aboutsummaryrefslogtreecommitdiffstats
AgeCommit message (Collapse)Author
2015-04-13erl_term.h: Add is_not_map() macroBjörn Gustavsson
For consistency with other data types, add the is_not_map() macro.
2015-04-13Tigthen code for the i_get_map_elements/3 instructionBjörn Gustavsson
2015-04-13Pre-compute hash values for the general get_map_elements instructionBjörn Gustavsson
See the previous commit for justification and use cases.
2015-04-13Teach the loader to pre-compute the hash value for single-key lookupsBjörn Gustavsson
Let the loader pre-compute the hash value when a single, literal key is matched as in: #{<<"some_key">>:=V} = Map In my measurements, this optimization resulted in a 30 percent speedup for short binary keys. Unfortunately, this optimizization makes no difference for small maps with less than 32 keys, since the hash value is not used. Still, there are the following use cases: * A map used instead of a record with more than 32 entries. I have seen some applications with huge records. * Lookup in JSON dictionaries represented as maps. The hash value will only be used when the map is a hash map (currently, that means at least 32 entries).
2015-04-13Optimize use of i_get_map_element/4Björn Gustavsson
In the i_get_map_element/4 instruction, for literal keys other than atoms, the key would be put into x[0] instead of used directly in the instruction. The reason is that the original implementation of maps only supported atom keys.
2015-04-13beam_emu: Slightly optimize update_map_{assoc,exact}Björn Gustavsson
In the update loop for big maps, the E variable is restored for each turn of the loop. It only needs to be restored if a garbage collection has been performed. Also add a new test case that attempts to force several garbage collections while updating a map, to help us find bugs with incorrect restoration of the E variable after a garbage collection.
2015-04-13v3_codegen: Don't sort map keys in map creation/updateBjörn Gustavsson
It is no longer necessary to sort the keys, since the loader does the sorting.
2015-04-13beam_validator: No longer require strict literal term orderBjörn Gustavsson
The BEAM loader will now sort keys for maps during loading, so beam_validator should not require the keys to be ordered any order. However, we must still ensure that literals keys are unique (which was implicitly guaranteed by the strict ordering requirement).
2015-04-13Sort maps keys in the loaderBjörn Gustavsson
The map instructions require that the keys in the instructions are sorted (for flatmaps). But that is an implementation detail that should not exposed outside of the BEAM virtual machine. Therefore, make the sorting of the keys the responsibility of the loader and not the compiler. Also note that the sort order for maps with numeric keys or keys with numeric components has changed in OTP 18. That means that code compiled for OTP 17 that operated on maps with map keys might not work in OTP 18 without the sorting in the loader (although it is unlikely to be an issue in practice).
2015-04-13De-optimize the has_map_fields instructionsBjörn Gustavsson
The has_map_fields instruction is infrequently used. Thus there is no need to have the fastest possible implementation; it is better to have an implementation that reduces the code size in the already big process_main() function. We can transform has_map_fields to a get_map_elements instruction, targeting the same unused x[0] register for all keys. That instruction will only be marginally slower than existing implementation.
2015-04-13erts/map_SUITE.erl: Add a test case that tests has_map_fieldsBjörn Gustavsson
The has_map_fields instruction was not tested at all by erts/map_SUITE.erl
2015-04-13Fully evaluate is_map/1 for literals at load-timeBjörn Gustavsson
The compiler will only emit is_map/1 instructions with literal argument if optimization is turned off. Therefore, the only reason for this commit is cleanliness.
2015-04-13map_SUITE: Add tests of is_map/1 with literal mapsBjörn Gustavsson
To be sure that the compiler and BEAM virtual machine correctly handles literals maps, we must test it.
2015-04-13Run a clone of map_SUITE without optimizationsBjörn Gustavsson
Create a clone of map_SUITE named map_no_opt_SUITE to ensure that the loader can cope with unoptimized map instructions.
2015-04-13Remove the fail label operand of the new_map instructionBjörn Gustavsson
The new_map instruction cannot fail, and thus needs no fail label.
2015-04-13Correct transformation of put_map_assoc to new_mapBjörn Gustavsson
A put_map_assoc instruction with an empty source map should be converted to a simpler new_map instruction. The transformation didn't happen because an empty source map is no longer represented as a NIL term (as it was in the beginning before map literals were implemented).
2015-04-13Remove support for put_map_exact without a source mapBjörn Gustavsson
Using the exact operator (':=') is only allowed when an existing map is being updated. Thus the following causes a compilation error: #{k:=v} Therefore there is no need to support the put_map_exact instruction without a source map.
2015-04-10Merge branch 'egil/maps-test-coverage'Björn-Egil Dahlberg
* egil/maps-test-coverage: erts: Remove code that was commented out erts: Cover maps:values/1 for large maps erts: Test maps:from_list/1 shrinking
2015-04-10Merge branch 'egil/fix-maps-deep-colliding-merge'Björn-Egil Dahlberg
* egil/fix-maps-deep-colliding-merge: erts: Fix deep colliding hash values in maps:from_list/1
2015-04-10Merge branch 'egil/fix-hash-float-zero/OTP-12641'Björn-Egil Dahlberg
* egil/fix-hash-float-zero/OTP-12641: erts: Ensure hashing of zero is consistent
2015-04-10erts: Ensure hashing of zero is consistentBjörn-Egil Dahlberg
Erlang treats positive and negative zero as equal, meaning, true = 0.0 =:= 0.0/-1 However, Erlangs hash functions: hash, phash and phash2 did not reflect this behaviour. Meaning, the hash values produced by the different hash functions would not be identical for positive and negative zero. This commit ensures that hash value of positive zero is always produced regardless of the signedness of the zero float, i.e. true = erlang:phash2(0.0) =:= erlang:phash2(0.0/-1)
2015-04-10erts: Remove code that was commented outBjörn-Egil Dahlberg
2015-04-10Merge branch 'maint-17'Zandra Hird
Conflicts: OTP_VERSION
2015-04-10Updated OTP versionOTP-17.5.1Erlang/OTP
2015-04-10Update release notesErlang/OTP
2015-04-10Merge branch 'ia/ssh/packet_size0/OTP-12645' into maint-17Erlang/OTP
* ia/ssh/packet_size0/OTP-12645: ssh: Reset upgrade instructions to default ssh: Change version (vsn.mk) ssh: Option minimal_remote_max_packet_size and test cases ssh: fix loop bug for zero maximum_packet_size. ssh: Add ssh_info:print/1 to write onto a file descriptor ssh: Add {active, false} to ssh listen socket ssh: Change send_buf implementation from list to queue ssh: Remove error report
2015-04-10Merge branch 'bjorn/stdlib/cuddle-with-tests'Björn Gustavsson
* bjorn/stdlib/cuddle-with-tests: Eliminate deprecated now/0 used as timestamp Eliminate use of now/0 for creating a unique filename Eliminate use of deprecated now/0 for measuring time Eliminate use of deprecated now/0 for random number generation Speed up timer_simple_SUITE:timer_perf/1 Speed up timer_SUITE Optimize gen_server_SUITE:hibernate/1 Optimize gen_event_SUITE:hibernate/1 Optimize gen_fsm_SUITE:hibernate/1 Optimize unicode_SUITE:ex_binaries_errors* test cases Optimize unicode_SUITE:binaries_errors/1 binary_module_SUITE: Remove unnecessary calls to the binref module Optimize binary_module:random_ref_comp/1 Optimize io_proto_SUITE:unicode_options_gen/1
2015-04-10Eliminate deprecated now/0 used as timestampBjörn Gustavsson
Rewrite the code to measure and print elapsed time instead of just printing timestamps for the start and stop of each operation.
2015-04-10Eliminate use of now/0 for creating a unique filenameBjörn Gustavsson
Use erlang:unique_integer/1 instead.
2015-04-10Eliminate use of deprecated now/0 for measuring timeBjörn Gustavsson
Use erlang:monotonic_time/1 instead of now/0.
2015-04-10Eliminate use of deprecated now/0 for random number generationBjörn Gustavsson
Use erlang:timestamp/0 instead now/0 when seeding the random number generator.
2015-04-10Speed up timer_simple_SUITE:timer_perf/1Björn Gustavsson
In big_test/1, there is 9 seconds sleep before spawning any timers. Why? Before the sleep nrev processes are started, but they are likely to finish running before the 9 seconds have elapsed. Since I see no reason at all for the sleep, I will remove it.
2015-04-10Speed up timer_SUITEBjörn Gustavsson
The single test case in timer_SUITE is annoyingly slow. On average, its running time is about 4 minutes (estimated by a back-of-the-envelope calculation). Scale down the delay times in the main loop and the length of timers started. The running average time should now be around 80 seconds.
2015-04-10Optimize gen_server_SUITE:hibernate/1Björn Gustavsson
The running time for gen_server_SUITE:hibernate/1 is about 12 seconds because of 1000 ms sleeps in the test case. Introduce a helper function to avoid sleeping more than necessary. While we are it, also remove all ?line macros in the test case.
2015-04-10Optimize gen_event_SUITE:hibernate/1Björn Gustavsson
The running time for gen_event_SUITE:hibernate/1 is about 20 seconds because of 1000 ms sleeps in the test case. Introduce helper functions to avoid sleeping more than necessary. While we are it, also remove all ?line macros in the test case.
2015-04-10Optimize gen_fsm_SUITE:hibernate/1Björn Gustavsson
The running time for gen_fsm_SUITE:hibernate/1 is almost 30 seconds because of 1000 ms sleeps in the test case. Introduce helper functions to avoid sleeping more than necessary. While we are it, also remove all ?line macros in the test case.
2015-04-10Optimize unicode_SUITE:ex_binaries_errors* test casesBjörn Gustavsson
Refactor the code so that the error test cases for utf16 can share the code for big and little endian. Do the same for utf32. The major optimization is the observation that the following code: byte_size(unicode:characters_to_binary(List, unicode, {utf16,big}) can be replaced with: 2*length(List) which is much faster. (That optimization is not safe in general, but for the lists used in the error tests cases, it is.)
2015-04-10Optimize unicode_SUITE:binaries_errors/1Björn Gustavsson
unicode_SUITE:binaries_errors/1 is the slowest test case in unicode_SUITE. Its running time is about 50 seconds on my computer. Break apart unicode_SUITE:binaries_errors/1 into several test cases that can run in parallel.
2015-04-10binary_module_SUITE: Remove unnecessary calls to the binref moduleBjörn Gustavsson
The compile_pattern/1 function in the binref module is a dummy function. Therefore, calling binref:match/2 once with the plain pattern and once with a "pre-compiled" pattern will take twice as long, but will not help us find more errors. I shaved off one fourth of the running time for random_ref_comp/1 by eliminating the unnecessary calls to the binref module.
2015-04-09Merge branch 'sverk/maps-bin2term-eqhash-bug/12585'Sverker Eriksson
* sverk/maps-bin2term-eqhash-bug/12585: erts: Fix bug in map_from_list when keys clash in both value and hash erts: Fix bug in binary_to_term for big maps with 32 bit hash-clash
2015-04-09erts: Cover maps:values/1 for large mapsBjörn-Egil Dahlberg
2015-04-09erts: Test maps:from_list/1 shrinkingBjörn-Egil Dahlberg
Repeated keys will shrink map to a flatmap if the number of pairs drops below the limit.
2015-04-09Merge branch 'ia/ssl/improve_docs'Ingela Anderton Andin
* ia/ssl/improve_docs: ssl: Document enhancements ssl: Align with alphabetical order ssl: Align with editorial changes ssl: Correct merge error ssl: Add links ssl: Correct typos ssl: Correct XML-tag ssl: Technically clarify description ssl: Editorial updates
2015-04-09ssl: Document enhancementsIngela Anderton Andin
2015-04-09ssl: Align with alphabetical orderIngela Anderton Andin
2015-04-09ssl: Align with editorial changesIngela Anderton Andin
2015-04-09ssl: Correct merge errorIngela Anderton Andin
2015-04-09ssl: Add linksIngela Anderton Andin
2015-04-09ssl: Correct typosIngela Anderton Andin
2015-04-09ssl: Correct XML-tagIngela Anderton Andin