Skip to content

Commit

Permalink
Add separate ASM file for XGETBV64 and CPUID64 (GH #1240)
Browse files Browse the repository at this point in the history 
This will allow us to define CRYPTOPP_DISABLE_ASM and completely avoid building x64dll.asm and x64masm.asm
  • Loading branch information
@noloader
noloader committed Oct 6, 2023
1 parent 3405dce commit 0432085
Show file tree
Hide file tree
Showing 5 changed files with 131 additions and 104 deletions.
63 changes: 63 additions & 0 deletions cpuid64.asm
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,63 @@
;; https://docs.microsoft.com/en-us/cpp/build/x64-calling-convention
;; The first four integer arguments are passed in registers.
;; Integer values are passed in left-to-right order in RCX,
;; RDX, R8, and R9, respectively. Arguments five and higher
;; are passed on the stack.

;; The registers RAX, RCX, RDX, R8, R9, R10, R11, XMM0-5,
;; and the upper portions of YMM0-15 and ZMM0-15 are
;; considered volatile and must be considered destroyed on
;; function calls.

.CODE

TITLE CPU features source file
SUBTITLE Microsoft specific ASM code to utilize CPUID and XGETBV64 for down level Microsoft toolchains

;; http://www.agner.org/optimize/vectorclass/read.php?i=65
;; word64 Xgetbv(word32 ctrl)
;; ctrl = rcx

ALIGN 8
XGETBV64 PROC FRAME
.endprolog
;; query
DB 0fh, 01h, 0d0h
;; xcr = (EDX << 32) | EAX
and rax, 0ffffffffh
shl rdx, 32
or rax, rdx
ret
XGETBV64 ENDP

;; word64 CpuId(word32 func, word32 subfunc, word32 output[4])
;; func = rcx
;; subfunc = rdx
;; output = r8

ALIGN 8
CPUID64 PROC FRAME
;; preserve per ABI
mov [rsp+8], rbx
.savereg rbx, 8
.endprolog
;; eax = func
mov rax, rcx
;; ecx = subfunc
mov rcx, rdx
;; query
cpuid
;; save
mov [r8+0], eax
mov [r8+4], ebx
mov [r8+8], ecx
mov [r8+12], edx
;; return value
mov rax, 1
;; restore
mov rbx, [rsp+8]
ret
CPUID64 ENDP

_TEXT ENDS
END
119 changes: 60 additions & 59 deletions cryptest.nmake
View file
Original file line number Diff line number Diff line change
Expand Up @@ -57,68 +57,69 @@

LIB_SRCS = \
cryptlib.cpp cpu.cpp integer.cpp 3way.cpp adler32.cpp algebra.cpp \
algparam.cpp allocate.cpp arc4.cpp aria.cpp ariatab.cpp \
asn.cpp authenc.cpp base32.cpp base64.cpp basecode.cpp bfinit.cpp \
blake2.cpp blake2b_simd.cpp blake2s_simd.cpp blowfish.cpp blumshub.cpp \
camellia.cpp cast.cpp casts.cpp cbcmac.cpp ccm.cpp chacha.cpp \
chacha_avx.cpp chacha_simd.cpp chachapoly.cpp cham.cpp cham_simd.cpp \
channels.cpp cmac.cpp crc.cpp crc_simd.cpp darn.cpp default.cpp des.cpp \
dessp.cpp dh.cpp dh2.cpp dll.cpp donna_32.cpp donna_64.cpp donna_sse.cpp \
dsa.cpp eax.cpp ec2n.cpp eccrypto.cpp ecp.cpp elgamal.cpp emsa2.cpp \
eprecomp.cpp esign.cpp files.cpp filters.cpp fips140.cpp fipstest.cpp \
gcm.cpp gcm_simd.cpp gf256.cpp gf2_32.cpp gf2n.cpp gf2n_simd.cpp \
gfpcrypt.cpp gost.cpp gzip.cpp hc128.cpp hc256.cpp hex.cpp hight.cpp \
hmac.cpp hrtimer.cpp ida.cpp idea.cpp iterhash.cpp kalyna.cpp \
kalynatab.cpp keccak.cpp keccak_core.cpp keccak_simd.cpp lea.cpp \
lea_simd.cpp lsh256.cpp lsh256_avx.cpp lsh256_sse.cpp lsh512.cpp \
lsh512_avx.cpp lsh512_sse.cpp luc.cpp mars.cpp marss.cpp md2.cpp md4.cpp \
md5.cpp misc.cpp modes.cpp mqueue.cpp mqv.cpp nbtheory.cpp oaep.cpp \
osrng.cpp padlkrng.cpp panama.cpp pkcspad.cpp poly1305.cpp polynomi.cpp \
pssr.cpp pubkey.cpp queue.cpp rabbit.cpp rabin.cpp randpool.cpp rc2.cpp \
rc5.cpp rc6.cpp rdrand.cpp rdtables.cpp rijndael.cpp rijndael_simd.cpp \
ripemd.cpp rng.cpp rsa.cpp rw.cpp safer.cpp salsa.cpp scrypt.cpp \
seal.cpp seed.cpp serpent.cpp sha.cpp sha3.cpp sha_simd.cpp shacal2.cpp \
shacal2_simd.cpp shake.cpp shark.cpp sharkbox.cpp simeck.cpp simon.cpp \
simon128_simd.cpp skipjack.cpp sm3.cpp sm4.cpp sm4_simd.cpp \
sosemanuk.cpp speck.cpp speck128_simd.cpp square.cpp squaretb.cpp \
sse_simd.cpp strciphr.cpp tea.cpp tftables.cpp threefish.cpp tiger.cpp \
tigertab.cpp ttmac.cpp tweetnacl.cpp twofish.cpp vmac.cpp wake.cpp \
whrlpool.cpp xed25519.cpp xtr.cpp xtrcrypt.cpp xts.cpp zdeflate.cpp \
zinflate.cpp zlib.cpp
algparam.cpp allocate.cpp arc4.cpp aria.cpp ariatab.cpp asn.cpp \
authenc.cpp base32.cpp base64.cpp basecode.cpp bfinit.cpp blake2.cpp \
blake2b_simd.cpp blake2s_simd.cpp blowfish.cpp blumshub.cpp camellia.cpp \
cast.cpp casts.cpp cbcmac.cpp ccm.cpp chacha.cpp chacha_avx.cpp \
chacha_simd.cpp chachapoly.cpp cham.cpp cham_simd.cpp channels.cpp \
cmac.cpp crc.cpp crc_simd.cpp darn.cpp default.cpp des.cpp dessp.cpp \
dh.cpp dh2.cpp dll.cpp donna_32.cpp donna_64.cpp donna_sse.cpp dsa.cpp \
eax.cpp ec2n.cpp eccrypto.cpp ecp.cpp elgamal.cpp emsa2.cpp eprecomp.cpp \
esign.cpp files.cpp filters.cpp fips140.cpp fipstest.cpp gcm.cpp \
gcm_simd.cpp gf256.cpp gf2_32.cpp gf2n.cpp gf2n_simd.cpp gfpcrypt.cpp \
gost.cpp gzip.cpp hc128.cpp hc256.cpp hex.cpp hight.cpp hmac.cpp \
hrtimer.cpp ida.cpp idea.cpp iterhash.cpp kalyna.cpp kalynatab.cpp \
keccak.cpp keccak_core.cpp keccak_simd.cpp lea.cpp lea_simd.cpp \
lsh256.cpp lsh256_avx.cpp lsh256_sse.cpp lsh512.cpp lsh512_avx.cpp \
lsh512_sse.cpp luc.cpp mars.cpp marss.cpp md2.cpp md4.cpp md5.cpp \
misc.cpp modes.cpp mqueue.cpp mqv.cpp nbtheory.cpp oaep.cpp osrng.cpp \
padlkrng.cpp panama.cpp pkcspad.cpp poly1305.cpp polynomi.cpp \
ppc_simd.cpp primetab.cpp pssr.cpp pubkey.cpp queue.cpp rabbit.cpp \
rabin.cpp randpool.cpp rc2.cpp rc5.cpp rc6.cpp rdrand.cpp rdtables.cpp \
rijndael.cpp rijndael_simd.cpp ripemd.cpp rng.cpp rsa.cpp rw.cpp \
safer.cpp salsa.cpp scrypt.cpp seal.cpp seed.cpp serpent.cpp sha.cpp \
sha3.cpp sha_simd.cpp shacal2.cpp shacal2_simd.cpp shake.cpp shark.cpp \
sharkbox.cpp simeck.cpp simon.cpp simon128_simd.cpp skipjack.cpp sm3.cpp \
sm4.cpp sm4_simd.cpp sosemanuk.cpp speck.cpp speck128_simd.cpp \
square.cpp squaretb.cpp sse_simd.cpp strciphr.cpp tea.cpp tftables.cpp \
threefish.cpp tiger.cpp tigertab.cpp ttmac.cpp tweetnacl.cpp twofish.cpp \
vmac.cpp wake.cpp whrlpool.cpp xed25519.cpp xtr.cpp xtrcrypt.cpp xts.cpp \
zdeflate.cpp zinflate.cpp zlib.cpp

LIB_OBJS = \
cryptlib.obj cpu.obj integer.obj 3way.obj adler32.obj algebra.obj \
algparam.obj allocate.obj arc4.obj aria.obj ariatab.obj \
asn.obj authenc.obj base32.obj base64.obj basecode.obj bfinit.obj \
blake2.obj blake2b_simd.obj blake2s_simd.obj blowfish.obj blumshub.obj \
camellia.obj cast.obj casts.obj cbcmac.obj ccm.obj chacha.obj \
chacha_avx.obj chacha_simd.obj chachapoly.obj cham.obj cham_simd.obj \
channels.obj cmac.obj crc.obj crc_simd.obj darn.obj default.obj des.obj \
dessp.obj dh.obj dh2.obj dll.obj donna_32.obj donna_64.obj donna_sse.obj \
dsa.obj eax.obj ec2n.obj eccrypto.obj ecp.obj elgamal.obj emsa2.obj \
eprecomp.obj esign.obj files.obj filters.obj fips140.obj fipstest.obj \
gcm.obj gcm_simd.obj gf256.obj gf2_32.obj gf2n.obj gf2n_simd.obj \
gfpcrypt.obj gost.obj gzip.obj hc128.obj hc256.obj hex.obj hight.obj \
hmac.obj hrtimer.obj ida.obj idea.obj iterhash.obj kalyna.obj \
kalynatab.obj keccak.obj keccak_core.obj keccak_simd.obj lea.obj \
lea_simd.obj lsh256.obj lsh256_avx.obj lsh256_sse.obj lsh512.obj \
lsh512_avx.obj lsh512_sse.obj luc.obj mars.obj marss.obj md2.obj md4.obj \
md5.obj misc.obj modes.obj mqueue.obj mqv.obj nbtheory.obj oaep.obj \
osrng.obj padlkrng.obj panama.obj pkcspad.obj poly1305.obj polynomi.obj \
pssr.obj pubkey.obj queue.obj rabbit.obj rabin.obj randpool.obj rc2.obj \
rc5.obj rc6.obj rdrand.obj rdtables.obj rijndael.obj rijndael_simd.obj \
ripemd.obj rng.obj rsa.obj rw.obj safer.obj salsa.obj scrypt.obj \
seal.obj seed.obj serpent.obj sha.obj sha3.obj sha_simd.obj shacal2.obj \
shacal2_simd.obj shake.obj shark.obj sharkbox.obj simeck.obj simon.obj \
simon128_simd.obj skipjack.obj sm3.obj sm4.obj sm4_simd.obj \
sosemanuk.obj speck.obj speck128_simd.obj square.obj squaretb.obj \
sse_simd.obj strciphr.obj tea.obj tftables.obj threefish.obj tiger.obj \
tigertab.obj ttmac.obj tweetnacl.obj twofish.obj vmac.obj wake.obj \
whrlpool.obj xed25519.obj xtr.obj xtrcrypt.obj xts.obj zdeflate.obj \
zinflate.obj zlib.obj
algparam.obj allocate.obj arc4.obj aria.obj ariatab.obj asn.obj \
authenc.obj base32.obj base64.obj basecode.obj bfinit.obj blake2.obj \
blake2b_simd.obj blake2s_simd.obj blowfish.obj blumshub.obj camellia.obj \
cast.obj casts.obj cbcmac.obj ccm.obj chacha.obj chacha_avx.obj \
chacha_simd.obj chachapoly.obj cham.obj cham_simd.obj channels.obj \
cmac.obj crc.obj crc_simd.obj darn.obj default.obj des.obj dessp.obj \
dh.obj dh2.obj dll.obj donna_32.obj donna_64.obj donna_sse.obj dsa.obj \
eax.obj ec2n.obj eccrypto.obj ecp.obj elgamal.obj emsa2.obj eprecomp.obj \
esign.obj files.obj filters.obj fips140.obj fipstest.obj gcm.obj \
gcm_simd.obj gf256.obj gf2_32.obj gf2n.obj gf2n_simd.obj gfpcrypt.obj \
gost.obj gzip.obj hc128.obj hc256.obj hex.obj hight.obj hmac.obj \
hrtimer.obj ida.obj idea.obj iterhash.obj kalyna.obj kalynatab.obj \
keccak.obj keccak_core.obj keccak_simd.obj lea.obj lea_simd.obj \
lsh256.obj lsh256_avx.obj lsh256_sse.obj lsh512.obj lsh512_avx.obj \
lsh512_sse.obj luc.obj mars.obj marss.obj md2.obj md4.obj md5.obj \
misc.obj modes.obj mqueue.obj mqv.obj nbtheory.obj oaep.obj osrng.obj \
padlkrng.obj panama.obj pkcspad.obj poly1305.obj polynomi.obj \
ppc_simd.obj primetab.obj pssr.obj pubkey.obj queue.obj rabbit.obj \
rabin.obj randpool.obj rc2.obj rc5.obj rc6.obj rdrand.obj rdtables.obj \
rijndael.obj rijndael_simd.obj ripemd.obj rng.obj rsa.obj rw.obj \
safer.obj salsa.obj scrypt.obj seal.obj seed.obj serpent.obj sha.obj \
sha3.obj sha_simd.obj shacal2.obj shacal2_simd.obj shake.obj shark.obj \
sharkbox.obj simeck.obj simon.obj simon128_simd.obj skipjack.obj sm3.obj \
sm4.obj sm4_simd.obj sosemanuk.obj speck.obj speck128_simd.obj \
square.obj squaretb.obj sse_simd.obj strciphr.obj tea.obj tftables.obj \
threefish.obj tiger.obj tigertab.obj ttmac.obj tweetnacl.obj twofish.obj \
vmac.obj wake.obj whrlpool.obj xed25519.obj xtr.obj xtrcrypt.obj xts.obj \
zdeflate.obj zinflate.obj zlib.obj

ASM_OBJS = \
rdrand-x86.obj rdrand-x64.obj rdseed-x86.obj rdseed-x64.obj x64masm.obj x64dll.obj
cpuid64.obj rdrand-x86.obj rdrand-x64.obj rdseed-x86.obj rdseed-x64.obj \
x64masm.obj x64dll.obj

TEST_SRCS = \
test.cpp bench1.cpp bench2.cpp bench3.cpp datatest.cpp \
Expand Down Expand Up @@ -211,8 +212,8 @@ RDSEED_OBJ = rdseed-x86.obj
# CXXFLAGS = $(CXXFLAGS) /DWINAPI_FAMILY=WINAPI_FAMILY_APP
AS = ml64.exe
ASFLAGS = /nologo /D_M_X64 /W3 /Cx /Zi
LIB_SRCS = $(LIB_SRCS) rdrand.cpp rdrand.asm rdseed.asm
LIB_OBJS = $(LIB_OBJS) rdrand-x64.obj rdseed-x64.obj x64masm.obj x64dll.obj
LIB_SRCS = $(LIB_SRCS) cpuid64.asm rdrand.cpp rdrand.asm rdseed.asm
LIB_OBJS = $(LIB_OBJS) cpuid64.obj rdrand-x64.obj rdseed-x64.obj x64masm.obj x64dll.obj
LDFLAGS = $(LDFLAGS) /MACHINE:X64
LDLIBS = $(LDLIBS) kernel32.lib
RDRAND_OBJ = rdrand-x64.obj
Expand Down
5 changes: 5 additions & 0 deletions cryptlib.vcxproj
View file
Original file line number Diff line number Diff line change
Expand Up @@ -368,6 +368,11 @@
<None Include="Readme.txt" />
</ItemGroup>
<ItemGroup>
<CustomBuild Condition="'$(Platform)'=='x64'" Include="cpuid64.asm">
<Message>Building and Assembling cpuid64.asm</Message>
<Command>ml64.exe /c /nologo /D_M_X64 /W3 /Zi /Fo"$(IntDir)cpuid64.obj" "%(FullPath)"</Command>
<Outputs>$(IntDir)cpuid64.obj;%(Outputs)</Outputs>
</CustomBuild>
<CustomBuild Condition="'$(Platform)'=='Win32'" Include="rdrand.asm">
<Message>Building and assembling rdrand.asm</Message>
<Command>ml.exe /c /nologo /D_M_X86 /W3 /Cx /Zi /safeseh /Fo"$(IntDir)rdrand-x86.obj" "%(FullPath)"</Command>
Expand Down
3 changes: 3 additions & 0 deletions cryptlib.vcxproj.filters
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1088,6 +1088,9 @@
</ClInclude>
</ItemGroup>
<ItemGroup>
<CustomBuild Include="cpuid64.asm">
<Filter>Source Files</Filter>
</CustomBuild>
<CustomBuild Include="rdrand.asm">
<Filter>Source Files</Filter>
</CustomBuild>
Expand Down
45 changes: 0 additions & 45 deletions x64dll.asm
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1975,50 +1975,5 @@ pop rsi
ret
SHA256_HashMultipleBlocks_SSE2 ENDP

;; http://www.agner.org/optimize/vectorclass/read.php?i=65
;; word64 Xgetbv(word32 ctrl)
;; ctrl = rcx

ALIGN 8
XGETBV64 PROC FRAME
.endprolog
;; query
DB 0fh, 01h, 0d0h
;; xcr = (EDX << 32) | EAX
and rax, 0ffffffffh
shl rdx, 32
or rax, rdx
ret
XGETBV64 ENDP

;; word64 CpuId(word32 func, word32 subfunc, word32 output[4])
;; func = rcx
;; subfunc = rdx
;; output = r8

ALIGN 8
CPUID64 PROC FRAME
;; preserve per ABI
mov [rsp+8], rbx
.savereg rbx, 8
.endprolog
;; eax = func
mov rax, rcx
;; ecx = subfunc
mov rcx, rdx
;; query
cpuid
;; save
mov [r8+0], eax
mov [r8+4], ebx
mov [r8+8], ecx
mov [r8+12], edx
;; return value
mov rax, 1
;; restore
mov rbx, [rsp+8]
ret
CPUID64 ENDP

_TEXT ENDS
END

4 comments on commit 0432085

@OgreTransporter
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good patch! But you forgot to change the files cryptdll.vcxproj and cryptdll.vcxproj.filters as well.

@noloader
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@OgreTransporter,

That should be cleared at Commit 1e20219ecddc.

The DLL project is scheduled to go away. You should start thinking about an alternative. Also see https://www.cryptopp.com/wiki/Wrapper_DLL.

@OgreTransporter
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you! It's no problem for me. I link the static library. Perhaps one should then delete the DLL project to avoid misunderstandings.

But wouldn't it perhaps make sense to replace the various VS project files and Makefiles with something like CMake? Then you only have to maintain one file.

@noloader
Copy link
Collaborator Author

@noloader noloader commented on 0432085 Oct 26, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you! It's no problem for me. I link the static library.

Very good.

Perhaps one should then delete the DLL project to avoid misunderstandings.

Yeah, that's been marked as deprecated for several years. We want to ensure folks have plenty of time to move away from it. See FIPS DLL deprecation in https://www.cryptopp.com/release800.html and friends.

Replace the various VS project files and Makefiles with something like CMake?

We tried to support CMake around Crypto++ 5.6.3. We had to remove the support around Crypto++ 6.0. There were too many problems with CMake, and it accounted for something like 23% of our bug reports. We did not have enough tribal knowledge or manpower to work the bugs. We consider it a failed experiment.

If CMake ever gets stable, then we may revisit it.

A CMake set of build files is available from Abdessattar Sassi (@abdes) at https://github.com/abdes/cryptopp-cmake. God bless him.

Please sign in to comment.