Age | Commit message (Collapse) | Author |
|
Clarify beam_load error message on file/module mismatch
|
|
This is particularly important in case insensitive filesystems,
where attempting to invoke a module with the wrong case leads
to confusing error messages:
1> erlpress_core:foo().
beam/beam_load.c(1428): Error loading module 'erlpress_core':
module name in object code is erlPress_core
Loading of erlPress_core.beam failed: :badfile
This commit replaces object code by BEAM file and improves
the readability of the message.
|
|
Use S operands instead of s operands for a slight speed increase
and reduction in code size of process_main(). Use micro instructions
for frequently executed instructions.
While at it, use safe multiplication in gen_get_integer() in
beam_load.c.
|
|
Starting in OTP 19 (in commit 9504c0dd71d0), the compiler emits
a test_unit instruction instead of a skip instruction at the end
of binary. We can do the same replacement in the loader to get
rid of the i_bs_skip_bits_all2 instruction.
|
|
The new compiler required adding support for Y register for all
binary matching instructions. That was (intentionally) done in a
naive way that simplicated duplicated the entire body of each
instruction.
Now it's time to be less naive. Rewrite the binary matching
instructions using micro instructions. Because some of the binary
instructions are huge, that will significantly decrease the size of
process_main().
When compiling with clang, a huge process_main() would mess up
profile-guide optimization resulting in a significant performance
degradation. On my Mac, profile-guide optimzation would decrease
the estone benchmark by 100K estones (about 20 percent). This commit
gives me back the lost estones.
|
|
|
|
Of the `move_dup` instructions, only `move_dup x x x` was
frequently used. Remove the other register combinations.
With those instruction `move_dup` instructions removed, it
is necessary to add new predicates to avoid unsafe translation
to `move_shift` and `move2_par`.
Also add additional transformations to transform more `move`
instructions into `move2_par`. The existing transformation
would require the `move` instructions to be in the "right"
order in order to be transformed.
Remove `move3 x y x y x y` because it turns out to be rarely
executed.
|
|
|
|
The is_function2 instruction is executed surprisingly
frequently when running dialyzer or the compiler. It
cannot hurt to optimize it a little.
|
|
Summary: This commit simplifies the implementation of the "GC BIFs" so
that they no longer need to do a garbage collection, removing duplicate
code for all GC BIFs in the runtime system, as well as potentially
reducing the size of the loaded BEAM code by using shorter
instructions calling those BIFs.
A GC BIF is a guard BIF that will do a garbage
collection if it needs to build anything on the heap.
For example, `abs/1` is a GC BIF because it might need to
allocate space on the heap (if the result is a floating point
number or the resulting integer is a bignum).
Before R12, a guard BIF (such as `abs/1`) that need to allocate
heap space would allocate outside of process's main heap, in
a heap fragment.
GC BIFs were introduced in R12B to support literals. During garbage
collection it become necessary to quickly test whether a term was
a literal. To make the check simple, guards BIFs were no longer
allowed to create heap fragments. Instead GC BIFs were introduced.
In OTP 19, the implementation of literals was changed to support
storing messages in heap fragments outside of the main heap for a
process. That change again made it allowed for guard BIFs to create
heap fragments when they need to build terms on the heap.
It would even be possible for the guard BIFs to build directly
on the main heap if there is room there, because the compiler
assumes that a new `test_heap/2` instruction must be emitted
when building anything after calling a GC BIF. (We don't do that
in this commit; see below.)
This commit simplifies the implementation of the GC BIFs in
the runtime system.
Each GC BIF had a dual implementation: one that was used when the GC
BIF was called directly and one used when it was called via
`apply/3`. For example, `abs/1` was implemented in `abs_1()` and
`erts_gc_abs_1()`. This commit removes the GC version of each BIF. The
other version that allocates heap space using `HAlloc()` is updated to
use the new `HeapFragOnlyAlloc()` macro that will allocate heap
space in a heap fragment outside of the main heap.
Because the BIFs will allocate outside of the main heap, the same
`bif` instructions used by nonbuilding BIFs can be used for the
(former) GC BIFs. Those instructions don't use the macros that save
and restore the heap and stack pointers (SWAPOUT/SWAPIN). If the
former GC BIFs would build on the main heap, either new instructions
would be needed, or SWAPOUT/SWAPIN instructions would need to be added
to the `bif` instructions.
Instructions that call the former GC BIFs don't need the operand
that specifies the number of live X registers. Therefore, the
instructions that call the BIFs are usually one word shorter.
|
|
The upcoming beam_ssa_bsm pass allows match contexts to be used
across function calls that take said context as an argument, which
means it's fairly common for them to end up in Y registers.
|
|
This introduces a similar optimisation for normal funs
to what was introduced for external funs in #1725.
It is possible to allocate the fun as a literal, if it does not capture
the environment (i.e. it does not close over any variables).
Unfortunately it's not possible to do this in the compiler due to
problems with representation of such functions in the `.beam` files.
Fortunately, we can do this in the loader.
Simple evaluation shows that functions that don't capture the
enviornment consistute over 60% of all funs in the source code of
Erlang/OTP itself.
The only downside is that we lose a meningful value in the `pid` field
of the fun. The goal of this field, beyond debugging, was to be
able to identify the original node of a function. To be able to still do
this, the functions that are created in the loader are assigned the init
pid as the creator.
To solve issues with staryp, initially set the `erts_init_process_id`
to `ERTS_INVALID_PID` and skip the described optimisation if the value
is still uninitialised.
|
|
Compile external fun expressions to literals
OTP-15003
|
|
The expressions fun M:F/A, when all elements are literals are also
treated as a literal. Since they have consistent representation and
don't depend on the code currently loaded in the VM, this is safe.
This can provide significant performance improvements in code using such
functions extensively - a full function call to erlang:make_fun/3 is
replaced by a single move instruction and no register shuffling or
saving registers to stack is necessary. Additionally, compound data
types that contain such external functions as elements can be treated as
literals too.
The commit also changes the representation of external funs to be a
valid Erlang syntax and adds support for literal external funs to core
Erlang.
|
|
|
|
|
|
|
|
|
|
|
|
particularly slow erlc when compiler is hipe compiled.
hipe_unified_loader:load did not patch external call sites
and instead caused a double hipe mode switch per call.
hipe_unified_loader:load is only used
for early modules first loaded as beam
and by code:atomic_load and friends.
|
|
* maint:
Fix encoding of filenames in stacktraces
|
|
|
|
|
|
|
|
When compiling Erlang source code, the literal area for the
module can only contain data types that have a literal
syntax.
However, it is possible to sneak in other data types
(such as references) in the literal pool by compiling from
abstract or assembly code. Those "fake literals" would work
fine, but would crash the runtime system when the module containing
the literals was purged.
Although fake literals are not officially supported, the
runtime should not crash when attempting to use them.
Therefore, fix the garbage collection of literals and releasing
of literal areas.
https://bugs.erlang.org/browse/ERL-508
|
|
The 's' operand overloads the tags for pids and ports to represent
X and Y registers, respectively. At load time, refuse to load the
module if the "literal" term is not a pid or port, as it would be
interpreted as a register.
This does not happen with normally compiled code, but it can happen
if the compiler (or beam_asm) is abused like in the following example:
make_bad() ->
Pid = self(),
Forms = [{attribute, 0, module, bad_s_operand},
{attribute, 0, export, [{test, 0}]},
{function, 0, test, 0,
[{clause, 0, [], [],
[{call,0,{atom,0,tuple_size},[{integer, 0, Pid}]}]}]}],
{ok, Module, Bin} = compile:forms(Forms, [no_copt,no_postopt,report_errors]),
code:load_binary(Module, "bad_s_operand.erl", Bin).
With this commit applied, the following message will be printed
when make_bad() is run:
=ERROR REPORT==== 10-Nov-2017::14:47:59 ===
Loading of bad_s_operand.erl failed: badfile
=ERROR REPORT==== 10-Nov-2017::14:47:59 ===
beam/beam_load.c(2396): Error loading function bad_s_operand:test/0: op bif1_body: bsd:
the term '<0.60.0>' would be confused with a register
|
|
* lukas/erts/misc_fixes:
erts: Fix a bunch of compiler warnings
kernel: Fix gen_tcp_misc indentation
erts: Fail port_SUITE:huge_env if error code > 127
erts: Add lcnt prototype for dist locks update
|
|
|
|
* sverker/on_load-nonblocking/OTP-14680:
erts: Remove scheduler blocking during finish_after_on_load_2
|
|
for normal case.
We still block for default trace and hipe.
|
|
On 64-bit machines where the C code is always at address below 4Gb,
pack one or more operands into the instruction word.
|
|
In the 'P' operator, don't assume that a packed target label ('f'
or 'j') is always the leftmost argument. Instead, transfer the
patch position from the accumulator to the stack.
|
|
The BeamOp() macro in erl_vm.h is clumsy to use. All users
cast the return value to BeamInstr.
Define new macros that are easier to use. In the future,
we might want to pack an operand into the same word as
the pointer to the instruction, so we will define two macros.
BeamIsOpCode() is used to rewrite code like this:
if (Instr == (BeamInstr) BeamOp(op_i_func_info_IaaI) {
...
}
to:
if (BeamIsOpCode(Instr, op_i_func_info_IaaI)) {
...
}
BeamOpCodeAddr(op_apply_bif) is used when we need the address
for an instruction.
Also elimiminate the global variables em_* in beam_emu.c.
They are not really needed. Use the BeamOpCodeAddr() macro
instead.
|
|
|
|
|
|
Pack sequences of trailing 'f' operands for instructions
such at jump_on_val or i_select_val_lins.
|
|
|
|
Relative failure in itself is not an optimization, but we plan to
pack failure labels in the future to save memory.
|
|
It has served its purpose.
|
|
Introduce new macros that can be used for relative jumps and
use them consistently.
Test that everything works by using a non-zero constant JUMP_OFFSET.
The loader subtracts JUMP_OFFSET from loaded labels, and all
instructions that use 'f' operands add it back.
|
|
De-duplicate bignums in the literal pool
|
|
Duplicate literals in the literal pool is a waste of memory.
Also, having unique literals can simplify some loader optimizations,
because we can known that if two literal indices are distinct, the
values must be distcint too.
|
|
Starting from R16B, floating point constants are stored
in the literal pool.
The backward compatibility support for the old representation
of floats has been kept long enough. Also, we might want to
ensure that all literals are unique in the future, and we certainly
don't want to update this code to ensure uniqueness.
|
|
The type 'd' could be used both for destination registers and
source register.
Restrict the 'd' type to only be used for destinations, and
introduce the new 'S' type to be used when a source must be
a register.
|
|
Update the pack engine to safely push literal operands to the pack
stack and to safely pop them back to another code address. That
will allow packing of more instructions.
|
|
As a preparation for potentially improving packing in the future,
we will need to make sure that packable types have a defined maximum
size.
The packer algorithm assumes that two 'I' operands can be packed
into one 64-bit word, but there are instructions that use an 'I'
operand to store a pointer. It only works because those instructions
are not packed for other reasons.
Introduce the 'W' type and use it for operands that don't fit in
32 bits.
|
|
The transformations were incorrect.
|
|
The instruction put_map_assoc/5 (used for updating a map) has a
failure operand, but it can't actually fail provided that its "map"
argument is a map.
The following code:
M#{key=>value}.
will be compiled to:
{test,is_map,{f,3},[{x,0}]}.
{line,[...]}.
{put_map_assoc,{f,0},{x,0},{x,0},1,{list,[{atom,key},{atom,value}]}}.
return.
{label,3}.
%% Code that produces a 'badmap' exception follows.
Because of the is_map instruction, {x,0} always contains a map when
the put_map_assoc instruction is executed. Therefore we can remove
the failure operand. That will save one word, and also eliminate
two tests at run-time.
The only problem is that the compiler in OTP 17 did not emit a
is_map instruction before the put_map_assoc instruction. Therefore,
we must add an instruction that tests for a map if the code was
compiled with the OTP 17 compiler.
Unfortunately, there is no safe and relatively easy way to known that
the OTP 17 compiler was used, so we will check whether a compiler
before OTP 20 was used. OTP 20 introduced a new chunk type for atoms,
which is trivial to check.
|
|
|
|
|