dlx user manual

Title	dlx (DLX functional model for ArchC)
Author	Nikolaos Kavvadias 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016
Contact	nikos@nkavvadias.com
Website	http://www.nkavvadias.com
Release Date	02 December 2014
Version	0.1.3
Rev. history
v0.1.3	2014-12-02 Added project logo in README.
v0.1.2	2014-10-23 Documentation corrections.
v0.1.1	2014-10-23 Fixed wrong reference to memory MEM (correct is DM. The simulator binary (dlx.x) is now correctly generated.
v0.1.0	2014-10-22 Updated documentation as README.rst. Changed func from 0x06 to 0x09 for multu. Removed sequ, sneu. Changed func for div (0x19) and divu (0x1A). Newly added files: defines_gdb, modifiers. Updated dlx_opcode_map.xls.
v0.0.4	2006-11-15 Added pcount instruction for basic-block profiling.
v0.0.3	2006-07-01 Corrected optimization instruction methods for j, jal, jr, jalr, beqz, bnez. Alternate behaviors for div, divu added. Fixed copyright notations to manually-written files: (*.ac, dlx-isa.cpp, dlx_syscall.cpp, dlx_gdb_funcs.cpp). Behaviors for addui, subui have been corrected.
v0.0.2	2006-01-01 Changed behavior of j, jal, beqz, bnez according to what is expected by the binutils DLX port. Fixed issue with jr instruction. New encoding for the halt instruction. Changed register notation to comply to DLX conventions: (r0-r31) and alternate notation: (zero,at,v0-v1,a0-a3,t0-t9,s0-s7, k0-k1,gp,sp,fp,ra) Both prefixed (by a dollar sign) and unprefixed symbolic register names should be accepted. Disabled non-standard DLX instructions, along with mvts, mvfs. The standard mult, multu, div, divu opcodes are now used. Testsuite directory removed. The acstone benchmarks should be used instead for the purpose of benchmarking the DLX model.
v0.0.1	2005-12-26 First public version. Most integer instruction set functionality has been added. Very few applications have been tested: fib.s (generated by dlxgcc-2.7.2.3 and slightly modified) loadi.s (tests load immediate pseudo-instructions)

1. Introduction

This is the DLX ArchC (http://www.archc.org) functional model. This model has the system call emulation functions implemented, so it is a good idea to turn on the ABI option.

2. File listing

The dlx distribution includes the following files:

/dlx	Top-level directory
AUTHORS	List of dlx authors.
LICENSE	The modified BSD license governs dlx.
README.html	HTML version of README.
README.pdf	PDF version of README.
README.rst	This file.
VERSION	Current version of the project sources.
bp_conf.ac	Branch predictor description (only for archc-1.5.1.bp2).
defines_gdb	Macro definitions for GDB integration.
dlx.ac	Register, memory and cache model for dlx.
dlx.png	PNG image for the dlx project logo.
dlx_gdb_funcs.cpp	GDB support for the DLX simulator.
dlx_isa.ac	Instruction encodings and assembly formats.
dlx_opcode_map.vsd	Incomplete MS Visio drawing of the DLX opcode map.
dlx_opcode_map.xls	Excel spreadsheet containing the DLX opcode map.
dlx_syscall.cpp	OS call emulation support for DLX.
dlx_isa.cpp	Instruction behaviors.
modifiers	Instruction encoding and decoding modifiers.
rst2docs.sh	Bash script for generating the HTML and PDF versions of the documentation (README).

3. Usage

To generate the interpreted simulator, the acsim executable is ran:

$ acsim dlx.ac [-g -abi -gdb]       # (create the simulator)
$ make -f Makefile.archc            # (compile)
$ ./dlx.x --load=<file-path> [args] # (run an application)

To generate the compiled application simulator, the accsim executable is ran:

$ accsim dlx.ac <file-path>         # (create specialized simulator)
$ make -f Makefile.archc            # (compile)
$ ./dlx.x [args]                    # (run the application)

The [args] are optional arguments for the application.

There are two formats recognized for application <file-path>:

• ELF binary matching ArchC specifications
• hexadecimal text file for ArchC

In order to generate the binary utilities port (binutils port), the acbingen.sh driver script must be used. This should be called as follows:

$ acbingen.sh -adlx -i`pwd`/../dlx-tools/ dlx.ac

for generating the binutils port executables. This includes the following tools:

• addr2line
• ar
• as
• c++filt
• gdb (the GDB port is also generated in the same directory)
• gdbtui
• ld
• nm
• objcopy
• objdump
• ranlib
• readelf
• size
• strings
• strip

4. General observations

1. Some non-classical DLX instructions (available in the DLX binutils target) might be added in the future. These are:

• bswap (BSWAPF) --> A byte swap instruction
• ldstbu (LSBUOP) --> Atomic load-store byte unsigned
• ldsthu (LSHUOP) --> Atomic load-store halfword unsigned
• ldstw (LSWOP) --> Atomic load-store word

2. mult, multu, div, divu instructions have different opcodes to the binutils DLX. Also, div, divu produce a single 32-bit result (the quotient). Probably, rem, remu instructions will be added to produce the remainder of a division. For 64-bit result multiplication maybe a good choice is to provide multl, multlu primitives, for which results are written in two consecutive registers (integer registers).
3. There are no HI/LO registers (I think this is the actual intent in the Patterson book).
4. Multiplication and division DONNOT use the floating-point register file. For this reason, mvts, mvfs instructions are currently unimplemented.
5. Loading 32-bit constants will be available via appropriate pseudo- instructions not requiring the HI/LO registers, and for the following formats:

li %dest, #hi-16bit-constant, #lo-16bit-constant

li %dest, #32bit-constant

6. For future provision of a coprocessor (maybe this is an overkill for the DLX?) some opcodes MIGHT be moved, e.g.:
   • Move opcode(J)=0x02, opcode(JAL)=0x03 to e.g. 0x06,0x07, respectively. (PREFERRED)
   • Move opcode(BEQZ), opcode(BNEZ) to 0x16, 0x17.
   • Then the 0x01-0x04 primary opcodes would be used for 4 optional coprocessors.