Dalvik "mterp" README
NOTE: Find rebuilding instructions at the bottom of this file.
==== Overview ====
This is the source code for the Dalvik interpreter. The core of the
original version was implemented as a single C function, but to improve
performance we rewrote it in assembly. To make this and future assembly
ports easier and less error-prone, we used a modular approach that allows
development of platform-specific code one opcode at a time.
The original all-in-one-function C version still exists as the "portable"
interpreter, and is generated using the same sources and tools that
generate the platform-specific versions.
Every configuration has a "config-*" file that controls how the sources
are generated. The sources are written into the "out" directory, where
they are picked up by the Android build system.
The best way to become familiar with the interpreter is to look at the
generated files in the "out" directory, such as out/InterpC-portstd.c,
rather than trying to look at the various component pieces in (say)
armv5te.
==== Platform-specific source generation ====
The architecture-specific config files determine what goes into two
generated output files (InterpC-<arch>.c, InterpAsm-<arch>.S). The goal is
to make it easy to swap C and assembly sources during initial development
and testing, and to provide a way to use architecture-specific versions of
some operations (e.g. making use of PLD instructions on ARMv6 or avoiding
CLZ on ARMv4T).
Depending on architecture, instruction-to-instruction transitions may
be done as either computed goto or jump table. In the computed goto
variant, each instruction handler is allocated a fixed-size area (e.g. 64
byte). "Overflow" code is tacked on to the end. In the jump table variant,
all of the instructions handlers are contiguous and may be of any size.
The interpreter style is selected via the "handler-size" command (see below).
When a C implementation for an instruction is desired, the assembly
version packs all local state into the Thread structure and passes
that to the C function. Updates to the state are pulled out of
"Thread" on return.
The "arch" value should indicate an architecture family with common
programming characteristics, so "armv5te" would work for all ARMv5TE CPUs,
but might not be backward- or forward-compatible. (We *might* want to
specify the ABI model as well, e.g. "armv5te-eabi", but currently that adds
verbosity without value.)
==== Config file format ====
The config files are parsed from top to bottom. Each line in the file
may be blank, hold a comment (line starts with '#'), or be a command.
The commands are:
handler-style <computed-goto|jump-table|all-c>
Specify which style of interpreter to generate. In computed-goto,
each handler is allocated a fixed region, allowing transitions to
be done via table-start-address + (opcode * handler-size). With
jump-table style, handlers may be of any length, and the generated
table is an array of pointers to the handlers. The "all-c" style is
for the portable interpreter (which is implemented completely in C).
[Note: all-c is distinct from an "allstubs" configuration. In both
configurations, all handlers are the C versions, but the allstubs
configuration uses the assembly outer loop and assembly stubs to
transition to the handlers]. This command is required, and must be
the first command in the config file.
handler-size <bytes>
Specify the size of the fixed region, in bytes. On most platforms
this will need to be a power of 2. For jump-table and all-c
implementations, this command is ignored.
import <filename>
The specified file is included immediately, in its entirety. No
substitutions are performed. ".cpp" and ".h" files are copied to the
C output, ".S" files are copied to the asm output.
asm-stub <filename>
The named file will be included whenever an assembly "stub" is needed
to transfer control to a handler written in C. Text substitution is
performed on the opcode name. This command is not applicable to
to "all-c" configurations.
asm-alt-stub <filename>
When present, this command will cause the generation of an alternate
set of entry points (for computed-goto interpreters) or an alternate
jump table (for jump-table interpreters).
op-start <directory>
Indicates the start of the opcode list. Must precede any "op"
commands. The specified directory is the default location to pull
instruction files from.
op <opcode> <directory>
Can only appear after "op-start" and before "op-end". Overrides the
default source file location of the specified opcode. The opcode
definition will come from the specified file, e.g. "op OP_NOP armv5te"
will load from "armv5te/OP_NOP.S". A substitution dictionary will be
applied (see below).
alt <opcode> <directory>
Can only appear after "op-start" and before "op-end". Similar to the
"op" command above, but denotes a source file to override the entry
in the alternate handler table. The opcode definition will come from
the specified file, e.g. "alt OP_NOP armv5te" will load from
"armv5te/ALT_OP_NOP.S". A substitution dictionary will be applied
(see below).
op-end
Indicates the end of the opcode list. All kNumPackedOpcodes
opcodes are emitted when this is seen, followed by any code that
didn't fit inside the fixed-size instruction handler space.
The order of "op" and "alt" directives are not significant; the generation
tool will extract ordering info from the VM sources.
Typically the form in which most opcodes currently exist is used in
the "op-start" directive. For a new port you would start with "c",
and add architecture-specific "op" entries as you write instructions.
When complete it will default to the target architecture, and you insert
"c" ops to stub out platform-specific code.
For the <directory> specified in the "op" command, the "c" directory
is special in two ways: (1) the sources are assumed to be C code, and
will be inserted into the generated C file; (2) when a C implementation
is emitted, a "glue stub" is emitted in the assembly source file.
(The generator script always emits kNumPackedOpcodes assembly
instructions, unless "asm-stub" was left blank, in which case it only
emits some labels.)
==== Instruction file format ====
The assembly instruction files are simply fragments of assembly sources.
The starting label will be provided by the generation tool, as will
declarations for the segment type and alignment. The expected target
assembler is GNU "as", but others will work (may require fiddling with
some of the pseudo-ops emitted by the generation tool).
The C files do a bunch of fancy things with macros in an attempt to share
code with the portable interpreter. (This is expected to be reduced in
the future.)
A substitution dictionary is applied to all opcode fragments as they are
appended to the output. Substitutions can look like "$value" or "${value}".
The dictionary always includes:
$opcode - opcode name, e.g. "OP_NOP"
$opnum - opcode number, e.g. 0 for OP_NOP
$handler_size_bytes - max size of an instruction handler, in bytes
$handler_size_bits - max size of an instruction handler, log 2
Both C and assembly sources will be passed through the C pre-processor,
so you can take advantage of C-style comments and preprocessor directives
like "#define".
Some generator operations are available.
%include "filename" [subst-dict]
Includes the file, which should look like "armv5te/OP_NOP.S". You can
specify values for the substitution dictionary, using standard Python
syntax. For example, this:
%include "armv5te/unop.S" {"result":"r1"}
would insert "armv5te/unop.S" at the current file position, replacing
occurrences of "$result" with "r1".
%default <subst-dict>
Specify default substitution dictionary values, using standard Python
syntax. Useful if you want to have a "base" version and variants.
%break
Identifies the split between the main portion of the instruction
handler (which must fit in "handler-size" bytes) and the "sister"
code, which is appended to the end of the instruction handler block.
In jump table implementations, %break is ignored.
%verify "message"
Leave a note to yourself about what needs to be tested. (This may
turn into something more interesting someday; for now, it just gets
stripped out before the output is generated.)
The generation tool does *not* print a warning if your instructions
exceed "handler-size", but the VM will abort on startup if it detects an
oversized handler. On architectures with fixed-width instructions this
is easy to work with, on others this you will need to count bytes.
==== Using C constants from assembly sources ====
The file "common/asm-constants.h" has some definitions for constant
values, structure sizes, and struct member offsets. The format is fairly
restricted, as simple macros are used to massage it for use with both C
(where it is verified) and assembly (where the definitions are used).
If a constant in the file becomes out of sync, the VM will log an error
message and abort during startup.
==== Development tips ====
If you need to debug the initial piece of an opcode handler, and your
debug code expands it beyond the handler size limit, you can insert a
generic header at the top:
b ${opcode}_start
%break
${opcode}_start:
If you already have a %break, it's okay to leave it in place -- the second
%break is ignored.
==== Rebuilding ====
If you change any of the source file fragments, you need to rebuild the
combined source files in the "out" directory. Make sure the files in
"out" are editable, then:
$ cd mterp
$ ./rebuild.sh
As of this writing, this requires Python 2.5. You may see inscrutible
error messages or just general failure if you have a different version
of Python installed.
The ultimate goal is to have the build system generate the necessary
output files without requiring this separate step, but we're not yet
ready to require Python in the build.
==== Interpreter Control ====
The central mechanism for interpreter control is the InterpBreak struture
that is found in each thread's Thread struct (see vm/Thread.h). There
is one mandatory field, and two optional fields:
subMode - required, describes debug/profile/special operation
breakFlags & curHandlerTable - optional, used lower subMode polling costs
The subMode field is a bitmask which records all currently active
special modes of operation. For example, when Traceview profiling
is active, kSubModeMethodTrace is set. This bit informs the interpreter
that it must notify the profiling subsystem on each method entry and
return. There are similar bits for an active debugging session,
instruction count profiling, pending thread suspension request, etc.
To support special subMode operation the simplest mechanism for the
interpreter is to poll the subMode field before interpreting each Dalvik
bytecode and take any required action. In fact, this is precisely
what the portable interpreter does. The "FINISH" macro expands to
include a test of subMode and subsequent call to the "dvmCheckBefore()".
Per-instruction polling, however, is expensive and subMode operation is
relative rare. For normal operation we'd like to avoid having to perform
any checks unless a special subMode is actually in effect. This is
where curHandlerTable and breakFlags come in to play.
The mterp fast interpreter achieves much of its performance advantage
over the portable interpreter through its efficient mechanism of
transitioning from one Dalvik bytecode to the next. Mterp for ARM targets
uses a computed-goto mechanism, in which the handler entrypoints are
located at the base of the handler table + (opcode * 64). Mterp for x86
targets instead uses a jump table of handler entry points indexed
by the Dalvik opcode. To support efficient handling of special subModes,
mterp supports two sets of handler entries (for ARM) or two jump
tables (for x86). One handler set is optimized for speed and performs no
inter-instruction checks (mainHandlerTable in the Thread structure), while
the other includes a test of the subMode field (altHandlerTable).
In normal operation (i.e. subMode == 0), the dedicated register rIBASE
(r8 for ARM, edx for x86) holds a mainHandlerTable. If we need to switch
to a subMode that requires inter-instruction checking, rIBASE is changed
to altHandlerTable. Note that this change is not immediate. What is actually
changed is the value of curHandlerTable - which is part of the interpBreak
structure. Rather than explicitly check for changes, each thread will
blindly refresh rIBASE at backward branches, exception throws and returns.
The breakFlags field tells the interpreter control mechanism whether
curHandlerTable should hold the real or alternate handler base. If
non-zero, we use the altHandlerBase. The bits within breakFlags
tells dvmCheckBefore which set of subModes need to be checked.
See dvmCheckBefore() for subMode handling, and dvmEnableSubMode(),
dvmDisableSubMode() for switching on and off.