| Age | Commit message (Collapse) | Author |
|---|
|
Using maps and cerl_sets instead of dict and sets will slightly
speed up the beam_clean pass for modules with many functions
and/or calls to local functions.
|
|
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.)
|
|
The removal of redundant bs_restore2 instructions is done easier
on the SSA format. Keep the rest of the optimizations, because
they are easier to do on the BEAM instructions.
|
|
|
|
During compilation, the bs_save2 and bs_restore2 instructions contain a match
context reference. That reference is the variable name that holds the match context.
beam_clean assumes that the reference always is an atom, which is not a safe assumption
since integers are legal variable names in Core Erlang.
|
|
1a029efd1ad47f started to run the beam_block pass a second time,
but it did not attempt to combine adjacent blocks.
Combining adjacent blocks leads to many more opportunities for
optimizations.
After doing some diffing in generated code, it turns out that
there is no benefit for beam_split to split out line instructions
from blocks. It seems that the only reason it was done was to
slightly simplify the implementation of the no_line_info option
in beam_clean.
|
|
|
|
|
|
|
|
There is an optimization in beam_clean that will remove values
having the same label as the failure label in a select_val
instruction. Conceptually, this optimization is in the wrong
module since ideally beam_clean is a mandatory pass that should
not do optimizations. Furthermore, this part of beam_clean is
called three times (from beam_dead, beam_peep, and as a compiler
pass from the 'compile' module), but it only does useful one of
the times it is called.
Therefore, move this optimization to the beam_peep pass.
The same optimization is done in beam_dead, but unfortunately it
misses some opportunities for optimization because the code sharing
optimization in beam_jump (share/1) runs after beam_dead. It would
be more satisfactory to have this optimization only in beam_dead,
but it turned out not to be trivial. If we try to run
beam_jump:share/1 before beam_dead, some optimizations will no
longer work in beam_dead because fallthroughs have been eliminated.
For the moment, the possible solutions to this problem seems to
involve more work and more complicated code than the gain from
eliminating the duplicated optimization would gain.
|
|
In 45f469ca0890, the BEAM loader started to use x(1023) as scratch
register for some instructions. Therefore we should not allow
x(1023) to be used in code emitted by the compiler.
|
|
|
|
The actual bs_match_string instruction has four operands:
bs_match_string {f,Lbl} Ctxt NumBits {string,ListOfBytes}
However, v3_codegen emits a more compact representation where
the bits to match are packaged in a bitstring:
bs_match_string {f,Lbl} Ctxt Bitstring
Currently, beam_clean:clean_labels/1 will rewrite the compact
representation to the final representation. That is unfortunate
since clean_labels/1 is called by beam_dead, which means that
the less compact representation will be introduced long before
it is actually needed by beam_asm. It will also complicate any
optimizations that we might want to do.
Move the rewriting of bs_match_string from beam_clean:clean_labels/1
to the beam_z pass, which is the last pass executed before
beam_validator and beam_asm.
|
|
Commits b44f86b7 and 97953704 introduced translations of instructions
in beam_a and beam_z, but forgot to remove the handling of them in
beam_clean.
|
|
* Combine multiple get values with one instruction
* Combine multiple check keys with one instruction
|
|
|
|
To make it possible to build the entire OTP system, also define
dummys for the instructions in ops.tab.
|
|
files as delimiters.
While working on a tool that processes Erlang code and testing it against this repo,
I found out about those little sneaky 0xff. I thought it may be of help to other
people build such tools to remove non-conforming-to-standard characters.
|
|
|
|
Somewhat reduce the code bloat by eliminating special cases.
|
|
Eliminate some code bloat.
|
|
Rewrite the five binary creation instructions to a bs_init
instruction, in order to somewhat reduce code bloat.
|
|
We can remove some code bloat by handling the special instructions
as BIF instructions in the optimization passes. Also note that
bs_utf*_size was not handled by beam_utils:check_liveness/3
(meaning the conservative answer instead of the correct answer
would be returned).
|
|
Seven bs_put_* instructions can be combined into one generic bs_put
instruction to avoid some code bloat. That will also improve some
optimizations (such as beam_trim) that did not handle all bs_put*
variants.
|
|
Also update the r12 and r13 options so that they imply no_line_info.
|
|
|
