Atomic, section of Accelerate ML/DL on OpenCL/SYC and refine Makefile…

… and install.rst

README.md

@@ -10,34 +10,29 @@ If you would like to view an up to date version of this book in your
 browser without checking out and building the book, please visit: 
 http://jonathan2251.github.io/lbd/index.html
 
+Linux prerequisite:
+Add swapfile for Linux: http://jonathan2251.github.io/lbd/install.html#build-steps
+
 BUILD steps:
 
 $ pwd
-
 $ $HOME  // HOME directory
-
-$ mkdir llvm
-
+$ mkdir git
 $ cd git
-
 $ pwd
-
 $ $HOME/git
-
 $ git clone https://github.com/Jonathan2251/lbd
-
 $ cd lbd/lbdex/install_llvm
-
 $ bash build-llvm.sh
-
+...
 # Please remember to add ${LLVM_RELEASE_DIR}/bin to variable ${PATH} to your
 # environment for clang++, clang. Reference last line of 
 # lbd/lbdex/install_llvm/build-llvm.sh
 
 $ cd ..
-
 $ pwd
-
 $ $HOME/git/lbd/lbdex
-
 $ bash build-cpu0.sh
+
+CHECK step:
+$ bash check.sh

check.sh

@@ -0,0 +1,46 @@
+#!/usr/bin/env bash
+
+pushd lbdex/input
+echo "bash build-run_backend.sh cpu032I eb"
+bash build-run_backend.sh cpu032I eb
+if [ "$?" != "0" ]; then
+  echo "FAIL"
+  exit 1;
+fi
+pushd ../verilog
+make
+./cpu0Is
+popd
+
+echo "bash build-run_backend.sh cpu032I el"
+bash build-run_backend.sh cpu032I el
+if [ "$?" != "0" ]; then
+  echo "FAIL"
+  exit 1;
+fi
+pushd ../verilog
+./cpu0Is
+popd
+
+echo "bash build-run_backend.sh cpu032II eb"
+bash build-run_backend.sh cpu032II eb
+if [ "$?" != "0" ]; then
+  echo "FAIL"
+  exit 1;
+fi
+pushd ../verilog
+./cpu0IIs
+popd
+
+echo "bash build-run_backend.sh cpu032II el"
+bash build-run_backend.sh cpu032II el
+if [ "$?" != "0" ]; then
+  echo "FAIL"
+  exit 1;
+fi
+pushd ../verilog
+./cpu0IIs
+popd
+
+echo "PASS"
+popd

lbdex/build-cpu0.sh

@@ -1,7 +1,7 @@
 #!/usr/bin/env bash
 
 LLVM_DIR=~/llvm
-LLVM_RELEASE_DIR=${LLVM_DIR}/release
+LLVM_DEBUG_DIR=${LLVM_DIR}/debug
 LLVM_TEST_DIR=${LLVM_DIR}/test
 
 if ! test -d ${LLVM_TEST_DIR}; then

lbdex/build-llvm-test-suite.sh

@@ -6,7 +6,7 @@ LLVM_DIR=~/llvm
 # Cpu0 set the other installed DIRs here
 LLVM_SRC_DIR=${LLVM_DIR}/llvm-project
 LLVM_TEST_SUITE_DIR=${LLVM_SRC_DIR}/test-suite
-LLVM_RELEASE_DIR=${LLVM_DIR}/release
+LLVM_DEBUG_DIR=${LLVM_DIR}/debug
 
 # ref.
 # https://llvm.org/docs/TestSuiteGuide.html
@@ -17,11 +17,11 @@ LLVM_RELEASE_DIR=${LLVM_DIR}/release
 # sudo apt-get install tcl tk tcl-dev tk-dev
 # On macos,
 # brew install tcl-tk
-# ${LLVM_RELEASE_DIR}/build: build with clang and compiler-rt, -DLLVM_ENABLE_PROJECTS="clang;compiler-rt" --> ref. https://github.com/Jonathan2251/lbd/blob/master/lbdex/install_llvm/build-llvm.sh
+# ${LLVM_DEBUG_DIR}/build: build with clang and compiler-rt, -DLLVM_ENABLE_PROJECTS="clang;compiler-rt" --> ref. https://github.com/Jonathan2251/lbd/blob/master/lbdex/install_llvm/build-llvm.sh
 
 build() {
   cd test-suite-build
-  cmake -DCMAKE_C_COMPILER=${LLVM_RELEASE_DIR}/build/bin/clang -C../test-suite/cmake/caches/O3.cmake -DCMAKE_C_FLAGS=-fPIE -DCMAKE_CXX_FLAGS=-fPIE ../test-suite
+  cmake -DCMAKE_C_COMPILER=${LLVM_DEBUG_DIR}/build/bin/clang -C../test-suite/cmake/caches/O3.cmake -DCMAKE_C_FLAGS=-fPIE -DCMAKE_CXX_FLAGS=-fPIE ../test-suite
   make
 }
 
@@ -31,14 +31,14 @@ if ! test -d ${LLVM_TEST_SUITE_DIR}; then
   cd test-suite
   git checkout -b 12.x origin/release/12.x
   popd
-  pushd ${LLVM_RELEASE_DIR}
+  pushd ${LLVM_DEBUG_DIR}
   ln -s ../llvm-project/test-suite test-suite
 #  sed 's/^add_subdirectory(XRay)/#add_subdirectory(XRay)/g' test-suite/MicroBenchmarks/CMakeLists.txt
   rm -rf test-suite-build
   mkdir test-suite-build
   build;
 else
-  pushd ${LLVM_RELEASE_DIR}
+  pushd ${LLVM_DEBUG_DIR}
   build;
   echo "${LLVM_TEST_SUITE_DIR} has existed already"
   exit 1

lbdex/gen-ref-output.sh

@@ -4,7 +4,7 @@ OS=`uname -s`
 echo "OS =" ${OS}
 
 TOOLDIR=~/llvm/test/build/bin
-CLANG=~/llvm/release/build/bin/clang
+CLANG=~/llvm/debug/build/bin/clang
 
 rm -rf output
 mkdir output

lbdex/input/build-run_backend.sh

@@ -1,14 +1,18 @@
 #!/usr/bin/env bash
 
+# for example:
+# bash build-run_backend.sh cpu032I el
+# bash build-run_backend.sh cpu032II eb
+
 source functions.sh
 
 sh_name=build-run_backend.sh
-argNum=$#
-arg1=$1
-arg2=$2
+ARG_NUM=$#
+CPU=$1
+ENDIAN=$2
 
 DEFFLAGS=""
-if [ "$arg1" == cpu032II ] ; then
+if [ "$CPU" == cpu032II ] ; then
   DEFFLAGS=${DEFFLAGS}" -DCPU032II"
 fi
 echo ${DEFFLAGS}
@@ -19,23 +23,26 @@ prologue;
 # on verilog.
 $CLANG ${DEFFLAGS} -target mips-unknown-linux-gnu -c ch8_2_select_global_pic.cpp \
 -emit-llvm -o ch8_2_select_global_pic.bc
-${TOOLDIR}/llc -march=cpu0${endian} -mcpu=${CPU} -relocation-model=pic \
+${TOOLDIR}/llc -march=cpu0${ENDIAN} -mcpu=${CPU} -relocation-model=pic \
 -filetype=obj ch8_2_select_global_pic.bc -o ch8_2_select_global_pic.cpu0.o
 
 $CLANG ${DEFFLAGS} -target mips-unknown-linux-gnu -c ch_run_backend.cpp \
 -emit-llvm -o ch_run_backend.bc
-echo "${TOOLDIR}/llc -march=cpu0${endian} -mcpu=${CPU} -relocation-model=static \
+echo "${TOOLDIR}/llc -march=cpu0${ENDIAN} -mcpu=${CPU} -relocation-model=static \
 -filetype=obj -enable-cpu0-tail-calls ch_run_backend.bc -o ch_run_backend.cpu0.o"
-${TOOLDIR}/llc -march=cpu0${endian} -mcpu=${CPU} -relocation-model=static \
+${TOOLDIR}/llc -march=cpu0${ENDIAN} -mcpu=${CPU} -relocation-model=static \
 -filetype=obj -enable-cpu0-tail-calls ch_run_backend.bc -o ch_run_backend.cpu0.o
 # print must at the same line, otherwise it will spilt into 2 lines
 ${TOOLDIR}/llvm-objdump --section=.text -d ch_run_backend.cpu0.o | tail -n +8| awk \
 '{print "/* " $1 " */\t" $2 " " $3 " " $4 " " $5 "\t/* " $6"\t" $7" " $8" " $9" " $10 "\t*/"}' \
  > ../verilog/cpu0.hex
 
-if [ "$arg2" == le ] ; then
-  echo "1   /* 0: big endian, 1: little endian */" > ../verilog/cpu0.config
+ENDIAN=`${TOOLDIR}/llvm-readobj -h ch_run_backend.cpu0.o|grep "DataEncoding"|awk '{print $2}'`
+isLittleEndian;
+
+if [ $LE == "true" ] ; then
+  echo "1   /* 0: big ENDIAN, 1: little ENDIAN */" > ../verilog/cpu0.config
 else
-  echo "0   /* 0: big endian, 1: little endian */" > ../verilog/cpu0.config
+  echo "0   /* 0: big ENDIAN, 1: little ENDIAN */" > ../verilog/cpu0.config
 fi
 cat ../verilog/cpu0.config

lbdex/input/build-run_backend2.sh

@@ -3,9 +3,9 @@
 source functions.sh
 
 sh_name=build-run_backend.sh
-argNum=$#
-arg1=$1
-arg2=$2
+ARG_NUM=$#
+CPU=$1
+ENDIAN=$2
 
 DEFFLAGS=""
 if [ "$arg1" == cpu032II ] ; then
@@ -17,13 +17,16 @@ prologue;
 
 ${CLANG} ${DEFFLAGS} -c ch_run_backend2.cpp \
 -emit-llvm -o ch_run_backend2.bc
-${TOOLDIR}/llc -march=cpu0${endian} -mcpu=${CPU} -relocation-model=static \
+${TOOLDIR}/llc -march=cpu0${ENDIAN} -mcpu=${CPU} -relocation-model=static \
 -filetype=obj ch_run_backend2.bc -o ch_run_backend2.cpu0.o
 ${TOOLDIR}/llvm-objdump -d ch_run_backend2.cpu0.o | tail -n +8| awk \
 '{print "/* " $1 " */\t" $2 " " $3 " " $4 " " $5 "\t/* " $6"\t" $7" " $8" \
 " $9" " $10 "\t*/"}' > ../verilog/cpu0.hex
 
-if [ "$arg2" == le ] ; then
+ENDIAN=`${TOOLDIR}/llvm-readobj -h ch_run_backend2.cpu0.o|grep "DataEncoding"|awk '{print $2}'`
+isLittleEndian;
+
+if [ $LE == "true" ] ; then
   echo "1   /* 0: big endian, 1: little endian */" > ../verilog/cpu0.config
 else
   echo "0   /* 0: big endian, 1: little endian */" > ../verilog/cpu0.config

lbdex/input/ch8_2_phinode.cpp

@@ -1,5 +1,5 @@
-// ~/llvm/release/build/bin/clang -O3 -target mips-unknown-linux-gnu -c ch8_2_phinode.cpp -emit-llvm -o ch8_2_phinode.bc
-// ~/llvm/release/build/bin/clang -O0 -target mips-unknown-linux-gnu -c ch8_2_phinode.cpp -emit-llvm -o ch8_2_phinode.bc
+// ~/llvm/debug/build/bin/clang -O3 -target mips-unknown-linux-gnu -c ch8_2_phinode.cpp -emit-llvm -o ch8_2_phinode.bc
+// ~/llvm/debug/build/bin/clang -O0 -target mips-unknown-linux-gnu -c ch8_2_phinode.cpp -emit-llvm -o ch8_2_phinode.bc
 // ~/llvm/test/build/bin/llvm-dis ch8_2_phinode.bc -o -
 // ~/llvm/test/build/bin/llc -march=cpu0 -mcpu=cpu032I -relocation-model=pic -filetype=asm ch8_2_phinode.bc -o -
 

lbdex/input/ch9_caller_callee_save_registers.cpp

@@ -1,6 +1,6 @@
 // clang -O0 -target mips-unknown-linux-gnu -c ch9_caller_callee_save_registers.cpp -emit-llvm -o ch9_caller_callee_save_registers.bc
 // ~/llvm/test/build/bin/llc -O0 -march=cpu0 -relocation-model=static -filetype=asm ch9_caller_callee_save_registers.bc -o -
-// ~/llvm/release/build/bin/llc -O0 -march=mips -relocation-model=static -filetype=asm ch9_caller_callee_save_registers.bc -o -
+// ~/llvm/debug/build/bin/llc -O0 -march=mips -relocation-model=static -filetype=asm ch9_caller_callee_save_registers.bc -o -
 
 /// start
 extern int add1(int x);
@@ -12,4 +12,4 @@ int callee()
   result = result - t1;
 
   return result;
-}
+}

lbdex/input/functions.sh

@@ -1,13 +1,13 @@
 prologue() {
-  if [ $argNum == 0 ]; then
+  if [ $ARG_NUM == 0 ]; then
     echo "useage: bash $sh_name cpu_type endian"
     echo "  cpu_type: cpu032I or cpu032II"
-    echo "  endian: be (big endian, default) or le (little endian)"
+    echo "  endian: eb (big endian, default) or el (little endian)"
     echo "for example:"
     echo "  bash build-slinker.sh cpu032I be"
     exit 1;
   fi
-  if [ $arg1 != cpu032I ] && [ $arg1 != cpu032II ]; then
+  if [ $CPU != cpu032I ] && [ $CPU != cpu032II ]; then
     echo "1st argument is cpu032I or cpu032II"
     exit 1
   fi
@@ -18,38 +18,36 @@ prologue() {
   TOOLDIR=~/llvm/test/build/bin
   CLANG=~/llvm/test/build/bin/clang
 
-  CPU=$arg1
+  CPU=$CPU
   echo "CPU =" "${CPU}"
 
-  if [ "$arg2" != "" ] && [ $arg2 != le ] && [ $arg2 != be ]; then
-    echo "2nd argument is be (big endian, default) or le (little endian)"
+  if [ "$ENDIAN" != "" ] && [ $ENDIAN != el ] && [ $ENDIAN != eb ]; then
+    echo "2nd argument is eb (big endian, default) or el (little endian)"
     exit 1
   fi
-  if [ "$arg2" == "" ] || [ $arg2 == be ]; then
-    endian=
-  else
-    endian=el
+  if [ $ENDIAN == eb ]; then
+    ENDIAN=
   fi
-  echo "endian =" "${endian}"
+  echo "ENDIAN =" "${ENDIAN}"
 
   bash clean.sh
 }
 
 isLittleEndian() {
-  echo "endian = " "$endian"
-  if [ "$endian" == "LittleEndian" ] ; then
-    le="true"
-  elif [ "$endian" == "BigEndian" ] ; then
-    le="false"
+  echo "ENDIAN = " "$ENDIAN"
+  if [ "$ENDIAN" == "LittleEndian" ] ; then
+    LE="true"
+  elif [ "$ENDIAN" == "BigEndian" ] ; then
+    LE="false"
   else
-    echo "!endian unknown"
+    echo "!ENDIAN unknown"
     exit 1
   fi
 }
 
 elf2hex() {
-  ${TOOLDIR}/llvm-objdump -elf2hex -le=${le} a.out > ../verilog/cpu0.hex
-  if [ ${le} == "true" ] ; then
+  ${TOOLDIR}/llvm-objdump -elf2hex -le=$LE a.out > ../verilog/cpu0.hex
+  if [ $LE == "true" ] ; then
     echo "1   /* 0: big endian, 1: little endian */" > ../verilog/cpu0.config
   else
     echo "0   /* 0: big endian, 1: little endian */" > ../verilog/cpu0.config

lbdex/install_llvm/build-llvm.sh

@@ -1,13 +1,20 @@
 #!/usr/bin/env bash
 
+# If your memory is large enough, add swap size ubuntu as follows,
+# 1. search "add swap size ubuntu"
+# 2. https://bogdancornianu.com/change-swap-size-in-ubuntu/
+# 3. add to boot by change to /swapfile                                 swap            swap    default              0       0
+# https://linuxize.com/post/how-to-add-swap-space-on-ubuntu-18-04/
+# 4. reboot
+
 LLVM_DIR=~/llvm
-LLVM_RELEASE_DIR=${LLVM_DIR}/release
+LLVM_DEBUG_DIR=${LLVM_DIR}/debug
 LLVM_TEST_DIR=${LLVM_DIR}/test
 
 pushd ${LLVM_DIR}
 
 if ! test -d ${LLVM_DIR}/llvm-project; then
-  git clone git://github.com/llvm/llvm-project.git
+  git clone https://github.com/llvm/llvm-project.git
   cd llvm-project
   git checkout -b 12.x origin/release/12.x
   git checkout e8a397203c67adbeae04763ce25c6a5ae76af52c
@@ -17,22 +24,22 @@ else
 #  exit 1
 fi
 
-if ! test -d ${LLVM_RELEASE_DIR}; then
-  mkdir ${LLVM_RELEASE_DIR}
-  cp -rf llvm-project/clang ${LLVM_RELEASE_DIR}
-  cp -rf llvm-project/llvm ${LLVM_RELEASE_DIR}
+if ! test -d ${LLVM_DEBUG_DIR}; then
+  mkdir ${LLVM_DEBUG_DIR}
+  cp -rf llvm-project/clang ${LLVM_DEBUG_DIR}
+  cp -rf llvm-project/llvm ${LLVM_DEBUG_DIR}
 # build compiler-rt for llvm-test-suite
-  cp -rf llvm-project/compiler-rt ${LLVM_RELEASE_DIR}
-  mkdir ${LLVM_RELEASE_DIR}/build
-  pushd ${LLVM_RELEASE_DIR}/build
+  cp -rf llvm-project/compiler-rt ${LLVM_DEBUG_DIR}
+  mkdir ${LLVM_DEBUG_DIR}/build
+  pushd ${LLVM_DEBUG_DIR}/build
   OS=`uname -s`
   echo "OS =" ${OS}
-  cmake -DCMAKE_BUILD_TYPE=Debug -DLLVM_ENABLE_PROJECTS="clang;compiler-rt" \
+  cmake -DCMAKE_BUILD_TYPE=Debug -DLLVM_ENABLE_PROJECTS="clang" \
   -DLLVM_PARALLEL_COMPILE_JOBS=4 -DLLVM_PARALLEL_LINK_JOBS=1 -G "Ninja" ../llvm
   ninja
   popd
 fi
 
 popd
-echo "Please remember to add ${LLVM_RELEASE_DIR}/bin to variable \${PATH} to your \
+echo "Please remember to add ${LLVM_DEBUG_DIR}/build/bin to variable \${PATH} to your \
   environment for clang++, clang."

lbdex/opt/add.ll

@@ -1,4 +1,4 @@
-; ~/llvm/3.9.0/release/cmake_debug_build/bin/opt -O3 -S add.ll -o -
+; ~/llvm/debug/build/bin/opt -O3 -S add.ll -o -
 
 define i32 @add(i32 %a, i32 %b) nounwind {
 entry:

lbdex/verilog/cpu0.v

@@ -264,7 +264,8 @@ module cpu0(input clock, reset, input [2:0] itype, output reg [2:0] tick,
       // Mathematic 
       ADDiu: regSet(a, Rb+c16);                   // ADDiu Ra, Rb+Cx; Ra<=Rb+Cx
       CMP:  begin 
-        if (Rb < Rc) `N=1; else `N=0; 
+        if (Rb < Rc) `N=1; else `N=0;
+        // `N=(Rb-Rc<0); // why not work for bash make.sh cpu032I el Makefile.builtins?
        `Z=(Rb-Rc==0); 
       end // CMP Rb, Rc; SW=(Rb >=< Rc)
       CMPu:  begin