SLIDE 1
XKCP internals
Gilles Van Assche1
1STMicroelectronics
SCA workshop Šibenik, Croatia, June 2019 Based on joint work with Ronny Van Keer
1 / 19
Outline
1
Introduction
2
Inside the XKCP
3
Below SnP and PlSnP
4
Build system
2 / 19
XKCP internals Gilles Van Assche 1 1 STMicroelectronics SCA workshop - - PowerPoint PPT Presentation
XKCP internals Gilles Van Assche 1 1 STMicroelectronics SCA workshop - - PowerPoint PPT Presentation
XKCP internals Gilles Van Assche 1 1 STMicroelectronics SCA workshop ibenik, Croatia, June 2019 Based on joint work with Ronny Van Keer 1 / 19 Outline 1 Introduction 2 Inside the XKCP 3 Below SnP and PlSnP 4 Build system 2 / 19
SLIDE 2
SLIDE 3
Introduction
Outline
1
Introduction
2
Inside the XKCP
3
Below SnP and PlSnP
4
Build system
3 / 19
SLIDE 4
Introduction
What is the XKCP?
Previously known as the Keccak Code Package… Repository of implementations of Keccak-p[200 to 1600], SHA-3, (c)SHAKE, KMAC, KangarooTwelve, Ketje, Keyak, Kravatte, … Xoodoo, Xoodyak, Xoofff, Xoofff-SANE, … … now eXtended KCP
4 / 19
SLIDE 5
Introduction
What is the XKCP?
Previously known as the Keccak Code Package… Repository of implementations of Keccak-p[200 to 1600], SHA-3, (c)SHAKE, KMAC, KangarooTwelve, Ketje, Keyak, Kravatte, … Xoodoo, Xoodyak, Xoofff, Xoofff-SANE, … … now eXtended KCP
4 / 19
SLIDE 6
Introduction
What is the XKCP?
Previously known as the Keccak Code Package… Repository of implementations of Keccak-p[200 to 1600], SHA-3, (c)SHAKE, KMAC, KangarooTwelve, Ketje, Keyak, Kravatte, … Xoodoo, Xoodyak, Xoofff, Xoofff-SANE, … … now eXtended KCP
4 / 19
SLIDE 7
Introduction
What is the XKCP?
Previously known as the Keccak Code Package… Repository of implementations of Keccak-p[200 to 1600], SHA-3, (c)SHAKE, KMAC, KangarooTwelve, Ketje, Keyak, Kravatte, … Xoodoo, Xoodyak, Xoofff, Xoofff-SANE, … … now eXtended KCP
4 / 19
SLIDE 8
Introduction
Where to fjnd it
https://github.com/XKCP/XKCP
5 / 19
SLIDE 9
Inside the XKCP
Outline
1
Introduction
2
Inside the XKCP
3
Below SnP and PlSnP
4
Build system
6 / 19
SLIDE 10
Inside the XKCP
A layered approach
Keccak-p[200] Keccak-p[1600] Xoodoo
Primitive
Sponge Duplex
Construction
Hashing MAC PRNG
- Auth. Enc.
Mode Generic focus on user
easy to use e.g., message queue
- ne implementation
pointers and arithmetic
Specifjc focus on developer
limited scope to optimize unit tests
tailored implementations
permutation bulk data processing
7 / 19
SLIDE 11
Inside the XKCP
A layered approach
Keccak-p[200] Keccak-p[1600] Xoodoo
Primitive
Sponge Duplex
Construction
Hashing MAC PRNG
- Auth. Enc.
Mode Generic focus on user
easy to use e.g., message queue
- ne implementation
pointers and arithmetic
Specifjc focus on developer
limited scope to optimize unit tests
tailored implementations
permutation bulk data processing
7 / 19
SLIDE 12
Inside the XKCP
A layered approach
Keccak-p[200] Keccak-p[1600] Xoodoo
Primitive
Sponge Duplex
Construction
Hashing MAC PRNG
- Auth. Enc.
Mode Generic focus on user
easy to use e.g., message queue
- ne implementation
pointers and arithmetic
Specifjc focus on developer
limited scope to optimize unit tests
tailored implementations
permutation bulk data processing
7 / 19
SLIDE 13
Inside the XKCP
A layered approach
Keccak-p[200] Keccak-p[1600] Xoodoo
Primitive
Sponge Duplex
Construction
Hashing MAC PRNG
- Auth. Enc.
Mode
SnP
Generic focus on user
easy to use e.g., message queue
- ne implementation
pointers and arithmetic
Specifjc focus on developer
limited scope to optimize unit tests
tailored implementations
permutation bulk data processing
7 / 19
SLIDE 14
Inside the XKCP
Parallel processing
4×Keccak-p[1600] 8×Keccak-p[1600] 16×Xoodoo
Primitive
Parallel Sponges Farfalle
Construction
KangarooTwelve Deck-SANE Deck-WBC
Mode SnP = State and Permutation PlSnP = Parallel States and Permutations
8 / 19
SLIDE 15
Inside the XKCP
Parallel processing
4×Keccak-p[1600] 8×Keccak-p[1600] 16×Xoodoo
Primitive
Parallel Sponges Farfalle
Construction
KangarooTwelve Deck-SANE Deck-WBC
Mode
PlSnP
SnP = State and Permutation PlSnP = Parallel States and Permutations
8 / 19
SLIDE 16
Below SnP and PlSnP
Outline
1
Introduction
2
Inside the XKCP
3
Below SnP and PlSnP
4
Build system
9 / 19
SLIDE 17
Below SnP and PlSnP
Multiple implementations
Keccak-p[1600]
- pt32
- pt64
AVX2 AVX512 ARMv7M ARMv8A 2×Keccak-p[1600] fallback AVX2 AVX512 NEON Xoodoo
- pt32
ARMv6M ARMv7M ARMv7A AVR8 4×Xoodoo fallback AVX2 AVX512 NEON
SnP or PlSnP
Assumption: the state representation is opaque!
10 / 19
SLIDE 18
Below SnP and PlSnP
Multiple implementations
Keccak-p[1600]
- pt32
- pt64
AVX2 AVX512 ARMv7M ARMv8A 2×Keccak-p[1600] fallback AVX2 AVX512 NEON Xoodoo
- pt32
ARMv6M ARMv7M ARMv7A AVR8 4×Xoodoo fallback AVX2 AVX512 NEON
SnP or PlSnP
Assumption: the state representation is opaque!
10 / 19
SLIDE 19
Below SnP and PlSnP
Operations on the state
initialize the state apply the permutation f XOR/overwrite bytes into the state extract bytes from the state
and optionally XOR them
11 / 19
SLIDE 20
Below SnP and PlSnP
Example for Keccak-p[1600]
Declarations in KeccakP-1600-SnP.h: #define KeccakP1600_implementation “generic 64-bit optimized implementation” #define KeccakP1600_stateSizeInBytes 200 #define KeccakP1600_stateAlignment 8 Typical functions to be implemented (or macro’ed): KeccakP1600_Initialize KeccakP1600_AddBytes KeccakP1600_OverwriteBytes KeccakP1600_Permute_Nrounds KeccakP1600_ExtractBytes KeccakP1600_ExtractAndAddBytes
12 / 19
SLIDE 21
Below SnP and PlSnP
Example for Keccak-p[1600]
Declarations in KeccakP-1600-SnP.h: #define KeccakP1600_implementation “generic 64-bit optimized implementation” #define KeccakP1600_stateSizeInBytes 200 #define KeccakP1600_stateAlignment 8 Typical functions to be implemented (or macro’ed): KeccakP1600_Initialize KeccakP1600_AddBytes KeccakP1600_OverwriteBytes KeccakP1600_Permute_Nrounds KeccakP1600_ExtractBytes KeccakP1600_ExtractAndAddBytes
12 / 19
SLIDE 22
Below SnP and PlSnP
Example for Keccak-p[1600]
Declarations in KeccakP-1600-SnP.h: #define KeccakP1600_implementation “generic 64-bit optimized implementation” #define KeccakP1600_stateSizeInBytes 200 #define KeccakP1600_stateAlignment 8 Typical functions to be implemented (or macro’ed): KeccakP1600_Initialize KeccakP1600_AddBytes KeccakP1600_OverwriteBytes KeccakP1600_Permute_Nrounds KeccakP1600_ExtractBytes KeccakP1600_ExtractAndAddBytes
12 / 19
SLIDE 23
Below SnP and PlSnP
Example for Keccak-p[1600]
Declarations in KeccakP-1600-SnP.h: #define KeccakP1600_implementation “generic 64-bit optimized implementation” #define KeccakP1600_stateSizeInBytes 200 #define KeccakP1600_stateAlignment 8 Typical functions to be implemented (or macro’ed): KeccakP1600_Initialize KeccakP1600_AddBytes KeccakP1600_OverwriteBytes KeccakP1600_Permute_Nrounds KeccakP1600_ExtractBytes KeccakP1600_ExtractAndAddBytes
12 / 19
SLIDE 24
Below SnP and PlSnP
Example for Keccak-p[1600]
Declarations in KeccakP-1600-SnP.h: #define KeccakP1600_implementation “generic 64-bit optimized implementation” #define KeccakP1600_stateSizeInBytes 200 #define KeccakP1600_stateAlignment 8 Typical functions to be implemented (or macro’ed): KeccakP1600_Initialize KeccakP1600_AddBytes KeccakP1600_OverwriteBytes KeccakP1600_Permute_Nrounds KeccakP1600_ExtractBytes KeccakP1600_ExtractAndAddBytes
12 / 19
SLIDE 25
Below SnP and PlSnP
Fast loop optimization
Specialized repeated application of some operations (optional)
13 / 19
SLIDE 26
Below SnP and PlSnP
Parallel operations on the states
f f f
…
f f f
…
f f f
…
Functions on individual instances Functions on all instances
Parallel application of f XOR blocks into state
Optional fast loop optimization
14 / 19
SLIDE 27
Below SnP and PlSnP
Parallel operations on the states
f f f
…
f f f
…
f f f
…
Functions on individual instances Functions on all instances
Parallel application of f XOR blocks into state
Optional fast loop optimization
14 / 19
SLIDE 28
Below SnP and PlSnP
Parallel operations on the states
f f f
…
f f f
…
f f f
…
Functions on individual instances Functions on all instances
Parallel application of f XOR blocks into state
Optional fast loop optimization
14 / 19
SLIDE 29
Build system
Outline
1
Introduction
2
Inside the XKCP
3
Below SnP and PlSnP
4
Build system
15 / 19
SLIDE 30
Build system
Making targets
Making a library (.a or .so)
make generic64/libXKCP.a make generic32/libXKCP.a make Skylake/libXKCP.a make ARMv7A/libXKCP.a make compact/libXKCP.a
Extracting the source fjles
make generic64/libXKCP.a.pack
Running the unit tests
make generic64/UnitTests UnitTests --SnP --KangarooTwelve --Xoofff
And more: benchmarks, KeccakSum utility
16 / 19
SLIDE 31
Build system
Making targets
Making a library (.a or .so)
make generic64/libXKCP.a make generic32/libXKCP.a make Skylake/libXKCP.a make ARMv7A/libXKCP.a make compact/libXKCP.a
Extracting the source fjles
make generic64/libXKCP.a.pack
Running the unit tests
make generic64/UnitTests UnitTests --SnP --KangarooTwelve --Xoofff
And more: benchmarks, KeccakSum utility
16 / 19
SLIDE 32
Build system
Making targets
Making a library (.a or .so)
make generic64/libXKCP.a make generic32/libXKCP.a make Skylake/libXKCP.a make ARMv7A/libXKCP.a make compact/libXKCP.a
Extracting the source fjles
make generic64/libXKCP.a.pack
Running the unit tests
make generic64/UnitTests UnitTests --SnP --KangarooTwelve --Xoofff
And more: benchmarks, KeccakSum utility
16 / 19
SLIDE 33
Build system
Making targets
Making a library (.a or .so)
make generic64/libXKCP.a make generic32/libXKCP.a make Skylake/libXKCP.a make ARMv7A/libXKCP.a make compact/libXKCP.a
Extracting the source fjles
make generic64/libXKCP.a.pack
Running the unit tests
make generic64/UnitTests UnitTests --SnP --KangarooTwelve --Xoofff
And more: benchmarks, KeccakSum utility
16 / 19
SLIDE 34
Build system
XML-driven Makefjle
Makele.build
HighLevel.build LowLevel.build HOWTO-Customize.build
Target makele
In *.build, one defjnes fragments that … … are sets of fjles and compilation options … represent
a concrete service, mode, construction the implementation of a permutation
17 / 19
SLIDE 35
Build system
Customizing targets
HOWTO-Customize.build Examples: <target name=“XoodyakCortexM3”/> <target name=“K12onAVX2”/>
18 / 19
SLIDE 36
Build system
Customizing targets
HOWTO-Customize.build Examples: <target name=“XoodyakCortexM3”/> <target name=“K12onAVX2”/>
18 / 19
SLIDE 37
Build system
Customizing targets
HOWTO-Customize.build Examples: <target name=“XoodyakCortexM3” inherits=“Xoodyak”/> <target name=“K12onAVX2”/>
18 / 19
SLIDE 38
Build system
Customizing targets
HOWTO-Customize.build Examples: <target name=“XoodyakCortexM3” inherits=“Xoodyak
- ptimizedARMv7M”/>
<target name=“K12onAVX2”/>
18 / 19
SLIDE 39
Build system
Customizing targets
HOWTO-Customize.build Examples: <target name=“XoodyakCortexM3” inherits=“Xoodyak
- ptimizedARMv7M”/>
<target name=“K12onAVX2”/>
18 / 19
SLIDE 40
Build system
Customizing targets
HOWTO-Customize.build Examples: <target name=“XoodyakCortexM3” inherits=“Xoodyak
- ptimizedARMv7M”/>
<target name=“K12onAVX2” inherits=“KangarooTwelve”/>
18 / 19
SLIDE 41
Build system
Customizing targets
HOWTO-Customize.build Examples: <target name=“XoodyakCortexM3” inherits=“Xoodyak
- ptimizedARMv7M”/>
<target name=“K12onAVX2” inherits=“KangarooTwelve optimized1600AVX2”/>
18 / 19
SLIDE 42
Build system
Customizing targets
HOWTO-Customize.build Examples: <target name=“XoodyakCortexM3” inherits=“Xoodyak
- ptimizedARMv7M”/>
<target name=“K12onAVX2” inherits=“KangarooTwelve
- ptimized1600AVX2 SIMD256-AVX2u12”/>
18 / 19
SLIDE 43
Conclusions
Any questions?
Thanks for your attention!
https://keccak.team/
19 / 19