| Age | Commit message (Collapse) | Author |
|---|
|
Move size=all binary clause pruning to v3_kernel
|
|
The advantage of moving it up is that it reduces the
size of the code emitted by v3_kernel, speeding
v3_kernel itself and beam_kernel_to_ssa pass.
|
|
Prior to this patch, v3_kernel would do multiple
passes on the clauses to group them. This commit
unrolls those passes, making v3_kernel up to 10%
faster in those cases.
|
|
The expansion of record field updates, when more than one field is
updated, but not a majority of the fields, will create a sequence of
calls to `erlang:setelement(Index, Value, Tuple)` where Tuple in the
first call is the original record tuple, and in the subsequent calls
Tuple is the result of the previous call. Furthermore, all Index
values are constant positive integers, and the first call to
`setelement` will have the greatest index. Thus all the following
calls do not actually need to test at run-time whether Tuple has type
tuple, nor that the index is within the tuple bounds.
Since OTP R7, the `sys_core_dsetel` pass, run as the very last Core
Erlang pass, has optimized this sequence of `setelement` calls to use
a special destructive version of `setelement` (called
`set_tuple_element`) for all but the very first `setelement` in the
sequence.
It turns out that the presence of the `set_tuple_element` in SSA code
is awkward and can prevent or complicate type analysis and aggressive
optimizations.
Therefore, this commit removes the `sys_core_dsetel` pass and
reimplements it for SSA code. The optimization will be done in the
`beam_ssa_pre_codegen` pass (that is, just before code generation and
after running all other SSA code optimization passes).
In most cases, the resulting BEAM code is identical to previous
code. For a few modules, the BEAM code is actually slightly better,
with smaller stack frames.
|
|
This has been superseded by bs_get_tail/3. Note that it is NOT
removed from the emulator or beam_disasm, as old modules are still
legal.
|
|
The upcoming beam_ssa_bsm pass makes this redundant.
|
|
v3_codegen is replaced by three new passes:
* beam_kernel_to_ssa which translates the Kernel Erlang format
to a new SSA-based intermediate format.
* beam_ssa_pre_codegen which prepares the SSA-based format
for code generation, including register allocation. Registers
are allocated using the linear scan algorithm.
* beam_ssa_codegen which generates BEAM assembly code from the
SSA-based format.
It easier and more effective to optimize the SSA-based format before X
and Y registers have been assigned. The current optimization passes
constantly have to make sure no "holes" in the X register assignments
are created (that is, that no X register becomes undefined that an
allocation instruction depends on).
This commit also introduces the following optimizations:
* Replacing of tuple matching of records with the is_tagged_tuple
instruction. (Replacing beam_record.)
* Sinking of get_tuple_element instructions to just before the first
use of the extracted values. As well as potentially avoiding
extracting tuple elements when they are not actually used on all
executions paths, this optimization could also reduce the number
values that will need to be stored in Y registers. (Similar to
beam_reorder, but more effective.)
* Live optimizations, removing the definition of a variable that is
not subsequently used (provided that the operation has no side
effects), as well strength reduction of binary matching by replacing
the extraction of value from a binary with a skip instruction. (Used
to be done by beam_block, beam_utils, and v3_codegen.)
* Removal of redundant bs_restore2 instructions. (Formerly done
by beam_bs.)
* Type-based optimizations across branches. More effective than
the old beam_type pass that only did type-based optimizations in
basic blocks.
* Optimization of floating point instructions. (Formerly done
by beam_type.)
* Optimization of receive statements to introduce recv_mark and
recv_set instructions. More effective with far fewer restrictions
on what instructions are allowed between creating the reference
and entering the receive statement.
* Common subexpression elimination. (Formerly done by beam_block.)
|
|
|
|
Use integer variable names instead of atoms in v3_core, sys_core_fold,
and v3_kernel to avoid overflowing the atom table.
It is a deliberate design decision to calculate the first free integer
variable name (in sys_core_fold and v3_kernel) instead of somehow
passing it from one pass to another. I don't want that kind of
dependency between compiler passes. Also note that the next free
variable name is not easily available after running the inliner.
|
|
For unclear reasons, v3_kernel attempts to guarantee that #k_try{}
always has at least one return value, even if it will never be
used. I said "attempts", because the handler block that is executed
when an exception is caught does not have the same guarantee. That
means that if an exception is thrown, the return value will not
actually be set.
In practice, however, this is not a problem for the existing code
generator (v3_codegen). The generated code will still be safe.
If we are to rewrite the code generator to generate an SSA-based
intermediate format, this inconsistency *will* cause problems
when creating phi nodes.
While at it, also remove an unecessary creation of new variables
in generation of #k_try_enter{}.
|
|
|
|
v3_kernel could generate a #k_break{} with only one variable, even if
the preceding code and succeding code expected more than one value. It
happened to work anyway because the value returned from the break was
not actually used.
|
|
The v3_life pass does not do enough to be worth being its own
pass. Essentially it does two things:
* Calculates life-time information starting from the annotations
that v3_kernel provides. That part can be moved into v3_codegen.
* Rewrites the Kernel Erlang records to similar plain tuples
(for example, #k_cons{hd=Hd,tl=Tl} is rewritten to {cons,Hd,Tl}).
That rewriting is not needed and can be eliminated.
|
|
If a type only has one clause and if the pattern is literal,
the matching can be done more efficiently by directly comparing
with the literal.
Example:
find(String, "") -> String;
find(String, <<>>) -> String;
find(String, SearchPattern) ->
.
.
.
Without this optimization, the relevant part of the code would look
this:
{test,bs_start_match2,{f,3},2,[{x,1},0],{x,2}}.
{test,bs_test_tail2,{f,4},[{x,2},0]}.
return.
{label,3}.
{test,is_nil,{f,4},[{x,1}]}.
return.
{label,4}.
.
.
.
That is, if {x,1} is a binary, a match context will be built to
test whether {x,1} is an empty binary.
With the optimization, the code will look this:
{test,is_eq_exact,{f,3},[{x,1},{literal,<<>>}]}.
return.
{label,3}.
{test,is_nil,{f,4},[{x,1}]}.
return.
{label,4}.
.
.
.
|
|
|
|
|
|
The new Dbgi chunk returns data in the following format:
{debug_info_v1, Backend, Data}
This allows compilers to store the debug info in different
formats. In order to retrieve a particular format, for
instance, Erlang Abstract Format, one may invoke:
Backend:debug_info(erlang_v1, Module, Data, Opts)
Besides introducing the chunk above, this commit also:
* Changes beam_lib:chunk(Beam, [:abstract_code]) to
read from the new Dbgi chunk while keeping backwards
compatibility with old .beams
* Adds the {debug_info, {Backend, Data}} option to
compile:file/2 and friends that are stored in the
Dbgi chunk. This allows the debug info encryption
mechanism to work across compilers
* Improves dialyzer to work directly on Core Erlang,
allowing languages that do not have the Erlang
Abstract Format to be dialyzer as long as they emit
the new chunk and their backend implementation is
available
Backwards compatibility is kept across the board except
for those calling beam_lib:chunk(Beam, ["Abst"]), as the
old chunk is no longer available. Note however the "Abst"
chunk has always been optional.
Future OTP versions may remove parsing the "Abst" chunk
altogether from beam_lib once Erlang 19 and earlier is no
longer supported.
The current Dialyzer implementation still supports earlier
.beam files and such may also be removed in future versions.
|
|
* egil/compiler/opt-maps-pattern-matching/OTP-14072:
compiler: Add regression tests
compiler: Optimize maps pattern matching
compiler: Allow for unaligned match argument in value groups
|
|
|
|
The compiler produces poor code for complex guard expressions with andalso/orelse.
Here is an example from the filename module:
-define(IS_DRIVELETTER(Letter),(((Letter >= $A) andalso (Letter =< $Z)) orelse
((Letter >= $a) andalso (Letter =< $z)))).
skip_prefix(Name, false) ->
Name;
skip_prefix([L, DrvSep|Name], DrvSep) when ?IS_DRIVELETTER(L) ->
Name;
skip_prefix(Name, _) ->
Name.
beam_bool fails to simplify the code for the guard, leaving several 'bif'
instructions:
{function, skip_prefix, 2, 49}.
{label,48}.
{line,[{location,"filename.erl",187}]}.
{func_info,{atom,filename},{atom,skip_prefix},2}.
{label,49}.
{test,is_ne_exact,{f,52},[{x,1},{atom,false}]}.
{test,is_nonempty_list,{f,52},[{x,0}]}.
{get_list,{x,0},{x,2},{x,3}}.
{test,is_nonempty_list,{f,52},[{x,3}]}.
{get_list,{x,3},{x,4},{x,5}}.
{bif,'=:=',{f,52},[{x,1},{x,4}],{x,6}}.
{test,is_ge,{f,50},[{x,2},{integer,65}]}.
{bif,'=<',{f,52},[{x,2},{integer,90}],{x,7}}.
{test,is_eq_exact,{f,51},[{x,7},{atom,false}]}.
{test,is_ge,{f,50},[{x,2},{integer,97}]}.
{bif,'=<',{f,52},[{x,2},{integer,122}],{x,7}}.
{jump,{f,51}}.
{label,50}.
{move,{atom,false},{x,7}}.
{label,51}.
{bif,'=:=',{f,52},[{x,7},{atom,true}],{x,7}}.
{test,is_eq_exact,{f,52},[{x,6},{atom,true}]}.
{test,is_eq_exact,{f,52},[{x,7},{atom,true}]}.
{move,{x,5},{x,0}}.
return.
{label,52}.
return.
We can add optimizations of guard tests to v3_kernel to achive a better result:
{function, skip_prefix, 2, 49}.
{label,48}.
{line,[{location,"filename.erl",187}]}.
{func_info,{atom,filename},{atom,skip_prefix},2}.
{label,49}.
{test,is_ne_exact,{f,51},[{x,1},{atom,false}]}.
{test,is_nonempty_list,{f,51},[{x,0}]}.
{get_list,{x,0},{x,2},{x,3}}.
{test,is_nonempty_list,{f,51},[{x,3}]}.
{get_list,{x,3},{x,4},{x,5}}.
{test,is_eq_exact,{f,51},[{x,1},{x,4}]}.
{test,is_ge,{f,51},[{x,2},{integer,65}]}.
{test,is_lt,{f,50},[{integer,90},{x,2}]}.
{test,is_ge,{f,51},[{x,2},{integer,97}]}.
{test,is_ge,{f,51},[{integer,122},{x,2}]}.
{label,50}.
{move,{x,5},{x,0}}.
return.
{label,51}.
return.
Looking at the STDLIB application, there were 112 lines of BIF calls in guards
that beam_bool failed to convert to test instructions. This commit eliminates
all those BIF calls.
Here is how I counted the instructions:
$ PATH=$ERL_TOP/bin:$PATH erlc -I ../include -I ../../kernel/include -S *.erl
$ grep "bif,'[=<>]" *.S | grep -v f,0
dets.S: {bif,'=:=',{f,547},[{x,4},{atom,read_write}],{x,4}}.
dets.S: {bif,'=:=',{f,547},[{x,5},{atom,saved}],{x,5}}.
dets.S: {bif,'=:=',{f,589},[{x,5},{atom,read}],{x,5}}.
.
.
.
$ grep "bif,'[=<>]" *.S | grep -v f,0 | wc
112 224 6765
$
|
|
At this stage in match compilation we are allowed to change the
alignment of arguments and constructors as long as they are the
same aligned in the same group.
|
|
* josevalim/compiler/at-var/PR-1081/OTP-13924:
Use @ in variable names generated by core and kernel
|
|
The previous variable names can be generated by
projects like LFE and Elixir, leading to possible
conflicts. Our first to choice to solve such conflicts
was to use $ but that's not a valid variable name in core.
Therefore we picked @ which is currently supported and
still reduces the chance of conflicts.
|
|
Do a simpler translation of internal BIFs.
While we are it, also remove the dummy values of Index and Uniq
from the make_fun internal operation.
|
|
Remove the special handling #k_try{} in guards in v3_life. If we
introduce a new #k_protected{} record in v3_kernel, v3_life no longer
needs to know whether it is processing guards or bodies.
|
|
* kostis/compiler/v3_kernel/eta-conversion:
v3_kernel: Fix typo in comment
|
|
|
|
'callback' and 'optional_callbacks' are no longer wild attributes.
|
|
Use cerl:make_list/1 instead of a home-made make_list/1 to ensure that
literal lists are constructed as literals. In a future release, we
would like to forbid in the loader construction of literal lists using
instructions like:
put_list {atom,a} [] Dst
The proper way is:
move {literal,[a]} {x,0}
Also update the comment about "put_list Const [] Dst" in ops.tab.
|
|
|
|
Using a map to store the number of free variables for funs instead of
an orddict will speed up the v3_kernel pass for modules with a huge
number of funs (such as NBAP-PDU-Contents in the asn1 test suite).
|
|
|
|
|
|
Profiling shows that subst_vsub/3 dominates the running time. It
is therefore worthwhile optimizing it.
|
|
If we have a sequence of put_map_* instructions operating on the
same map, it will be more efficient if we can have one is_map/2
instruction before put_map_* instructions, so that each put_map_*
does not need to test whether the argument is a map.
|
|
Essentially, core_lib:literal_value/1 became useless when literals
were introduced in R12. Since we always create #c_literal{} records
whenever possible, literal_value/1 would *only* succeed when it was
passed a #c_literal{} argument.
|
|
When translating a function with map construction:
f(A) ->
B = b,
C = c,
#{A=>1,B=>2,C=>3}.
v3_core would break apart the map construction into three
parts because of the way the map instructions in BEAM work --
variable keys need to be in their own instruction.
In the example, constant propagation will turn two of the
keys to literal keys. But the initial breaking apart will
not be undone, so there will still be three map constructions:
'f'/1 =
fun (_cor0) ->
let <_cor3> = ~{::<_cor0,1>}~
in let <_cor4> = ~{::<'b',2>|_cor3}~
in ~{::<'c',3>|_cor4}~
It would be possible to complicate the sys_core_fold pass
to regroup map operations so that we would get:
'f'/1 =
fun (_cor0) ->
let <_cor3> = ~{::<_cor0,1>}~
in ~{::<'b',2>,::<'c',3>|_cor3}~
A simpler way that allows to simplify the translation is
to skip the grouping in v3_core and translate the function
to:
'f'/1 =
fun (_cor0) ->
~{::<_cor0,1>,::<'b',2>,::<'c',3>}~
We will then let v3_kernel do the grouping while translating
from Core Erlang to Kernel Erlang.
|
|
|
|
No need to check for variables in Map keys.
|
|
Not only variables are allowed as arguments, the name should reflect that.
Change cerl Map argument interface
* cerl:map_arg/1 is more suitable then cerl:map_val/1 in this case.
|
|
Reject all expressions that are known to fail.
Emit 'badarg' for those expressions.
Ex.
[]#{ a => 1}
Is not a valid map update expression.
|
|
The Kernel instructions were not properly ordered when compiling maps with
complex values mixed in assoc and exact pairs.
Reported-by: Ulf Norell
|
|
* Combine multiple get values with one instruction
* Combine multiple check keys with one instruction
|
|
* egil/compiler/maps-fix-codegen:
compiler: Fix codegen multiple updates for Maps
erts,compiler: Correct and amend tests for Maps
|
|
This fixes an error on multiple updates optimization for map pairs.
The error was introduced with moving to term order in Maps.
This also fixes an error where register life time was lost for values
and could result in erroneuos values being emitted in for map pairs.
Simplified v3_codegen by moving multiple update optimizations to v3_kernel.
|
|
* hsv/using_lists_droplast:
lib/mnesia/test/ - Replace reverse(tl(reverse(L))) with lists:droplast/1
lib/ssh - Replace reverse(tl(reverse(L))) with lists:droplast/1
lib/wx - Replace reverse(tl(reverse(L))) with lists:droplast/1
Use lists:droplast/1 in orber/orber_interceptors.erl
Import and use lists:droplast/1 in v3_core/v3_kernel
OTP-11678
OTP-11677
|
|
Simplify compiler internals and parsing of core format.
|
|
Simplify compiler internals for kernel passes.
|
|
|
|
|
