| Age | Commit message (Collapse) | Author |
|---|
|
|
|
* Combine multiple get values with one instruction
* Combine multiple check keys with one instruction
|
|
The instruction get_map_element has a faillabel so you may not
use the instruction within a allocate/deallocate block.
|
|
|
|
To make it possible to build the entire OTP system, also define
dummys for the instructions in ops.tab.
|
|
Any init instruction following an allocate is put in the Inits list of the
corresponding alloc tuple.
|
|
If an allocate_zero instruction is fed to beam_block and the beam_type
pass is not used afterwards (e.g. with erlc +no_topt), the 'no_opt' atom
will be rejected by beam_flatten.
|
|
The compiler shouldn't crash when fed an already-optimised BEAM assembly file.
|
|
|
|
Eliminate some code bloat.
|
|
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.
|
|
Since we always want to remove unused labels directly after code
generation (whether we'll run the optimization passes or not),
we can simplify the code by doing it in beam_a.
|
|
In modules with huge functions with many bs_match_string
instructions, we can speed up the compilation by combining
adjacent bs_match_strings instruction in v3_codegen (as opposed
to in beam_block where we used to do it).
For instance, on my computer the v3_codegen became more than
twice as fast when compiling the re_testoutput1_split_test module
in the STDLIB test suites.
|
|
|
|
|
|
Moving of allocation instructions upwards in the instruction
stream (in order to enable further optimizations) in beam_block,
is implemented with the assumption that if a register {x,X}
contains a valid term, then all other x register with lower
numbers than X also contain valid terms. That assumption is
true after code generation.
The beam_utils:live_opt/1 optimization, however, may invalidate
that assumption. For instance, if a receive statement exports a
variable that is used, but the return value of the receive statement
is not used, then {x,1} but not {x,0} contains a valid term at the
end of the receive statement. If the receive statement is
followed by
{bif,self,{f,0},[],{x,0}}.
{test_heap,NumberOfWords,2}.
moving the allocation upwards will produce
{test_heap,NumberOfWords,2}.
{bif,self,{f,0},[],{x,0}}.
which will cause the beam_validator pass to scream loudly that
{x,0} is not live at the test_heap instruction.
Fix the problem by doing the optimizations in reverse order.
Reported-by: Jim Engquist
|
|
Since the introduction of literals in R12B, empty tuples
are literals. Thus the put_tuple/2 instruction is always
followed by at least one put/1 instruction. Therefore
the alloc_may_pass/1 function in beam_block no longer needs
a clause for for "put_tuple", because the clause for "put"
will always be reached first (since the instruction stream
is scanned in reverse execution order).
Note that if the compiler would generate a "put_tuple 0 Dst"
instruction for some unfathomable reason, we should still be
because the run-time system will now refuse to load a module
containing such an instruction.
|
|
* bg/compiler-remove-r11-support:
compiler: Don't support the no_binaries option
erts: Don't support the put_string/3 instruction
compiler: Don't support the no_constant_pool option
compiler: Don't support the r11 option
test_server: Don't support communication with R11 nodes
binary_SUITE: Don't test bit-level binary roundtrips with R11 nodes
erts: Test compatibility of funs with R12 instead of R11
OTP-8531 bg/compiler-remove-r11-support
|
|
The no_constant_pool option was implied by the r11 option. It turns
off the usage of the constant (literal) pool, so that BEAM
instructions that use constants can be loaded in an R11 system.
Since the r11 option has been removed, there is no need to
retain the no_constant_pool option.
|
|
|
