmaFmtVal;
|
||||
else FmaModFmt = FmaFmtVal === `FMT;
|
||||
if(`FMTBITS == 1) ModFmt = FmtVal == `FMT;
|
||||
else ModFmt = FmtVal;
|
||||
if(`FMTBITS == 1) FmaModFmt = FmaFmtVal == `FMT;
|
||||
else FmaModFmt = FmaFmtVal;
|
||||
end
|
||||
|
||||
// extract the inputs (X, Y, Z, SrcA) and the output (Ans, AnsFlg) from the current test vector
|
||||
|
|
@ -1028,111 +1028,111 @@ end
|
|||
// - the sign of the NaN does not matter for the opperations being tested
|
||||
// - when 2 or more NaNs are inputed the NaN that is propigated doesn't matter
|
||||
case (FmaFmtVal)
|
||||
4'b11: FmaRneNaNGood =(((`IEEE754==0)&(FmaRneRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
4'b11: FmaRneNaNGood =(((`IEEE754==0)&FmaRneAnsNaN&(FmaRneRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRneAnsFlg[4]&(FmaRneRes[`Q_LEN-2:0] === {{`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRneXNaN&(FmaRneRes[`Q_LEN-2:0] === {FmaRneX[`Q_LEN-2:`Q_NF],1'b1,FmaRneX[`Q_NF-2:0]})) |
|
||||
(FmaRneYNaN&(FmaRneRes[`Q_LEN-2:0] === {FmaRneY[`Q_LEN-2:`Q_NF],1'b1,FmaRneY[`Q_NF-2:0]})) |
|
||||
(FmaRneZNaN&(FmaRneRes[`Q_LEN-2:0] === {FmaRneZ[`Q_LEN-2:`Q_NF],1'b1,FmaRneZ[`Q_NF-2:0]})));
|
||||
4'b01: FmaRneNaNGood =(((`IEEE754==0)&(FmaRneRes === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
4'b01: FmaRneNaNGood =(((`IEEE754==0)&FmaRneAnsNaN&(FmaRneRes[`D_LEN-1:0] === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRneAnsFlg[4]&(FmaRneRes[`D_LEN-2:0] === {{`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRneXNaN&(FmaRneRes[`D_LEN-2:0] === {FmaRneX[`D_LEN-2:`D_NF],1'b1,FmaRneX[`D_NF-2:0]})) |
|
||||
(FmaRneYNaN&(FmaRneRes[`D_LEN-2:0] === {FmaRneY[`D_LEN-2:`D_NF],1'b1,FmaRneY[`D_NF-2:0]})) |
|
||||
(FmaRneZNaN&(FmaRneRes[`D_LEN-2:0] === {FmaRneZ[`D_LEN-2:`D_NF],1'b1,FmaRneZ[`D_NF-2:0]})));
|
||||
4'b00: FmaRneNaNGood =(((`IEEE754==0)&(FmaRneRes === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
4'b00: FmaRneNaNGood =(((`IEEE754==0)&FmaRneAnsNaN&(FmaRneRes[`S_LEN-1:0] === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRneAnsFlg[4]&(FmaRneRes[`S_LEN-2:0] === {{`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRneXNaN&(FmaRneRes[`S_LEN-2:0] === {FmaRneX[`S_LEN-2:`S_NF],1'b1,FmaRneX[`S_NF-2:0]})) |
|
||||
(FmaRneYNaN&(FmaRneRes[`S_LEN-2:0] === {FmaRneY[`S_LEN-2:`S_NF],1'b1,FmaRneY[`S_NF-2:0]})) |
|
||||
(FmaRneZNaN&(FmaRneRes[`S_LEN-2:0] === {FmaRneZ[`S_LEN-2:`S_NF],1'b1,FmaRneZ[`S_NF-2:0]})));
|
||||
4'b10: FmaRneNaNGood =(((`IEEE754==0)&(FmaRneRes === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
4'b10: FmaRneNaNGood =(((`IEEE754==0)&FmaRneAnsNaN&(FmaRneRes[`H_LEN-1:0] === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRneAnsFlg[4]&(FmaRneRes[`H_LEN-2:0] === {{`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRneXNaN&(FmaRneRes[`H_LEN-2:0] === {FmaRneX[`H_LEN-2:`H_NF],1'b1,FmaRneX[`H_NF-2:0]})) |
|
||||
(FmaRneYNaN&(FmaRneRes[`H_LEN-2:0] === {FmaRneY[`H_LEN-2:`H_NF],1'b1,FmaRneY[`H_NF-2:0]})) |
|
||||
(FmaRneZNaN&(FmaRneRes[`H_LEN-2:0] === {FmaRneZ[`H_LEN-2:`H_NF],1'b1,FmaRneZ[`H_NF-2:0]})));
|
||||
endcase
|
||||
case (FmaFmtVal)
|
||||
4'b11: FmaRzNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
4'b11: FmaRzNaNGood = (((`IEEE754==0)&FmaRzAnsNaN&(FmaRzRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRzAnsFlg[4]&(FmaRzRes[`Q_LEN-2:0] === {{`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRzXNaN&(FmaRzRes[`Q_LEN-2:0] === {FmaRzX[`Q_LEN-2:`Q_NF],1'b1,FmaRzX[`Q_NF-2:0]})) |
|
||||
(FmaRzYNaN&(FmaRzRes[`Q_LEN-2:0] === {FmaRzY[`Q_LEN-2:`Q_NF],1'b1,FmaRzY[`Q_NF-2:0]})) |
|
||||
(FmaRzZNaN&(FmaRzRes[`Q_LEN-2:0] === {FmaRzZ[`Q_LEN-2:`Q_NF],1'b1,FmaRzZ[`Q_NF-2:0]})));
|
||||
4'b01: FmaRzNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
4'b01: FmaRzNaNGood = (((`IEEE754==0)&FmaRzAnsNaN&(FmaRzRes[`D_LEN-1:0] === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRzAnsFlg[4]&(FmaRzRes[`D_LEN-2:0] === {{`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRzXNaN&(FmaRzRes[`D_LEN-2:0] === {FmaRzX[`D_LEN-2:`D_NF],1'b1,FmaRzX[`D_NF-2:0]})) |
|
||||
(FmaRzYNaN&(FmaRzRes[`D_LEN-2:0] === {FmaRzY[`D_LEN-2:`D_NF],1'b1,FmaRzY[`D_NF-2:0]})) |
|
||||
(FmaRzZNaN&(FmaRzRes[`D_LEN-2:0] === {FmaRzZ[`D_LEN-2:`D_NF],1'b1,FmaRzZ[`D_NF-2:0]})));
|
||||
4'b00: FmaRzNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
4'b00: FmaRzNaNGood = (((`IEEE754==0)&FmaRzAnsNaN&(FmaRzRes[`S_LEN-1:0] === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRzAnsFlg[4]&(FmaRzRes[`S_LEN-2:0] === {{`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRzXNaN&(FmaRzRes[`S_LEN-2:0] === {FmaRzX[`S_LEN-2:`S_NF],1'b1,FmaRzX[`S_NF-2:0]})) |
|
||||
(FmaRzYNaN&(FmaRzRes[`S_LEN-2:0] === {FmaRzY[`S_LEN-2:`S_NF],1'b1,FmaRzY[`S_NF-2:0]})) |
|
||||
(FmaRzZNaN&(FmaRzRes[`S_LEN-2:0] === {FmaRzZ[`S_LEN-2:`S_NF],1'b1,FmaRzZ[`S_NF-2:0]})));
|
||||
4'b10: FmaRzNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
4'b10: FmaRzNaNGood = (((`IEEE754==0)&FmaRzAnsNaN&(FmaRzRes[`H_LEN-1:0] === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRzAnsFlg[4]&(FmaRzRes[`H_LEN-2:0] === {{`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRzXNaN&(FmaRzRes[`H_LEN-2:0] === {FmaRzX[`H_LEN-2:`H_NF],1'b1,FmaRzX[`H_NF-2:0]})) |
|
||||
(FmaRzYNaN&(FmaRzRes[`H_LEN-2:0] === {FmaRzY[`H_LEN-2:`H_NF],1'b1,FmaRzY[`H_NF-2:0]})) |
|
||||
(FmaRzZNaN&(FmaRzRes[`H_LEN-2:0] === {FmaRzZ[`H_LEN-2:`H_NF],1'b1,FmaRzZ[`H_NF-2:0]})));
|
||||
endcase
|
||||
case (FmaFmtVal)
|
||||
4'b11: FmaRuNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
4'b11: FmaRuNaNGood = (((`IEEE754==0)&FmaRuAnsNaN&(FmaRuRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRuAnsFlg[4]&(FmaRuRes[`Q_LEN-2:0] === {{`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRuXNaN&(FmaRuRes[`Q_LEN-2:0] === {FmaRuX[`Q_LEN-2:`Q_NF],1'b1,FmaRuX[`Q_NF-2:0]})) |
|
||||
(FmaRuYNaN&(FmaRuRes[`Q_LEN-2:0] === {FmaRuY[`Q_LEN-2:`Q_NF],1'b1,FmaRuY[`Q_NF-2:0]})) |
|
||||
(FmaRuZNaN&(FmaRuRes[`Q_LEN-2:0] === {FmaRuZ[`Q_LEN-2:`Q_NF],1'b1,FmaRuZ[`Q_NF-2:0]})));
|
||||
4'b01: FmaRuNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
4'b01: FmaRuNaNGood = (((`IEEE754==0)&FmaRuAnsNaN&(FmaRuRes[`D_LEN-1:0] === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRuAnsFlg[4]&(FmaRuRes[`D_LEN-2:0] === {{`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRuAnsFlg[4]&(FmaRuRes[`Q_LEN-2:0] === {{`D_NE+1{1'b1}}, {`D_NF{1'b0}}})) |
|
||||
(FmaRuXNaN&(FmaRuRes[`D_LEN-2:0] === {FmaRuX[`D_LEN-2:`D_NF],1'b1,FmaRuX[`D_NF-2:0]})) |
|
||||
(FmaRuYNaN&(FmaRuRes[`D_LEN-2:0] === {FmaRuY[`D_LEN-2:`D_NF],1'b1,FmaRuY[`D_NF-2:0]})) |
|
||||
(FmaRuZNaN&(FmaRuRes[`D_LEN-2:0] === {FmaRuZ[`D_LEN-2:`D_NF],1'b1,FmaRuZ[`D_NF-2:0]})));
|
||||
4'b00: FmaRuNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
4'b00: FmaRuNaNGood = (((`IEEE754==0)&FmaRuAnsNaN&(FmaRuRes[`S_LEN-1:0] === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRuAnsFlg[4]&(FmaRuRes[`S_LEN-2:0] === {{`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRuXNaN&(FmaRuRes[`S_LEN-2:0] === {FmaRuX[`S_LEN-2:`S_NF],1'b1,FmaRuX[`S_NF-2:0]})) |
|
||||
(FmaRuYNaN&(FmaRuRes[`S_LEN-2:0] === {FmaRuY[`S_LEN-2:`S_NF],1'b1,FmaRuY[`S_NF-2:0]})) |
|
||||
(FmaRuZNaN&(FmaRuRes[`S_LEN-2:0] === {FmaRuZ[`S_LEN-2:`S_NF],1'b1,FmaRuZ[`S_NF-2:0]})));
|
||||
4'b10: FmaRuNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
4'b10: FmaRuNaNGood = (((`IEEE754==0)&FmaRuAnsNaN&(FmaRuRes[`H_LEN-1:0] === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRuAnsFlg[4]&(FmaRuRes[`H_LEN-2:0] === {{`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRuXNaN&(FmaRuRes[`H_LEN-2:0] === {FmaRuX[`H_LEN-2:`H_NF],1'b1,FmaRuX[`H_NF-2:0]})) |
|
||||
(FmaRuYNaN&(FmaRuRes[`H_LEN-2:0] === {FmaRuY[`H_LEN-2:`H_NF],1'b1,FmaRuY[`H_NF-2:0]})) |
|
||||
(FmaRuZNaN&(FmaRuRes[`H_LEN-2:0] === {FmaRuZ[`H_LEN-2:`H_NF],1'b1,FmaRuZ[`H_NF-2:0]})));
|
||||
endcase
|
||||
case (FmaFmtVal)
|
||||
4'b11: FmaRdNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
4'b11: FmaRdNaNGood = (((`IEEE754==0)&FmaRdAnsNaN&(FmaRdRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRdAnsFlg[4]&(FmaRdRes[`Q_LEN-2:0] === {{`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRdXNaN&(FmaRdRes[`Q_LEN-2:0] === {FmaRdX[`Q_LEN-2:`Q_NF],1'b1,FmaRdX[`Q_NF-2:0]})) |
|
||||
(FmaRdYNaN&(FmaRdRes[`Q_LEN-2:0] === {FmaRdY[`Q_LEN-2:`Q_NF],1'b1,FmaRdY[`Q_NF-2:0]})) |
|
||||
(FmaRdZNaN&(FmaRdRes[`Q_LEN-2:0] === {FmaRdZ[`Q_LEN-2:`Q_NF],1'b1,FmaRdZ[`Q_NF-2:0]})));
|
||||
4'b01: FmaRdNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
4'b01: FmaRdNaNGood = (((`IEEE754==0)&FmaRdAnsNaN&(FmaRdRes[`D_LEN-1:0] === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRdAnsFlg[4]&(FmaRdRes[`D_LEN-2:0] === {{`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRdXNaN&(FmaRdRes[`D_LEN-2:0] === {FmaRdX[`D_LEN-2:`D_NF],1'b1,FmaRdX[`D_NF-2:0]})) |
|
||||
(FmaRdYNaN&(FmaRdRes[`D_LEN-2:0] === {FmaRdY[`D_LEN-2:`D_NF],1'b1,FmaRdY[`D_NF-2:0]})) |
|
||||
(FmaRdZNaN&(FmaRdRes[`D_LEN-2:0] === {FmaRdZ[`D_LEN-2:`D_NF],1'b1,FmaRdZ[`D_NF-2:0]})));
|
||||
4'b00: FmaRdNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
4'b00: FmaRdNaNGood = (((`IEEE754==0)&FmaRdAnsNaN&(FmaRdRes[`S_LEN-1:0] === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRdAnsFlg[4]&(FmaRdRes[`S_LEN-2:0] === {{`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRdXNaN&(FmaRdRes[`S_LEN-2:0] === {FmaRdX[`S_LEN-2:`S_NF],1'b1,FmaRdX[`S_NF-2:0]})) |
|
||||
(FmaRdYNaN&(FmaRdRes[`S_LEN-2:0] === {FmaRdY[`S_LEN-2:`S_NF],1'b1,FmaRdY[`S_NF-2:0]})) |
|
||||
(FmaRdZNaN&(FmaRdRes[`S_LEN-2:0] === {FmaRdZ[`S_LEN-2:`S_NF],1'b1,FmaRdZ[`S_NF-2:0]})));
|
||||
4'b10: FmaRdNaNGood = (((`IEEE754==0)&(FmaRneRes === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
4'b10: FmaRdNaNGood = (((`IEEE754==0)&FmaRdAnsNaN&(FmaRdRes[`H_LEN-1:0] === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRdAnsFlg[4]&(FmaRdRes[`H_LEN-2:0] === {{`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRdXNaN&(FmaRdRes[`H_LEN-2:0] === {FmaRdX[`H_LEN-2:`H_NF],1'b1,FmaRdX[`H_NF-2:0]})) |
|
||||
(FmaRdYNaN&(FmaRdRes[`H_LEN-2:0] === {FmaRdY[`H_LEN-2:`H_NF],1'b1,FmaRdY[`H_NF-2:0]})) |
|
||||
(FmaRdZNaN&(FmaRdRes[`H_LEN-2:0] === {FmaRdZ[`H_LEN-2:`H_NF],1'b1,FmaRdZ[`H_NF-2:0]})));
|
||||
endcase
|
||||
case (FmaFmtVal)
|
||||
4'b11: FmaRnmNaNGood =(((`IEEE754==0)&(FmaRneRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
4'b11: FmaRnmNaNGood =(((`IEEE754==0)&FmaRnmAnsNaN&(FmaRnmRes === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRnmAnsFlg[4]&(FmaRnmRes[`Q_LEN-2:0] === {{`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(FmaRnmXNaN&(FmaRnmRes[`Q_LEN-2:0] === {FmaRnmX[`Q_LEN-2:`Q_NF],1'b1,FmaRnmX[`Q_NF-2:0]})) |
|
||||
(FmaRnmYNaN&(FmaRnmRes[`Q_LEN-2:0] === {FmaRnmY[`Q_LEN-2:`Q_NF],1'b1,FmaRnmY[`Q_NF-2:0]})) |
|
||||
(FmaRnmZNaN&(FmaRnmRes[`Q_LEN-2:0] === {FmaRnmZ[`Q_LEN-2:`Q_NF],1'b1,FmaRnmZ[`Q_NF-2:0]})));
|
||||
4'b01: FmaRnmNaNGood =(((`IEEE754==0)&(FmaRneRes === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
4'b01: FmaRnmNaNGood =(((`IEEE754==0)&FmaRnmAnsNaN&(FmaRnmRes[`D_LEN-1:0] === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRnmAnsFlg[4]&(FmaRnmRes[`D_LEN-2:0] === {{`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(FmaRnmXNaN&(FmaRnmRes[`D_LEN-2:0] === {FmaRnmX[`D_LEN-2:`D_NF],1'b1,FmaRnmX[`D_NF-2:0]})) |
|
||||
(FmaRnmYNaN&(FmaRnmRes[`D_LEN-2:0] === {FmaRnmY[`D_LEN-2:`D_NF],1'b1,FmaRnmY[`D_NF-2:0]})) |
|
||||
(FmaRnmZNaN&(FmaRnmRes[`D_LEN-2:0] === {FmaRnmZ[`D_LEN-2:`D_NF],1'b1,FmaRnmZ[`D_NF-2:0]})));
|
||||
4'b00: FmaRnmNaNGood =(((`IEEE754==0)&(FmaRneRes === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
4'b00: FmaRnmNaNGood =(((`IEEE754==0)&FmaRnmAnsNaN&(FmaRnmRes[`S_LEN-1:0] === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRnmAnsFlg[4]&(FmaRnmRes[`S_LEN-2:0] === {{`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(FmaRnmXNaN&(FmaRnmRes[`S_LEN-2:0] === {FmaRnmX[`S_LEN-2:`S_NF],1'b1,FmaRnmX[`S_NF-2:0]})) |
|
||||
(FmaRnmYNaN&(FmaRnmRes[`S_LEN-2:0] === {FmaRnmY[`S_LEN-2:`S_NF],1'b1,FmaRnmY[`S_NF-2:0]})) |
|
||||
(FmaRnmZNaN&(FmaRnmRes[`S_LEN-2:0] === {FmaRnmZ[`S_LEN-2:`S_NF],1'b1,FmaRnmZ[`S_NF-2:0]})));
|
||||
4'b10: FmaRnmNaNGood =(((`IEEE754==0)&(FmaRneRes === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
4'b10: FmaRnmNaNGood =(((`IEEE754==0)&FmaRnmAnsNaN&(FmaRnmRes[`H_LEN-1:0] === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRnmAnsFlg[4]&(FmaRnmRes[`H_LEN-2:0] === {{`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(FmaRnmXNaN&(FmaRnmRes[`H_LEN-2:0] === {FmaRnmX[`H_LEN-2:`H_NF],1'b1,FmaRnmX[`H_NF-2:0]})) |
|
||||
(FmaRnmYNaN&(FmaRnmRes[`H_LEN-2:0] === {FmaRnmY[`H_LEN-2:`H_NF],1'b1,FmaRnmY[`H_NF-2:0]})) |
|
||||
|
|
@ -1140,22 +1140,22 @@ end
|
|||
endcase
|
||||
if (UnitVal !== `CVTFPUNIT & UnitVal !== `CVTINTUNIT)
|
||||
case (FmtVal)
|
||||
4'b11: NaNGood = (((`IEEE754==0)&(Res === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
4'b11: NaNGood = (((`IEEE754==0)&AnsNaN&(Res === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(AnsFlg[4]&(Res[`Q_LEN-2:0] === {{`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(XNaN&(Res[`Q_LEN-2:0] === {X[`Q_LEN-2:`Q_NF],1'b1,X[`Q_NF-2:0]})) |
|
||||
(YNaN&(Res[`Q_LEN-2:0] === {Y[`Q_LEN-2:`Q_NF],1'b1,Y[`Q_NF-2:0]})) |
|
||||
(ZNaN&(Res[`Q_LEN-2:0] === {Z[`Q_LEN-2:`Q_NF],1'b1,Z[`Q_NF-2:0]})));
|
||||
4'b01: NaNGood = (((`IEEE754==0)&(Res === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
4'b01: NaNGood = (((`IEEE754==0)&AnsNaN&(Res[`D_LEN-1:0] === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(AnsFlg[4]&(Res[`D_LEN-2:0] === {{`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(XNaN&(Res[`D_LEN-2:0] === {X[`D_LEN-2:`D_NF],1'b1,X[`D_NF-2:0]})) |
|
||||
(YNaN&(Res[`D_LEN-2:0] === {Y[`D_LEN-2:`D_NF],1'b1,Y[`D_NF-2:0]})) |
|
||||
(ZNaN&(Res[`D_LEN-2:0] === {Z[`D_LEN-2:`D_NF],1'b1,Z[`D_NF-2:0]})));
|
||||
4'b00: NaNGood = (((`IEEE754==0)&(Res === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
4'b00: NaNGood = (((`IEEE754==0)&AnsNaN&(Res[`S_LEN-1:0] === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(AnsFlg[4]&(Res[`S_LEN-2:0] === {{`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(XNaN&(Res[`S_LEN-2:0] === {X[`S_LEN-2:`S_NF],1'b1,X[`S_NF-2:0]})) |
|
||||
(YNaN&(Res[`S_LEN-2:0] === {Y[`S_LEN-2:`S_NF],1'b1,Y[`S_NF-2:0]})) |
|
||||
(ZNaN&(Res[`S_LEN-2:0] === {Z[`S_LEN-2:`S_NF],1'b1,Z[`S_NF-2:0]})));
|
||||
4'b10: NaNGood = (((`IEEE754==0)&(Res === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
4'b10: NaNGood = (((`IEEE754==0)&AnsNaN&(Res[`H_LEN-1:0] === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(AnsFlg[4]&(Res[`H_LEN-2:0] === {{`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(XNaN&(Res[`H_LEN-2:0] === {X[`H_LEN-2:`H_NF],1'b1,X[`H_NF-2:0]})) |
|
||||
(YNaN&(Res[`H_LEN-2:0] === {Y[`H_LEN-2:`H_NF],1'b1,Y[`H_NF-2:0]})) |
|
||||
|
|
@ -1163,22 +1163,22 @@ end
|
|||
endcase
|
||||
else if (UnitVal === `CVTFPUNIT) // if converting from floating point to floating point OpCtrl contains the final FP format
|
||||
case (OpCtrlVal[1:0])
|
||||
2'b11: NaNGood = (((`IEEE754==0)&(Res === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
2'b11: NaNGood = (((`IEEE754==0)&AnsNaN&(Res === {1'b0, {`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(AnsFlg[4]&(Res[`Q_LEN-2:0] === {{`Q_NE+1{1'b1}}, {`Q_NF-1{1'b0}}})) |
|
||||
(AnsNaN&(Res[`Q_LEN-2:0] === Ans[`Q_LEN-2:0])) |
|
||||
(XNaN&(Res[`Q_LEN-2:0] === {X[`Q_LEN-2:`Q_NF],1'b1,X[`Q_NF-2:0]})) |
|
||||
(YNaN&(Res[`Q_LEN-2:0] === {Y[`Q_LEN-2:`Q_NF],1'b1,Y[`Q_NF-2:0]})));
|
||||
2'b01: NaNGood = (((`IEEE754==0)&(Res === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
2'b01: NaNGood = (((`IEEE754==0)&AnsNaN&(Res[`D_LEN-1:0] === {1'b0, {`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(AnsFlg[4]&(Res[`D_LEN-2:0] === {{`D_NE+1{1'b1}}, {`D_NF-1{1'b0}}})) |
|
||||
(AnsNaN&(Res[`D_LEN-2:0] === Ans[`D_LEN-2:0])) |
|
||||
(XNaN&(Res[`D_LEN-2:0] === {X[`D_LEN-2:`D_NF],1'b1,X[`D_NF-2:0]})) |
|
||||
(YNaN&(Res[`D_LEN-2:0] === {Y[`D_LEN-2:`D_NF],1'b1,Y[`D_NF-2:0]})));
|
||||
2'b00: NaNGood = (((`IEEE754==0)&(Res === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
2'b00: NaNGood = (((`IEEE754==0)&AnsNaN&(Res[`S_LEN-1:0] === {1'b0, {`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(AnsFlg[4]&(Res[`S_LEN-2:0] === {{`S_NE+1{1'b1}}, {`S_NF-1{1'b0}}})) |
|
||||
(AnsNaN&(Res[`S_LEN-2:0] === Ans[`S_LEN-2:0])) |
|
||||
(XNaN&(Res[`S_LEN-2:0] === {X[`S_LEN-2:`S_NF],1'b1,X[`S_NF-2:0]})) |
|
||||
(YNaN&(Res[`S_LEN-2:0] === {Y[`S_LEN-2:`S_NF],1'b1,Y[`S_NF-2:0]})));
|
||||
2'b10: NaNGood = (((`IEEE754==0)&(Res === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
2'b10: NaNGood = (((`IEEE754==0)&AnsNaN&(Res[`H_LEN-1:0] === {1'b0, {`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(AnsFlg[4]&(Res[`H_LEN-2:0] === {{`H_NE+1{1'b1}}, {`H_NF-1{1'b0}}})) |
|
||||
(AnsNaN&(Res[`H_LEN-2:0] === Ans[`H_LEN-2:0])) |
|
||||
(XNaN&(Res[`H_LEN-2:0] === {X[`H_LEN-2:`H_NF],1'b1,X[`H_NF-2:0]})) |
|
||||
|
|
@ -1452,14 +1452,14 @@ module readvectors (
|
|||
case (Fmt)
|
||||
2'b11: begin // quad
|
||||
X = TestVector[8+3*(`Q_LEN)-1:8+2*(`Q_LEN)];
|
||||
if(OpCtrl === `MUL_OPCTRL) Y = TestVector[8+2*(`Q_LEN)-1:8+(`Q_LEN)]; else Y = {2'b0, {`Q_NE-1{1'b1}}, `Q_NF'h0};
|
||||
if(OpCtrl === `MUL_OPCTRL) Y = TestVector[8+2*(`Q_LEN)-1:8+(`Q_LEN)]; else Y = {2'b0, {`Q_NE-1{1'b1}}, (`Q_NF)'(0)};
|
||||
if(OpCtrl === `MUL_OPCTRL) Z = 0; else Z = TestVector[8+2*(`Q_LEN)-1:8+(`Q_LEN)];
|
||||
Ans = TestVector[8+(`Q_LEN-1):8];
|
||||
end
|
||||
2'b01: begin // double
|
||||
X = {{`FLEN-`D_LEN{1'b1}}, TestVector[8+3*(`D_LEN)-1:8+2*(`D_LEN)]};
|
||||
if(OpCtrl === `MUL_OPCTRL) Y = {{`FLEN-`D_LEN{1'b1}}, TestVector[8+2*(`D_LEN)-1:8+(`D_LEN)]};
|
||||
else Y = {{`FLEN-`D_LEN{1'b1}}, 2'b0, {`D_NE-1{1'b1}}, `D_NF'h0};
|
||||
else Y = {{`FLEN-`D_LEN{1'b1}}, 2'b0, {`D_NE-1{1'b1}}, (`D_NF)'(0)};
|
||||
if(OpCtrl === `MUL_OPCTRL) Z = {{`FLEN-`D_LEN{1'b1}}, {`D_LEN{1'b0}}};
|
||||
else Z = {{`FLEN-`D_LEN{1'b1}}, TestVector[8+2*(`D_LEN)-1:8+(`D_LEN)]};
|
||||
Ans = {{`FLEN-`D_LEN{1'b1}}, TestVector[8+(`D_LEN-1):8]};
|
||||
|
|
@ -1467,7 +1467,7 @@ module readvectors (
|
|||
2'b00: begin // single
|
||||
X = {{`FLEN-`S_LEN{1'b1}}, TestVector[8+3*(`S_LEN)-1:8+2*(`S_LEN)]};
|
||||
if(OpCtrl === `MUL_OPCTRL) Y = {{`FLEN-`S_LEN{1'b1}}, TestVector[8+2*(`S_LEN)-1:8+(`S_LEN)]};
|
||||
else Y = {{`FLEN-`S_LEN{1'b1}}, 2'b0, {`S_NE-1{1'b1}}, `S_NF'h0};
|
||||
else Y = {{`FLEN-`S_LEN{1'b1}}, 2'b0, {`S_NE-1{1'b1}}, (`S_NF)'(0)};
|
||||
if(OpCtrl === `MUL_OPCTRL) Z = {{`FLEN-`S_LEN{1'b1}}, {`S_LEN{1'b0}}};
|
||||
else Z = {{`FLEN-`S_LEN{1'b1}}, TestVector[8+2*(`S_LEN)-1:8+(`S_LEN)]};
|
||||
Ans = {{`FLEN-`S_LEN{1'b1}}, TestVector[8+(`S_LEN-1):8]};
|
||||
|
|
@ -1475,7 +1475,7 @@ module readvectors (
|
|||
2'b10: begin // half
|
||||
X = {{`FLEN-`H_LEN{1'b1}}, TestVector[8+3*(`H_LEN)-1:8+2*(`H_LEN)]};
|
||||
if(OpCtrl === `MUL_OPCTRL) Y = {{`FLEN-`H_LEN{1'b1}}, TestVector[8+2*(`H_LEN)-1:8+(`H_LEN)]};
|
||||
else Y = {{`FLEN-`H_LEN{1'b1}}, 2'b0, {`H_NE-1{1'b1}}, `H_NF'h0};
|
||||
else Y = {{`FLEN-`H_LEN{1'b1}}, 2'b0, {`H_NE-1{1'b1}}, (`H_NF)'(0)};
|
||||
if(OpCtrl === `MUL_OPCTRL) Z = {{`FLEN-`H_LEN{1'b1}}, {`H_LEN{1'b0}}};
|
||||
else Z = {{`FLEN-`H_LEN{1'b1}}, TestVector[8+2*(`H_LEN)-1:8+(`H_LEN)]};
|
||||
Ans = {{`FLEN-`H_LEN{1'b1}}, TestVector[8+(`H_LEN-1):8]};
|
||||
|
|
|
|||
|
|
@ -666,7 +666,7 @@ module testbench;
|
|||
// turn on waves
|
||||
if (AttemptedInstructionCount == INSTR_WAVEON) $stop;
|
||||
// end sim
|
||||
if ((AttemptedInstructionCount == INSTR_LIMIT) & (INSTR_LIMIT!=0)) $stop;
|
||||
if ((AttemptedInstructionCount == INSTR_LIMIT) & (INSTR_LIMIT!=0)) begin $stop; $stop; end
|
||||
fault = 0;
|
||||
if (`DEBUG_TRACE >= 1) begin
|
||||
`checkEQ("PCW",PCW,ExpectedPCW)
|
||||
|
|
|
|||
|
|
@ -127,7 +127,7 @@ logic [3:0] dummy;
|
|||
end
|
||||
|
||||
string signame, memfilename, pathname, objdumpfilename, adrstr, outputfile;
|
||||
integer outputFilePointer;
|
||||
integer outputFilePointer, ProgramLabelMap, ProgramAddrMap;
|
||||
|
||||
logic [31:0] GPIOPinsIn, GPIOPinsOut, GPIOPinsEn;
|
||||
logic UARTSin, UARTSout;
|
||||
|
|
@ -215,8 +215,31 @@ logic [3:0] dummy;
|
|||
// Termination condition (i.e. we finished running current test)
|
||||
if (DCacheFlushDone) begin
|
||||
// Gets the memory location of begin_signature
|
||||
testadr = (`RAM_BASE+tests[test+1].atohex())/(`XLEN/8);
|
||||
testadrNoBase = (tests[test+1].atohex())/(`XLEN/8);
|
||||
adrstr = "0";
|
||||
ProgramLabelMap = $fopen(ProgramLabelMapFile, "r");
|
||||
ProgramAddrMap = $fopen(ProgramAddrMapFile, "r");
|
||||
if (ProgramLabelMap & ProgramAddrMap) begin // check we found both files
|
||||
while (!$feof(ProgramLabelMap)) begin
|
||||
string addr, label;
|
||||
integer returncode;
|
||||
returncode = $fgets(label, ProgramLabelMap);
|
||||
returncode = $fgets(addr, ProgramAddrMap);
|
||||
if (label == "begin_signature\n") begin
|
||||
adrstr = addr[1:7];
|
||||
if (adrstr=="0000000") // if running on rv64 we get the address at a later
|
||||
adrstr = addr[9:15];
|
||||
if (DEBUG) $display("%s begin_signature adrstr: %s", TEST, adrstr);
|
||||
end
|
||||
end
|
||||
end
|
||||
if (adrstr == "0") begin
|
||||
$display("begin_signature addr not found in %s", ProgramLabelMapFile);
|
||||
end
|
||||
$fclose(ProgramLabelMap);
|
||||
$fclose(ProgramAddrMap);
|
||||
|
||||
testadr = (`RAM_BASE+adrstr.atohex())/(`XLEN/8);
|
||||
testadrNoBase = (adrstr.atohex())/(`XLEN/8);
|
||||
#600; // give time for instructions in pipeline to finish
|
||||
if (TEST == "embench") begin
|
||||
// Writes contents of begin_signature to .sim.output file
|
||||
|
|
@ -294,7 +317,8 @@ logic [3:0] dummy;
|
|||
end
|
||||
end
|
||||
// move onto the next test, check to see if we're done
|
||||
test = test + 2;
|
||||
// test = test + 2;
|
||||
test = test + 1;
|
||||
if (test == tests.size()) begin
|
||||
if (totalerrors == 0) $display("SUCCESS! All tests ran without failures.");
|
||||
else $display("FAIL: %d test programs had errors", totalerrors);
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load diff
Loading…
Add table
Add a link
Reference in a new issue