Comparison of compilers for development on microcontrollers with the ARM Cortex-M core
In this article, we test the 3rd compiler for Kinetis microcontrollers with the ARM Cortex-M4 core .
Run the tests CoreMark , Whetstone , Dhrystone .
We study compression algorithms with minimal RAM consumption and find out how different compilers affect their performance.
And we ’ll even try to find out how far Kinetis is behind in performance from Intel Core I7 .
Previous articles on development on Kinetis microcontrollers:
- Start of development on Kinetis microcontrollers. Fast start
- Kinetis Microcontrollers from NXP-Freescale for Embedded Electronics
- Open project universal microcontroller module
- Smart board for managing 3-phase power loads
In the article “Start of development on Kinetis microcontrollers. Quick Start ” was described the development environment Kinetis Design Studio v3.0.0 (KDS) . Wednesday KDS uses as a tool to compile and build the software package the GNU for the Tools on ARM Embedded Processors . This includes the GCC compiler for ARM in particular . The compiler is not bad in every sense. But if we look at how many branches he has in the gcc.gnu.org/svn/gcc/branches directory , then we will see more than 300 of them! How much effort is needed to support such a huge number of branches !? But the project is not commercial. On the other hand, the ARM Cortex-M seriesnot mainstream, not used on smartphones, they have not attracted so much attention from the community that supports Open Source.
This all makes you think, especially if you intend to get involved in the long and exhausting programming of microcontrollers with megabytes of impenetrable sources, limited memory space, unknown requirements but with a lightning reaction to their changes.
Fortunately, KDS is provided with plugins for converting to other well-known compilation and assembly packages - Keil and IAR . Most of the demo projects for Kinetis are accompanied by 3 configurations for building tools GCC , Keil and IAR . However in the software suite forKinetis is not a comprehensive test enough to compare these compilation tools with each other. Therefore, I had to create my own.
The project was created in KDS using the ProcessorExpert plug-in for the MK60FN1M0VLQ12 microcontroller on the board presented in the article “Smart board for controlling 3-phase power loads”
Here is the project screen in KDS :
What is included in the test suite
Whetstone project . A very old and simple floating point performance test. The test is very easy to port and can be performed on simple microcontrollers. In our case, two implementations are presented: a calculation option with a float typeand a variant with a double type. This is important because in the first case, the Cortex-M4 uses a coprocessor, while in the second it does not.
Dhrystone project . Performance test of integer operations. The float and double typesare not used anywhere. The test is also easy to port and there is a huge archive of results for all kinds of microcontrollers and computers.
CoreMark Project. A modern test specifically designed for embedded systems. May become an industry standard. Its results are certified by EEMBC , and therefore protected to some extent from cheating and falsification.
The results of the CoreMark test are the most credible in the comprehensive testing of microcontroller performance and compiler comparison. On it, in principle, it would be possible to limit it, but the tests themselves do not have any practical benefit for embedded applications, and to give more practical value to these tests, I added more testing of compression algorithms.
First I had to find a suitable memory manager (heap manager or heap).
Umm_malloc project . Found on Github. In appearance it seemed quite suitable. The standard heap manager from the C library, and I refused to use for two reasons. Firstly, each of the Keil , IAR , GCC environments will have its own heap library and did not want to spend time researching the features of each of them. Secondly, compression algorithms are critically dependent on the amount of memory allocated to them. A memory barely enough. We needed a mechanism for precise control of allocated memory. And here umm_malloc provides such an opportunity. And along the way, we will test the performance of the umm_malloc algorithm itself .
Compression algorithms
- LZSS Algorithm
- Zlib algorithm
- S-LZW algorithm. Found somewhere on the Internet. But I was strongly refactored.
- FastLZ Algorithm
All the algorithms had to be adjusted in order to use as little RAM as possible.
How much memory these algorithms need, how much they can compress the data, and what their performance can be found in the test results and in the listing of output to the terminal.
And here are the results:
-All compilers had the option of maximum optimization enabled. The exact composition of the options can be found in the attached project.
-Compression algorithms compressed a section of the binary firmware code obtained from the project of this test compiled in the IAR.
-GCC compile time measured when compiling from KDS.
-In addition to the speed of the algorithms, lines are also inserted in the table with additional metrics that are important for the developer, these are: the time the project is completely recompiled, the size of the binary file when optimizing for speed, and the size of the binary file when optimizing for size.
Output to the terminal of the test results compiled in IAR
----------- Speed Test -------------
Reference time = 100001 us
----------- umm malloc test -------------
...........................................................
|0x1fff1b44|B 0|NB 1|PB 0|Z 1|NF 1|PF 0|USED
|0x1fff1b4c|B 1|NB 4223|PB 0|Z 4222|NF 4223|PF 0|FREE
|0x1fff9f3c|B 4223|NB 0|PB 1|Z 1|NF 0|PF 1|USED
...........................................................
Total Entries 1 Used Entries 0 Free Entries 1
Total Blocks 4223 Used Blocks 0 Free Blocks 4223
Used space = 0 Free space = 33784
...........................................................
umm malloc test time = 1447071 uS
Allocations = 40000, uSec per allocation = 36
----------- Compressor LZSS test -------------
Uncompressed size = 2048
Compressed size = 1858. Max. alloc.= 14384. Time = 6627 uS
Decompress Max. alloc.= 1048. Time = 981 uS
Decompress Ok!
----------- Compressor Zlib test -------------
Uncompressed size = 2048
Compressed size = 1740. Max. alloc.= 24288. Time = 12179 uS
Decompress Max. alloc.= 15768. Time = 764 uS
Decompress Ok!
----------- Compressor S-LZW test -------------
Uncompressed size = 2048
Compressed size = 2035. Max. alloc.= 4320. Time = 2542 uS
Decompress Max. alloc.= 4320. Time = 3816 uS
Decompress Ok!
----------- Compressor FastLZ test -------------
Uncompressed size = 2048
Compressed size = 1834. Max. alloc.= 32776. Time = 1010 uS
Decompress Max. alloc.= 0. Time = 152 uS
Decompress Ok!
----------- WHETSTONE FLOAT -------------
.........MODULE 1: simple identifiers...
xx1 = -0.06679254770278930664
xx2 = -0.46633863449096679688
xx3 = -0.73303699493408203125
xx4 = -1.13254797458648681641
.........MODULE 2: array elements.......
e10 = -0.06834230571985244751
e11 = -0.46263590455055236816
e12 = -0.72971796989440917969
e13 = -1.12397670745849609375
.........MODULE 3: array as parameter...
e1_0 = -0.05533060804009437561
e1_1 = -0.44743216037750244141
e1_2 = -0.71096724271774291992
e1_3 = -1.10309338569641113281
.........MODULE 4: conditional jumps....
val = 1
.........MODULE 6: integer arithmetic...
e1k = 6.00000000000000000000
e1l = 6.00000000000000000000
.........MODULE 7: trig. functions......
x1 = 0.49041154980659484863
y1 = 0.49039667844772338867
.........MODULE 8: procedure calls......
z = 0.99993747472763061523
.........MODULE9: array references......
e1_j = -1.10309338569641113281
e1_k = 3.00000000000000000000
e1_l = -1.10309338569641113281
.........MODULE10: integer arithmetic...
j = 2
k = 3
.........MODULE11: standard functions...
x = 0.83466047048568725586
........................................
Whetstone last time (uS) = 10960.00
Whetstone max time (uS) = 10961.00
Whetstone min time (uS) = 10960.00
----------- WHETSTONE DOUBLE -------------
.........MODULE 1: simple identifiers...
xx1 = -0.06679268039452399826
xx2 = -0.46633881454398036003
xx3 = -0.73303694842681344599
xx4 = -1.13254799829068817198
.........MODULE 2: array elements.......
e10 = -0.06834219862995164003
e11 = -0.46263765626356890425
e12 = -0.72971838784369053319
e13 = -1.12397907004612833261
.........MODULE 3: array as parameter...
e1_0 = -0.05533645259179445915
e1_1 = -0.44743656275474680588
e1_2 = -0.71097338928518248722
e1_3 = -1.10309805692560095340
.........MODULE 4: conditional jumps....
val = 1
.........MODULE 6: integer arithmetic...
e1k = 6.00000000000000000000
e1l = 6.00000000000000000000
.........MODULE 7: trig. functions......
x1 = 0.49040731615907084197
y1 = 0.49039249795612543821
.........MODULE 8: procedure calls......
z = 0.99993750062499996275
.........MODULE9: array references......
e1_j = -1.10309805692560095340
e1_k = 3.00000000000000000000
e1_l = -1.10309805692560095340
.........MODULE10: integer arithmetic...
j = 2
k = 3
.........MODULE11: standard functions...
x = 0.83466551951904967427
........................................
Whetstone last time (uS) = 142583.00
Whetstone max time (uS) = 142583.00
Whetstone min time (uS) = 142582.00
----------- DHRYSTONE 2 -------------
Dhrystone Benchmark, Version 2.1 (Language: C)
Program compiled without 'register' attribute
Execution starts, 20000 runs through Dhrystone
Execution ends
Final values of the variables used in the benchmark:
Int_Glob: 5
should be: 5
Bool_Glob: 1
should be: 1
Ch_1_Glob: A
should be: A
Ch_2_Glob: B
should be: B
Arr_1_Glob[8]: 7
should be: 7
Arr_2_Glob[8][7]: 20010
should be: Number_Of_Runs + 10
Ptr_Glob->
Ptr_Comp: 536867932
should be: (implementation-dependent)
Discr: 0
should be: 0
Enum_Comp: 2
should be: 2
Int_Comp: 17
should be: 17
Str_Comp: DHRYSTONE PROGRAM, SOME STRING
should be: DHRYSTONE PROGRAM, SOME STRING
Next_Ptr_Glob->
Ptr_Comp: 536867932
should be: (implementation-dependent), same as above
Discr: 0
should be: 0
Enum_Comp: 1
should be: 1
Int_Comp: 18
should be: 18
Str_Comp: DHRYSTONE PROGRAM, SOME STRING
should be: DHRYSTONE PROGRAM, SOME STRING
Int_1_Loc: 5
should be: 5
Int_2_Loc: 13
should be: 13
Int_3_Loc: 7
should be: 7
Enum_Loc: 1
should be: 1
Str_1_Loc: DHRYSTONE PROGRAM, 1'ST STRING
should be: DHRYSTONE PROGRAM, 1'ST STRING
Str_2_Loc: DHRYSTONE PROGRAM, 2'ND STRING
should be: DHRYSTONE PROGRAM, 2'ND STRING
Time = 93000 us
Microseconds for one run through Dhrystone: 4.65
Dhrystones per Second: 215053
----------- COREMARK -------------
2K performance run parameters for coremark.
CoreMark Size : 666
Total ticks : 29079282
Total time (secs): 29.079282
Iterations/Sec : 343.887445
Iterations : 10000
Compiler version : IAR 7.40.7
Compiler flags : -O3 -Otime
Memory location : IRAM
seedcrc : 0xe9f5
[0]crclist : 0xe714
[0]crcmatrix : 0x1fd7
[0]crcstate : 0x8e3a
[0]crcfinal : 0x988c
Correct operation validated. See readme.txt for run and reporting rules.
CoreMark 1.0 : 343.887445 / IAR 7.40.7 -O3 -Otime / IRAM
END.
Output to the terminal the results of a test compiled in Keil
----------- Speed Test -------------
Reference time = 100001 us
----------- umm malloc test -------------
...........................................................
|0x1fff60f8|B 0|NB 1|PB 0|Z 1|NF 1|PF 0|USED
|0x1fff6100|B 1|NB 4223|PB 0|Z 4222|NF 4223|PF 0|FREE
|0x1fffe4f0|B 4223|NB 0|PB 1|Z 1|NF 0|PF 1|USED
...........................................................
Total Entries 1 Used Entries 0 Free Entries 1
Total Blocks 4223 Used Blocks 0 Free Blocks 4223
Used space = 0 Free space = 33784
...........................................................
umm malloc test time = 1507118 uS
Allocations = 40000, uSec per allocation = 37
----------- Compressor LZSS test -------------
Uncompressed size = 2048
Compressed size = 1858. Max. alloc.= 14384. Time = 7406 uS
Decompress Max. alloc.= 1048. Time = 989 uS
Decompress Ok!
----------- Compressor Zlib test -------------
Uncompressed size = 2048
Compressed size = 1740. Max. alloc.= 24288. Time = 11716 uS
Decompress Max. alloc.= 15768. Time = 912 uS
Decompress Ok!
----------- Compressor S-LZW test -------------
Uncompressed size = 2048
Compressed size = 2035. Max. alloc.= 4320. Time = 2839 uS
Decompress Max. alloc.= 4320. Time = 4194 uS
Decompress Ok!
----------- Compressor FastLZ test -------------
Uncompressed size = 2048
Compressed size = 1834. Max. alloc.= 32776. Time = 1028 uS
Decompress Max. alloc.= 0. Time = 160 uS
Decompress Ok!
----------- WHETSTONE FLOAT -------------
.........MODULE 1: simple identifiers...
xx1 = -0.06679254770278930700
xx2 = -0.46633863449096680000
xx3 = -0.73303699493408203000
xx4 = -1.13254797458648680000
.........MODULE 2: array elements.......
e10 = -0.06834230571985244800
e11 = -0.46263590455055237000
e12 = -0.72971796989440918000
e13 = -1.12397670745849610000
.........MODULE 3: array as parameter...
e1_0 = -0.05533060804009437600
e1_1 = -0.44743216037750244000
e1_2 = -0.71096724271774292000
e1_3 = -1.10309338569641110000
.........MODULE 4: conditional jumps....
val = 1
.........MODULE 6: integer arithmetic...
e1k = 6.00000000000000000000
e1l = 6.00000000000000000000
.........MODULE 7: trig. functions......
x1 = 0.49040567874908447000
y1 = 0.49039086699485779000
.........MODULE 8: procedure calls......
z = 0.99993747472763062000
.........MODULE9: array references......
e1_j = -1.10309338569641110000
e1_k = 3.00000000000000000000
e1_l = -1.10309338569641110000
.........MODULE10: integer arithmetic...
j = 2
k = 3
.........MODULE11: standard functions...
x = 0.83466011285781860000
........................................
Whetstone last time (uS) = 08925.00
Whetstone max time (uS) = 08926.00
Whetstone min time (uS) = 08925.00
----------- WHETSTONE DOUBLE -------------
.........MODULE 1: simple identifiers...
xx1 = -0.06679268039452399000
xx2 = -0.46633881454398041000
xx3 = -0.73303694842681344000
xx4 = -1.13254799829068810000
.........MODULE 2: array elements.......
e10 = -0.06834219862995163900
e11 = -0.46263765626356895000
e12 = -0.72971838784369047000
e13 = -1.12397907004612830000
.........MODULE 3: array as parameter...
e1_0 = -0.05533645259179446200
e1_1 = -0.44743656275474680000
e1_2 = -0.71097338928518250000
e1_3 = -1.10309805692560080000
.........MODULE 4: conditional jumps....
val = 1
.........MODULE 6: integer arithmetic...
e1k = 6.00000000000000000000
e1l = 6.00000000000000000000
.........MODULE 7: trig. functions......
x1 = 0.49040731615903904000
y1 = 0.49039249795609352000
.........MODULE 8: procedure calls......
z = 0.99993750062499998000
.........MODULE9: array references......
e1_j = -1.10309805692560080000
e1_k = 3.00000000000000000000
e1_l = -1.10309805692560080000
.........MODULE10: integer arithmetic...
j = 2
k = 3
.........MODULE11: standard functions...
x = 0.83466551951905033000
........................................
Whetstone last time (uS) = 166298.00
Whetstone max time (uS) = 166298.00
Whetstone min time (uS) = 166298.00
----------- DHRYSTONE 2 -------------
Dhrystone Benchmark, Version 2.1 (Language: C)
Program compiled without 'register' attribute
Execution starts, 20000 runs through Dhrystone
Execution ends
Final values of the variables used in the benchmark:
Int_Glob: 5
should be: 5
Bool_Glob: 1
should be: 1
Ch_1_Glob: A
should be: A
Ch_2_Glob: B
should be: B
Arr_1_Glob[8]: 7
should be: 7
Arr_2_Glob[8][7]: 20010
should be: Number_Of_Runs + 10
Ptr_Glob->
Ptr_Comp: 536870816
should be: (implementation-dependent)
Discr: 0
should be: 0
Enum_Comp: 2
should be: 2
Int_Comp: 17
should be: 17
Str_Comp: DHRYSTONE PROGRAM, SOME STRING
should be: DHRYSTONE PROGRAM, SOME STRING
Next_Ptr_Glob->
Ptr_Comp: 536870816
should be: (implementation-dependent), same as above
Discr: 0
should be: 0
Enum_Comp: 1
should be: 1
Int_Comp: 18
should be: 18
Str_Comp: DHRYSTONE PROGRAM, SOME STRING
should be: DHRYSTONE PROGRAM, SOME STRING
Int_1_Loc: 5
should be: 5
Int_2_Loc: 13
should be: 13
Int_3_Loc: 7
should be: 7
Enum_Loc: 1
should be: 1
Str_1_Loc: DHRYSTONE PROGRAM, 1'ST STRING
should be: DHRYSTONE PROGRAM, 1'ST STRING
Str_2_Loc: DHRYSTONE PROGRAM, 2'ND STRING
should be: DHRYSTONE PROGRAM, 2'ND STRING
Time = 91000 us
Microseconds for one run through Dhrystone: 4.55
Dhrystones per Second: 219780
----------- COREMARK -------------
2K performance run parameters for coremark.
CoreMark Size : 666
Total ticks : 37386523
Total time (secs): 37.386523
Iterations/Sec : 267.476064
Iterations : 10000
Compiler version : armcc V5.06 (build 20)
Compiler flags : -O3 -Otime
Memory location : IRAM
seedcrc : 0xe9f5
[0]crclist : 0xe714
[0]crcmatrix : 0x1fd7
[0]crcstate : 0x8e3a
[0]crcfinal : 0x988c
Correct operation validated. See readme.txt for run and reporting rules.
CoreMark 1.0 : 267.476064 / armcc V5.06 (build 20) -O3 -Otime / IRAM
END.
Output to the terminal the results of a test compiled in GCC
----------- Speed Test -------------
Reference time = 100001 us
----------- umm malloc test -------------
...........................................................
|0x1fff5280|B 0|NB 1|PB 0|Z 1|NF 1|PF 0|USED
|0x1fff5288|B 1|NB 4223|PB 0|Z 4222|NF 4223|PF 0|FREE
|0x1fffd678|B 4223|NB 0|PB 1|Z 1|NF 0|PF 1|USED
...........................................................
Total Entries 1 Used Entries 0 Free Entries 1
Total Blocks 4223 Used Blocks 0 Free Blocks 4223
Used space = 0 Free space = 33784
...........................................................
umm malloc test time = 983944 uS
Allocations = 40000, uSec per allocation = 24
----------- Compressor LZSS test -------------
Uncompressed size = 2048
Compressed size = 1858. Max. alloc.= 14384. Time = 8504 uS
Decompress Max. alloc.= 1048. Time = 673 uS
Decompress Ok!
----------- Compressor Zlib test -------------
Uncompressed size = 2048
Compressed size = 1740. Max. alloc.= 24288. Time = 12384 uS
Decompress Max. alloc.= 15768. Time = 1070 uS
Decompress Ok!
----------- Compressor S-LZW test -------------
Uncompressed size = 2048
Compressed size = 2035. Max. alloc.= 4320. Time = 2577 uS
Decompress Max. alloc.= 4320. Time = 4189 uS
Decompress Ok!
----------- Compressor FastLZ test -------------
Uncompressed size = 2048
Compressed size = 1834. Max. alloc.= 32776. Time = 1045 uS
Decompress Max. alloc.= 0. Time = 234 uS
Decompress Ok!
----------- WHETSTONE FLOAT -------------
.........MODULE 1: simple identifiers...
xx1 = -0.06679254770278930664
xx2 = -0.46633863449096679688
xx3 = -0.73303699493408203125
xx4 = -1.13254797458648681641
.........MODULE 2: array elements.......
e10 = -0.06834230571985244751
e11 = -0.46263590455055236816
e12 = -0.72971796989440917969
e13 = -1.12397670745849609375
.........MODULE 3: array as parameter...
e1_0 = -0.05533138290047645569
e1_1 = -0.44743290543556213379
e1_2 = -0.71096915006637573242
e1_3 = -1.10309529304504394531
.........MODULE 4: conditional jumps....
val = 1
.........MODULE 6: integer arithmetic...
e1k = 6.00000000000000000000
e1l = 6.00000000000000000000
.........MODULE 7: trig. functions......
x1 = 0.49040564894676208496
y1 = 0.49039086699485778809
.........MODULE 8: procedure calls......
z = 0.99993747472763061523
.........MODULE9: array references......
e1_j = -1.10309529304504394531
e1_k = 3.00000000000000000000
e1_l = -1.10309529304504394531
.........MODULE10: integer arithmetic...
j = 2
k = 3
.........MODULE11: standard functions...
x = 0.83466064929962158203
........................................
Whetstone last time (uS) = 13629.00
Whetstone max time (uS) = 13630.00
Whetstone min time (uS) = 13629.00
----------- WHETSTONE DOUBLE -------------
.........MODULE 1: simple identifiers...
xx1 = -0.06679268039452398997
xx2 = -0.46633881454398040667
xx3 = -0.73303694842681343946
xx4 = -1.13254799829068808492
.........MODULE 2: array elements.......
e10 = -0.06834219862995163930
e11 = -0.46263765626356895266
e12 = -0.72971838784369047470
e13 = -1.12397907004612829240
.........MODULE 3: array as parameter...
e1_0 = -0.05533645259179446191
e1_1 = -0.44743656275474680273
e1_2 = -0.71097338928518249990
e1_3 = -1.10309805692560081170
.........MODULE 4: conditional jumps....
val = 1
.........MODULE 6: integer arithmetic...
e1k = 6.00000000000000000000
e1l = 6.00000000000000000000
.........MODULE 7: trig. functions......
x1 = 0.49040731615904653573
y1 = 0.49039249795610123650
.........MODULE 8: procedure calls......
z = 0.99993750062499997533
.........MODULE9: array references......
e1_j = -1.10309805692560081170
e1_k = 3.00000000000000000000
e1_l = -1.10309805692560081170
.........MODULE10: integer arithmetic...
j = 2
k = 3
.........MODULE11: standard functions...
x = 0.83466551951905032514
........................................
Whetstone last time (uS) = 219895.00
Whetstone max time (uS) = 219895.00
Whetstone min time (uS) = 219894.00
----------- DHRYSTONE 2 -------------
Dhrystone Benchmark, Version 2.1 (Language: C)
Program compiled without 'register' attribute
Execution starts, 20000 runs through Dhrystone
Execution ends
Final values of the variables used in the benchmark:
Int_Glob: 5
should be: 5
Bool_Glob: 1
should be: 1
Ch_1_Glob: A
should be: A
Ch_2_Glob: B
should be: B
Arr_1_Glob[8]: 7
should be: 7
Arr_2_Glob[8][7]: 20010
should be: Number_Of_Runs + 10
Ptr_Glob->
Ptr_Comp: 536870792
should be: (implementation-dependent)
Discr: 0
should be: 0
Enum_Comp: 2
should be: 2
Int_Comp: 17
should be: 17
Str_Comp: DHRYSTONE PROGRAM, SOME STRING
should be: DHRYSTONE PROGRAM, SOME STRING
Next_Ptr_Glob->
Ptr_Comp: 536870792
should be: (implementation-dependent), same as above
Discr: 0
should be: 0
Enum_Comp: 1
should be: 1
Int_Comp: 18
should be: 18
Str_Comp: DHRYSTONE PROGRAM, SOME STRING
should be: DHRYSTONE PROGRAM, SOME STRING
Int_1_Loc: 5
should be: 5
Int_2_Loc: 13
should be: 13
Int_3_Loc: 7
should be: 7
Enum_Loc: 1
should be: 1
Str_1_Loc: DHRYSTONE PROGRAM, 1'ST STRING
should be: DHRYSTONE PROGRAM, 1'ST STRING
Str_2_Loc: DHRYSTONE PROGRAM, 2'ND STRING
should be: DHRYSTONE PROGRAM, 2'ND STRING
Time = 74000 us
Microseconds for one run through Dhrystone: 3.70
Dhrystones per Second: 270270
----------- COREMARK -------------
2K performance run parameters for coremark.
CoreMark Size : 666
Total ticks : 45291107
Total time (secs): 45.291107
Iterations/Sec : 220.793897
Iterations : 10000
Compiler version : GCC4.8.4 20140725 (release) [ARM/embedded-4_8-branch revision 213147]
Compiler flags : -O3 -Otime
Memory location : IRAM
seedcrc : 0xe9f5
[0]crclist : 0xe714
[0]crcmatrix : 0x1fd7
[0]crcstate : 0x8e3a
[0]crcfinal : 0x988c
Correct operation validated. See readme.txt for run and reporting rules.
CoreMark 1.0 : 220.793897 / GCC4.8.4 20140725 (release) [ARM/embedded-4_8-branch revision 213147] -O3 -Otime / IRAM
END.
Output to the terminal the results of a test running in Visual Studio 2015
----------- Speed Test -------------
Reference time = 100000 us
----------- umm malloc test -------------
...........................................................
|0x00ba5980|B 0|NB 1|PB 0|Z 1|NF 1|PF 0|USED
|0x00ba5988|B 1|NB 4223|PB 0|Z 4222|NF 4223|PF 0|FREE
|0x00badd78|B 4223|NB 0|PB 1|Z 1|NF 0|PF 1|USED
...........................................................
Total Entries 1 Used Entries 0 Free Entries 1
Total Blocks 4223 Used Blocks 0 Free Blocks 4223
Used space = 0 Free space = 33784
...........................................................
umm malloc test time = 13586 uS
Allocations = 40000, uSec per allocation = 0
----------- Compressor LZSS test -------------
Uncompressed size = 2048
Compressed size = 1858. Max. alloc.= 14384. Time = 146 uS
Decompress Max. alloc.= 1048. Time = 10 uS
Decompress Ok!
----------- Compressor Zlib test -------------
Uncompressed size = 2048
Compressed size = 1740. Max. alloc.= 24288. Time = 241 uS
Decompress Max. alloc.= 15768. Time = 20 uS
Decompress Ok!
----------- Compressor S-LZW test -------------
Uncompressed size = 2048
Compressed size = 2035. Max. alloc.= 4320. Time = 47 uS
Decompress Max. alloc.= 4320. Time = 31 uS
Decompress Ok!
----------- Compressor FastLZ test -------------
Uncompressed size = 2048
Compressed size = 1834. Max. alloc.= 32776. Time = 13 uS
Decompress Max. alloc.= 0. Time = 4 uS
Decompress Ok!
----------- WHETSTONE FLOAT -------------
.........MODULE 1: simple identifiers...
xx1 = -0.06679254770278930700
xx2 = -0.46633863449096680000
xx3 = -0.73303699493408203000
xx4 = -1.13254797458648680000
.........MODULE 2: array elements.......
e10 = -0.06834230571985244800
e11 = -0.46263590455055237000
e12 = -0.72971796989440918000
e13 = -1.12397670745849610000
.........MODULE 3: array as parameter...
e1_0 = -0.05533060804009437600
e1_1 = -0.44743216037750244000
e1_2 = -0.71096724271774292000
e1_3 = -1.10309338569641110000
.........MODULE 4: conditional jumps....
val = 1
.........MODULE 6: integer arithmetic...
e1k = 6.00000000000000000000
e1l = 6.00000000000000000000
.........MODULE 7: trig. functions......
x1 = 0.49040612578392029000
y1 = 0.49039128422737122000
.........MODULE 8: procedure calls......
z = 0.99993747472763062000
.........MODULE9: array references......
e1_j = -1.10309338569641110000
e1_k = 3.00000000000000000000
e1_l = -1.10309338569641110000
.........MODULE10: integer arithmetic...
j = 2
k = 3
.........MODULE11: standard functions...
x = 0.83466225862503052000
........................................
Whetstone last time (uS) = 00131.00
Whetstone max time (uS) = 00168.00
Whetstone min time (uS) = 00123.00
----------- WHETSTONE DOUBLE -------------
.........MODULE 1: simple identifiers...
xx1 = -0.06679268039452399000
xx2 = -0.46633881454398041000
xx3 = -0.73303694842681344000
xx4 = -1.13254799829068810000
.........MODULE 2: array elements.......
e10 = -0.06834219862995163900
e11 = -0.46263765626356895000
e12 = -0.72971838784369047000
e13 = -1.12397907004612830000
.........MODULE 3: array as parameter...
e1_0 = -0.05533645259179446200
e1_1 = -0.44743656275474680000
e1_2 = -0.71097338928518250000
e1_3 = -1.10309805692560080000
.........MODULE 4: conditional jumps....
val = 1
.........MODULE 6: integer arithmetic...
e1k = 6.00000000000000000000
e1l = 6.00000000000000000000
.........MODULE 7: trig. functions......
x1 = 0.49040731615904543000
y1 = 0.49039249795610007000
.........MODULE 8: procedure calls......
z = 0.99993750062499998000
.........MODULE9: array references......
e1_j = -1.10309805692560080000
e1_k = 3.00000000000000000000
e1_l = -1.10309805692560080000
.........MODULE10: integer arithmetic...
j = 2
k = 3
.........MODULE11: standard functions...
x = 0.83466551951905787000
........................................
Whetstone last time (uS) = 00152.00
Whetstone max time (uS) = 00186.00
Whetstone min time (uS) = 00142.00
----------- DHRYSTONE 2 -------------
Dhrystone Benchmark, Version 2.1 (Language: C)
Program compiled without 'register' attribute
Execution starts, 20000 runs through Dhrystone
Execution ends
Final values of the variables used in the benchmark:
Int_Glob: 5
should be: 5
Bool_Glob: 1
should be: 1
Ch_1_Glob: A
should be: A
Ch_2_Glob: B
should be: B
Arr_1_Glob[8]: 7
should be: 7
Arr_2_Glob[8][7]: 20010
should be: Number_Of_Runs + 10
Ptr_Glob->
Ptr_Comp: 3602560
should be: (implementation-dependent)
Discr: 0
should be: 0
Enum_Comp: 2
should be: 2
Int_Comp: 17
should be: 17
Str_Comp: DHRYSTONE PROGRAM, SOME STRING
should be: DHRYSTONE PROGRAM, SOME STRING
Next_Ptr_Glob->
Ptr_Comp: 3602560
should be: (implementation-dependent), same as above
Discr: 0
should be: 0
Enum_Comp: 1
should be: 1
Int_Comp: 18
should be: 18
Str_Comp: DHRYSTONE PROGRAM, SOME STRING
should be: DHRYSTONE PROGRAM, SOME STRING
Int_1_Loc: 5
should be: 5
Int_2_Loc: 13
should be: 13
Int_3_Loc: 7
should be: 7
Enum_Loc: 1
should be: 1
Str_1_Loc: DHRYSTONE PROGRAM, 1'ST STRING
should be: DHRYSTONE PROGRAM, 1'ST STRING
Str_2_Loc: DHRYSTONE PROGRAM, 2'ND STRING
should be: DHRYSTONE PROGRAM, 2'ND STRING
Time = 551 us
Microseconds for one run through Dhrystone: 0.03
Dhrystones per Second: 36297640
----------- COREMARK -------------
2K performance run parameters for coremark.
CoreMark Size : 666
Total ticks : 11012043
Total time (secs): 11.012043
Iterations/Sec : 9080.967083
Iterations : 100000
Compiler version : VS 2013
Compiler flags : -O3 -Otime
Memory location : IRAM
seedcrc : 0xe9f5
[0]crclist : 0xe714
[0]crcmatrix : 0x1fd7
[0]crcstate : 0x8e3a
[0]crcfinal : 0xd340
Correct operation validated. See readme.txt for run and reporting rules.
CoreMark 1.0 : 9080.967083 / VS 2015 -O3 -Otime / IRAM
END.
How did the Intel Core i7
The fact is that all compression algorithms and the memory manager were debugged in the Visual Studio environment and this is the processor of my computer. The archive contains the SpeedTest_VS directory, which contains the project for Visual Studio 2015 and in which you can continue to improve these algorithms if you wish.
Conclusion
IAR was the undisputed leader in the ranking. The most surprising thing is that while maintaining the effectiveness of IAR code, it is also a leader in compilation speed. The fly in the ointment unaccounted for here is that the IAR comes with quite a bit of a functional source editor. You can, of course, start compiling IAR tools directly from the KDS environment , since KDS is the same slightly modified Eclipse version of Mylyn . But the recompilation time then increases to 2 minutes! Have eclipsethere is also a drawback, it slowly opens, with a certain size of the source, contextual parsing starts to slow down. There is a SpeedTest_SE directory in the project of this test . It contains the file SpeedTest.vpw of the project workspace of this test for the SlickEdit editor . SlickEdit is a great, very fast editor, and the IAR IDE can automatically track file changes, so there is no problem sharing them.
The full project archive (97.7 MB) is available at https://drive.google.com/file/d/0B5dbvc_yPqJHYWI0OE9YZklKRjQ/view?usp=sharing