Refactor ECMA builtin template

[akosthekiss]

Avoid sorting the array of property magic string IDs and make it
const, thus ensuring that it gets placed in .rodata. Replace the
binary search in the array (which would not work anymore because
of unsortendness) with function that returns the index of the
looked-after magic string ID using a switch-based logic (we rely
on compiler optimization to generate optimal code from that
switch).

Extras:

• Getting rid of the superfluous macro argument from routine names.
• Fixing the list of undef'd macros
• Fixing builtin-related comments

JerryScript-DCO-1.0-Signed-off-by: Akos Kiss akiss@inf.u-szeged.hu

[akosthekiss]

The effect on text (=.text+.rodata) and data (=.data) is considerable in release builds (up to cca. 2.5K gain in total), size of r/w data drops below 100 bytes. Debug builds show strange results; something is still left in .data. Details below.

filename	master/text	master/data	pr905/text	pr905/data	gain/text	gain/data
release.armv7lhf_linux/jerry	197295	1096	195958	52	1337	1044
release.armv7lhf_linux-cp/jerry	143284	812	142156	52	1128	760
release.armv7lhf_linux-cp_minimal/jerry	88091	192	87973	0	118	192
release.armv7lhf_linux-mem_stats/jerry	197855	1096	196554	52	1301	1044
release.armv7lhf_linux-mem_stress_test/jerry	197231	1096	195590	52	1641	1044
release.mcu_stm32f3-cp.bin/jerry	144356	260	142748	72	1608	188
release.mcu_stm32f3-cp_minimal.bin/jerry	82704	68	82436	20	268	48
release.mcu_stm32f4-cp.bin/jerry	144356	260	142748	72	1608	188
release.mcu_stm32f4-cp_minimal.bin/jerry	82704	68	82436	20	268	48
release.x64_linux/jerry	279890	1373	278590	61	1300	1312
release.x64_linux-cp/jerry	211932	989	210766	61	1166	928
release.x64_linux-cp_minimal/jerry	133068	204	132993	0	75	204
release.x64_linux-mem_stats/jerry	280923	1373	280071	61	852	1312
release.x64_linux-mem_stress_test/jerry	279917	1373	279031	61	886	1312
debug.armv7lhf_linux/jerry	732048	3272	731864	2228	184	1044
debug.armv7lhf_linux-cp/jerry	536720	2988	535624	2228	1096	760
debug.armv7lhf_linux-cp_minimal/jerry	366772	2384	366120	2192	652	192
debug.armv7lhf_linux-mem_stats/jerry	735596	3272	735364	2228	232	1044
debug.armv7lhf_linux-mem_stress_test/jerry	732096	3272	731912	2228	184	1044
debug.mcu_stm32f4-cp.bin/jerry	547060	2484	545060	2244	2000	240
debug.mcu_stm32f4-cp_minimal.bin/jerry	365828	2272	365244	2212	584	60
debug.x64_linux/jerry	808414	5780	812955	4468	-4541	1312
debug.x64_linux-cp/jerry	628393	5492	630807	4468	-2414	1024
debug.x64_linux-cp_minimal/jerry	438533	4688	439054	4432	-521	256
debug.x64_linux-mem_stats/jerry	813019	5780	816760	4468	-3741	1312
debug.x64_linux-mem_stress_test/jerry	808434	5780	812975	4468	-4541	1312

[LaszloLango]

Missing comment for this argument

[LaszloLango]

Could you measure the performance on RPi2?

[zherczeg]

If there is no big perf regression, LGTM from my side.

[akosthekiss]

@LaszloLango I've updated the patch. Does it look better?

[LaszloLango]

@akiss77, yes. The patch looks good. I agree with @zherczeg. If there is no big perf regression then we can merge the PR.

[bzsolt]

RPi2 results
master (aa0b621)

                  Benchmark |                  Perf(+ is better)
                 3d-cube.js |  2.622s ->   2.640s :  -0.699%  (+-0.735%) : [≈]
     access-binary-trees.js |  1.407s ->   1.412s :  -0.355%  (+-1.301%) : [≈]
         access-fannkuch.js |  7.983s ->   7.995s :  -0.146%  (+-0.421%) : [≈]
            access-nbody.js |  3.102s ->   3.098s :  +0.108%  (+-0.939%) : [≈]
bitops-3bit-bits-in-byte.js |  1.733s ->   1.735s :  -0.096%  (+-0.715%) : [≈]
     bitops-bits-in-byte.js |  2.660s ->   2.655s :  +0.188%  (+-0.757%) : [≈]
      bitops-bitwise-and.js |  3.645s ->   3.685s :  -1.097%  (+-0.835%) : [-]
   controlflow-recursive.js |  0.890s ->   0.890s :  +0.000%  (+-0.000%) : [≈]
              crypto-aes.js |  4.640s ->   4.718s :  -1.688%  (+-0.654%) : [-]
              crypto-md5.js | 21.462s ->  21.540s :  -0.365%  (+-0.756%) : [≈]
             crypto-sha1.js |  9.752s ->   9.715s :  +0.376%  (+-0.649%) : [≈]
       date-format-tofte.js |  2.215s ->   2.230s :  -0.677%  (+-0.781%) : [≈]
       date-format-xparb.js |  1.065s ->   1.087s :  -2.035%  (+-1.176%) : [-]
             math-cordic.js |  3.010s ->   3.000s :  +0.332%  (+-0.597%) : [≈]
       math-partial-sums.js |  1.667s ->   1.660s :  +0.400%  (+-0.508%) : [≈]
      math-spectral-norm.js |  1.485s ->   1.513s :  -1.908%  (+-1.096%) : [-]
            string-fasta.js |  4.567s ->   4.565s :  +0.037%  (+-1.171%) : [≈]
            Geometric mean: |               Speed up: -0.446% (+-0.200%) : [-]

[LaszloLango]

@bzsolt, thanks for the measurement. @akiss77, you may land the patch.

[zherczeg]

Just noise. LGTM