better logic with new block class

lib/asm/block.rb

@@ -1,32 +1,111 @@
-require_relative "code"
+require_relative 'call_instruction'
+require_relative 'stack_instruction'
+require_relative 'logic_instruction'
+require_relative 'memory_instruction'
 
 module Asm
   
-  # Labels are, like in assembler, a point to jump/branch to. An address in the stream.
+  class Code ; end
-  # To allow for forward branches creation does not fix the position. Set does that.
+  
-  class Label < Code
+  # A Block is the smalles unit of code, a list of instructions as it were
-    def initialize(name , asm)
+  # It is also a point to jump/branch to. An address in the final stream.
-      super
+  # To allow for forward branches creation does not fix the position. Either set or assembling does that.
-      @name = name
+
+  # Blocks are also used to create instructions, and so Block has functions for every cpu instruction
+  # and to make using the apu function easier, there are functions that create registers as well
+  class Block < Code
+
+    def initialize(asm)
+      super()
+      @codes = []
+      @position = 0
       @asm = asm
     end
 
-    # setting a label fixes it's position in the stream. 
+    ArmMachine::REGISTERS.each do |reg , number|
-    # For backwards jumps, positions of labels are known at creation, but for forward off course not.
+      define_method(reg) { Asm::Register.new(reg , number) }
-    # So then one can create a label, branch to it and set it later. 
+    end
+
+    def instruction(clazz, opcode , condition_code , update_status , *args)
+      arg_nodes = []
+      args.each do |arg|
+        if (arg.is_a?(Asm::Register))
+          arg_nodes << arg
+        elsif (arg.is_a?(Integer))
+          arg_nodes << Asm::NumLiteral.new(arg)
+        elsif (arg.is_a?(String))
+          arg_nodes << @asm.add_string(arg)
+        elsif (arg.is_a?(Asm::Label))
+          arg_nodes << arg
+        else
+          raise "Invalid argument #{arg.inspect} for instruction"
+        end
+      end
+      add_code clazz.new(opcode , condition_code , update_status , arg_nodes)
+    end
+    
+    
+    def self.define_instruction(inst , clazz )
+      define_method(inst) do |*args|
+        instruction clazz , inst , :al , 0 , *args
+      end
+      define_method("#{inst}s") do |*args|
+        instruction clazz , inst , :al , 1 , *args
+      end
+      ArmMachine::COND_CODES.keys.each do |suffix|
+        define_method("#{inst}#{suffix}") do |*args|
+          instruction clazz , inst , suffix , 0 , *args
+        end
+        define_method("#{inst}s#{suffix}") do |*args|
+          instruction clazz , inst , suffix , 1 , *args
+        end
+      end
+    end
+
+    [:push, :pop].each do |inst|
+      define_instruction(inst , StackInstruction)
+    end
+
+    [:adc, :add, :and, :bic, :eor, :orr, :rsb, :rsc, :sbc, :sub].each do |inst|
+      define_instruction(inst , LogicInstruction)
+    end
+    [:mov, :mvn].each do |inst|
+      define_instruction(inst , MoveInstruction)
+    end
+    [:cmn, :cmp, :teq, :tst].each do |inst|
+      define_instruction(inst , CompareInstruction)
+    end
+    [:strb, :str , :ldrb, :ldr].each do |inst|
+      define_instruction(inst , MemoryInstruction)
+    end
+    [:b, :bl , :swi].each do |inst|
+      define_instruction(inst , CallInstruction)
+    end
+
+    # setting a block fixes it's position in the stream. 
+    # For backwards jumps, positions of blocks are known at creation, but for forward off course not.
+    # So then one can create a block, branch to it and set it later. 
     def set!
-      @asm.add_value self
+      @asm.add_block self
       self
     end
 
     # Label has no length , 0
     def length 
-      0
+      @codes.sum :length
+    end
+
+    def add_code(kode)
+      kode.at(@position)
+      length = kode.length
+      @position += length
+      @codes << kode
     end
 
-    # nothing to write, we check that the position is what was set
     def assemble(io)
-      raise "Hmm hmm hmm, me thinks i should be somewhere else" if self.position != io.tell
+      @codes.each do |obj|
+        obj.assemble io
+      end
     end
 
   end

lib/asm/call_instruction.rb

@@ -1,3 +1,5 @@
+require_relative "instruction"
+
 module Asm
   # There are only three call instructions in arm branch (b), call (bl) and syscall (swi)
   
@@ -17,7 +19,7 @@ module Asm
       case opcode
       when :b, :bl
         arg = args[0]
-        if arg.is_a? Label
+        if arg.is_a? Block
           diff = arg.position - self.position - 8
           arg = NumLiteral.new(diff)
         end

lib/asm/instruction.rb

@@ -1,9 +1,10 @@
-require_relative "label"
 require_relative "assembly_error"
 require_relative "arm_machine"
 
 module Asm
 
+  class Code ; end
+  
   # Not surprisingly represents an cpu instruction. 
   # This is an abstract base class, with derived classes 
   # Logic / Move / Compare / Stack / Memory (see there)

lib/asm/logic_instruction.rb

@@ -1,3 +1,5 @@
+require_relative "instruction"
+
 module Asm
   # ADDRESSING MODE 1 
   # Logic ,Maths, Move and compare instructions (last three below)

lib/asm/memory_instruction.rb

@@ -1,4 +1,5 @@
 require "asm/nodes"
+require_relative "instruction"
 
 module Asm
   # ADDRESSING MODE 2

lib/asm/program.rb

@@ -1,125 +1,90 @@
-require 'asm/call_instruction'
-require 'asm/stack_instruction'
-require 'asm/logic_instruction'
-require 'asm/memory_instruction'
 require 'asm/nodes'
+require 'asm/block'
 require 'stream_reader'
 require 'stringio'
 require "asm/string_literal"
 
 module Asm
-    
+  
-  class ArmAssembler
+  # Program is the the top-level of the code hierachy, except it is not derived from code
-
+  # instead a Program is a list of blocks (and string constants)
-    ArmMachine::REGISTERS.each do |reg , number|
+  
-      define_method(reg) { Asm::Register.new(reg , number) }
+  # All code is created in blocks (see there) and there are two styles for that, for forward of backward
-    end
+  # referencing. Read function block and add_block and Block.set
+  
+  
+  class Program
 
     def initialize
-      @codes = []
+      @blocks = []
-      @position = 0 # marks not set
-      @labels = []
       @string_table = {}
     end
-    attr_reader  :codes , :position 
 
-    def instruction(clazz, opcode , condition_code , update_status , *args)
+    attr_reader  :blocks 
-      arg_nodes = []
-      args.each do |arg|
-        if (arg.is_a?(Asm::Register))
-          arg_nodes << arg
-        elsif (arg.is_a?(Integer))
-          arg_nodes << Asm::NumLiteral.new(arg)
-        elsif (arg.is_a?(String))
-          arg_nodes << add_string(arg)
-        elsif (arg.is_a?(Asm::Label))
-          arg_nodes << arg
-        else
-          raise "Invalid argument #{arg.inspect} for instruction"
-        end
-      end
-      add_code clazz.new(opcode , condition_code , update_status , arg_nodes)
-    end
-    
-    
-    def self.define_instruction(inst , clazz )
-      define_method(inst) do |*args|
-        instruction clazz , inst , :al , 0 , *args
-      end
-      define_method("#{inst}s") do |*args|
-        instruction clazz , inst , :al , 1 , *args
-      end
-      ArmMachine::COND_CODES.keys.each do |suffix|
-        define_method("#{inst}#{suffix}") do |*args|
-          instruction clazz , inst , suffix , 0 , *args
-        end
-        define_method("#{inst}s#{suffix}") do |*args|
-          instruction clazz , inst , suffix , 1 , *args
-        end
-      end
-    end
 
-    [:push, :pop].each do |inst|
+    # Assembling to string will return a binary string of the whole program, ie all blocks and the 
-      define_instruction(inst , StackInstruction)
+    # strings they use
-    end
+    # As a memory reference this would be callable, but more likely you will hand it over to 
-
+    # an ObjectWriter as the .text section and then link it. And then execute it :-)
-    [:adc, :add, :and, :bic, :eor, :orr, :rsb, :rsc, :sbc, :sub].each do |inst|
-      define_instruction(inst , LogicInstruction)
-    end
-    [:mov, :mvn].each do |inst|
-      define_instruction(inst , MoveInstruction)
-    end
-    [:cmn, :cmp, :teq, :tst].each do |inst|
-      define_instruction(inst , CompareInstruction)
-    end
-    [:strb, :str , :ldrb, :ldr].each do |inst|
-      define_instruction(inst , MemoryInstruction)
-    end
-    [:b, :bl , :swi].each do |inst|
-      define_instruction(inst , CallInstruction)
-    end
-    
     def assemble_to_string
       #put the strings at the end of the assembled code.
       # adding them will fix their position and make them assemble after
       @string_table.values.each do |data|
-        add_code data
+        add_block data
       end
       io = StringIO.new
       assemble(io)
       io.string
     end
 
+    # Add a string to the string table. Strings are global and constant. So only one copy of each 
+    # string exists
+    # Internally StringLiterals are created and stored and during assembly written after the blocks
     def add_string str
       code = @string_table[str]
       return code if code
       data = Asm::StringLiteral.new(str)
       @string_table[str] = data
     end
-  
+
-    def strings
+    # Length of all blocks. Does not take strings into account as they are added after all blocks.
-      @string_table.values
+    # This is used to determine where a block when it is added after creation (see add_block)
-    end
+    def length
-  
+      @blocks.inject(0) {| sum  , item | sum + item.length}
-    def add_code(kode)
-      kode.at(@position)
-      length = kode.length
-      @position += length
-      @codes << kode
-    end
-  
-    def label name
-      label = Label.new(name , self)
-      @labels << label
-      label 
     end
 
+    # call block to create a new (code) block. The simple way is to do this with a block and 
+    # use the yielded block to add code, ie something like:
+    #   prog.block do |loop|
+    #         loop.instance_eval do           #this part you can acheive with calls too
+    #                   mov r0 , 10
+    #                   subs r0 , 1
+    #                   bne block
+    #         end
+    #   end
+    # Easy, because it's a backward jump. For forward branches that doesn't work and so you have to 
+    # create the block without a ruby block. You can then jumpt to it immediately
+    # But the block is not part of the program (since we don't know where) and so you have to add it later
+    def block
+      block = Block.new(self)
+      yield block.set! if block_given? #yield the block (which set returns)
+      block
+    end
+
+    # This is how you add a forward declared block. This is called automatically when you 
+    # call block with ruby block, but has to be done manually if not
+    def add_block block
+      block.at self.length
+      @blocks << block
+    end
+
+    private
+    
     def assemble(io)
-      @codes.each do |obj|
+      @blocks.each do |obj|
         obj.assemble io
       end
     end
-
   end
 end
 

lib/asm/stack_instruction.rb

@@ -1,4 +1,4 @@
-require "asm/instruction"
+require_relative "instruction"
 
 module Asm
   # ADDRESSING MODE 4

lib/crystal.rb

@@ -2,8 +2,7 @@
 $LOAD_PATH.unshift(File.join(File.dirname(__FILE__), '..', ".." , "parslet",'lib'))
 require 'parslet'
 
-require "asm/stack_instruction"
+require "asm/program"
-require "asm/arm_assembler"
 require "elf/object_writer"
 require 'vm/parser'
 require 'vm/nodes'

test/test_crystal.rb

@@ -6,17 +6,20 @@ require_relative 'helper'
 #  adc add and bic eor orr rsb rsc sbc sub mov mvn cmn cmp teq tst b bl bx swi strb
 
 class TestArmAsm < MiniTest::Test
-  # need a code generator, for arm 
+  # need Program and a block (see those classes) 
   def setup
-    @assembler = Asm::ArmAssembler.new
+    @program = Asm::Program.new
+    #no ruby block given, ie need to add this block later by hand
+    @block = @program.block
   end
 
   # code is what the generator spits out, at least one instruction worth (.first)
   # the op code is wat was witten as assembler in the first place and the binary result
   # is reversed and in 4 bytes as ruby can only do 31 bits and so we can't test with just one int (?)
-  def assert_code code , op , should 
+  def assert_code code , op , should
+    @program.add_block @block #assume not added before
     assert_equal op ,  code.opcode
-    binary = @assembler.assemble_to_string
+    binary = @program.assemble_to_string
     assert_equal 4 , binary.length
     index = 0
     binary.each_byte do |byte |
@@ -25,111 +28,111 @@ class TestArmAsm < MiniTest::Test
     end
   end
   def test_adc
-    code = @assembler.instance_eval { adc	r1, r3, r5}.first
+    code = @block.instance_eval { adc	r1, r3, r5}.first
     assert_code code , :adc , [0x05,0x10,0xa3,0xe0] #e0 a3 10 05
   end
   def test_add
-    code = @assembler.instance_eval { add	r1 , r1, r3}.first
+    code = @block.instance_eval { add	r1 , r1, r3}.first
     assert_code code , :add , [0x03,0x10,0x81,0xe0] #e0 81 10 03
   end
   def test_and # inst eval doesn't really work with and
-    code = @assembler.and(	 @assembler.r1 , @assembler.r2 , @assembler.r3).first
+    code = @block.and(	 @block.r1 , @block.r2 , @block.r3).first
     assert_code code , :and , [0x03,0x10,0x02,0xe0] #e0 01 10 03
   end
   def test_b
     # the address is what an assembler calculates (a signed number for the amount of instructions), 
     # ie the relative (to pc) address -8 (pipeline) /4 so save space
     # so the cpu adds the value*4 and starts loading that (load, decode, execute)
-    code = @assembler.instance_eval { b	-1 }.first #this jumps to the next instruction
+    code = @block.instance_eval { b	-1 }.first #this jumps to the next instruction
     assert_code code , :b , [0xff,0xff,0xff,0xea] #ea ff ff fe
   end
   def test_bl #see comment above. bx not implemented (as it means into thumb, and no thumb here)
-    code = @assembler.instance_eval { bl	-1 }.first #this jumps to the next instruction
+    code = @block.instance_eval { bl	-1 }.first #this jumps to the next instruction
     assert_code code , :bl , [0xff,0xff,0xff,0xeb] #ea ff ff fe
   end
   def test_bic
-    code = @assembler.instance_eval { bic	r2 , r2 , r3 }.first
+    code = @block.instance_eval { bic	r2 , r2 , r3 }.first
     assert_code code , :bic , [0x03,0x20,0xc2,0xe1] #e3 c2 20 44
   end
   def test_cmn
-    code = @assembler.instance_eval { cmn	r1 , r2  }.first
+    code = @block.instance_eval { cmn	r1 , r2  }.first
     assert_code code , :cmn , [0x02,0x00,0x71,0xe1] #e1 71 00 02
   end
   def test_cmp
-    code = @assembler.instance_eval { cmp	r1 , r2  }.first
+    code = @block.instance_eval { cmp	r1 , r2  }.first
     assert_code code , :cmp , [0x02,0x00,0x51,0xe1] #e1 51 00 02
   end
   def test_eor
-    code = @assembler.instance_eval { eor	r2 , r2 , r3 }.first
+    code = @block.instance_eval { eor	r2 , r2 , r3 }.first
     assert_code code , :eor , [0x03,0x20,0x22,0xe0] #e0 22 20 03
   end
   def test_ldr
-    code = @assembler.instance_eval { ldr r0, r0 }.first
+    code = @block.instance_eval { ldr r0, r0 }.first
     assert_code code, :ldr ,  [0x00,0x00,0x90,0xe5] #e5 90 00 00
   end
   def test_ldr2
-    code = @assembler.instance_eval { ldr r0, r0 + 4 }.first
+    code = @block.instance_eval { ldr r0, r0 + 4 }.first
     assert_code code, :ldr ,  [0x04,0x00,0x90,0xe5] #e5 90 00 04
   end
   def test_ldrb
-    code = @assembler.instance_eval { ldrb r0, r0 }.first
+    code = @block.instance_eval { ldrb r0, r0 }.first
     assert_code code, :ldrb ,  [0x00,0x00,0xd0,0xe5] #e5 d0 00 00
   end
   def test_orr
-    code = @assembler.instance_eval { orr	r2 , r2 , r3 }.first
+    code = @block.instance_eval { orr	r2 , r2 , r3 }.first
     assert_code code , :orr , [0x03,0x20,0x82,0xe1] #e1 82 20 03
   end
   def test_push
-    code = @assembler.instance_eval { push lr }.first
+    code = @block.instance_eval { push lr }.first
     assert_code code , :push ,  [0x00,0x40,0x2d,0xe9] #e9 2d 40 00
   end
   def test_pop
-    code = @assembler.instance_eval { pop	pc }.first
+    code = @block.instance_eval { pop	pc }.first
     assert_code code , :pop , [0x00,0x80,0xbd,0xe8] #e8 bd 80 00
   end
   
   def test_rsb
-    code = @assembler.instance_eval { rsb	r1 , r2 , r3 }.first
+    code = @block.instance_eval { rsb	r1 , r2 , r3 }.first
     assert_code code , :rsb , [0x03,0x10,0x62,0xe0]#e0 62 10 03
   end
   def test_rsc
-    code = @assembler.instance_eval { rsc	r2 , r3 , r4 }.first
+    code = @block.instance_eval { rsc	r2 , r3 , r4 }.first
     assert_code code , :rsc , [0x04,0x20,0xe3,0xe0]#e0 e3 20 04
   end
   def test_sbc
-    code = @assembler.instance_eval { sbc	r3, r4 , r5 }.first
+    code = @block.instance_eval { sbc	r3, r4 , r5 }.first
     assert_code code , :sbc , [0x05,0x30,0xc4,0xe0]#e0 c4 30 05
   end
   def test_str
-    code = @assembler.instance_eval { str r0, r0 }.first
+    code = @block.instance_eval { str r0, r0 }.first
     assert_code code, :str ,  [0x00,0x00,0x80,0xe5] #e5 81 00 00
   end
   def test_strb
-    code = @assembler.instance_eval { strb r0, r0 }.first
+    code = @block.instance_eval { strb r0, r0 }.first
     assert_code code, :strb ,  [0x00,0x00,0xc0,0xe5] #e5 c0 00 00
   end
   def test_sub
-    code = @assembler.instance_eval { sub r2, r0, 1 }.first
+    code = @block.instance_eval { sub r2, r0, 1 }.first
     assert_code code, :sub ,  [0x01,0x20,0x40,0xe2] #e2 40 20 01 
   end
   def test_swi
-    code = @assembler.instance_eval { swi	0x05 }.first
+    code = @block.instance_eval { swi	0x05 }.first
     assert_code code , :swi , [0x05,0x00,0x00,0xef]#ef 00 00 05
   end
   def test_teq
-    code = @assembler.instance_eval{	teq r1 , r2}.first
+    code = @block.instance_eval{	teq r1 , r2}.first
     assert_code code , :teq , [0x02,0x00,0x31,0xe1] #e1 31 00 02
   end
   def test_tst
-    code = @assembler.instance_eval{	tst r1 , r2}.first
+    code = @block.instance_eval{	tst r1 , r2}.first
     assert_code code , :tst , [0x02,0x00,0x11,0xe1] #e1 11 00 02
   end
   def test_mov
-    code = @assembler.instance_eval { mov r0, 5 }.first
+    code = @block.instance_eval { mov r0, 5 }.first
     assert_code code , :mov , [0x05,0x00,0xa0,0xe3] #e3 a0 10 05
   end
   def test_mvn
-    code = @assembler.instance_eval { mvn r1, 5 }.first
+    code = @block.instance_eval { mvn r1, 5 }.first
     assert_code code , :mvn , [0x05,0x10,0xe0,0xe3] #e3 e0 10 05
   end
 end