ExternalVendorCode/ZeroTierOne
1
0
Fork 0
mirror of https://github.com/zerotier/ZeroTierOne.git synced 2024-12-19 13:07:55 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

more fastitude

Browse Source
This commit is contained in:
Adam Ierymenko 2019-09-05 09:54:30 -07:00
parent 0599c1710b
commit 43e6a9e9ee
No known key found for this signature in database
GPG Key ID: C8877CF2D7A5D7F3
2 changed files with 48 additions and 68 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

95
node/AES.hpp
View File

@ -505,73 +505,54 @@ private:
const __m128i k13 = _k.ni.k[13]; const __m128i k13 = _k.ni.k[13];
const __m128i k14 = _k.ni.k[14]; const __m128i k14 = _k.ni.k[14];
while (len >= 64) { #define ZT_AES_CTR_AESNI_ROUND(k) \
c0 = _mm_aesenc_si128(c0,k); \
c1 = _mm_aesenc_si128(c1,k); \
c2 = _mm_aesenc_si128(c2,k); \
c3 = _mm_aesenc_si128(c3,k); \
c4 = _mm_aesenc_si128(c4,k); \
c5 = _mm_aesenc_si128(c5,k); \
c6 = _mm_aesenc_si128(c6,k); \
c7 = _mm_aesenc_si128(c7,k)
while (len >= 128) {
__m128i c0 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton(ctr),iv0),k0); __m128i c0 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton(ctr),iv0),k0);
__m128i c1 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+1ULL)),iv0),k0); __m128i c1 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+1ULL)),iv0),k0);
__m128i c2 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+2ULL)),iv0),k0); __m128i c2 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+2ULL)),iv0),k0);
__m128i c3 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+3ULL)),iv0),k0); __m128i c3 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+3ULL)),iv0),k0);
ctr += 4; __m128i c4 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+4ULL)),iv0),k0);
c0 = _mm_aesenc_si128(c0,k1); __m128i c5 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+5ULL)),iv0),k0);
c1 = _mm_aesenc_si128(c1,k1); __m128i c6 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+6ULL)),iv0),k0);
c2 = _mm_aesenc_si128(c2,k1); __m128i c7 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton((uint64_t)(ctr+7ULL)),iv0),k0);
c3 = _mm_aesenc_si128(c3,k1); ctr += 8;
c0 = _mm_aesenc_si128(c0,k2); ZT_AES_CTR_AESNI_ROUND(k1);
c1 = _mm_aesenc_si128(c1,k2); ZT_AES_CTR_AESNI_ROUND(k2);
c2 = _mm_aesenc_si128(c2,k2); ZT_AES_CTR_AESNI_ROUND(k3);
c3 = _mm_aesenc_si128(c3,k2); ZT_AES_CTR_AESNI_ROUND(k4);
c0 = _mm_aesenc_si128(c0,k3); ZT_AES_CTR_AESNI_ROUND(k5);
c1 = _mm_aesenc_si128(c1,k3); ZT_AES_CTR_AESNI_ROUND(k6);
c2 = _mm_aesenc_si128(c2,k3); ZT_AES_CTR_AESNI_ROUND(k7);
c3 = _mm_aesenc_si128(c3,k3); ZT_AES_CTR_AESNI_ROUND(k8);
c0 = _mm_aesenc_si128(c0,k4); ZT_AES_CTR_AESNI_ROUND(k9);
c1 = _mm_aesenc_si128(c1,k4); ZT_AES_CTR_AESNI_ROUND(k10);
c2 = _mm_aesenc_si128(c2,k4); ZT_AES_CTR_AESNI_ROUND(k11);
c3 = _mm_aesenc_si128(c3,k4); ZT_AES_CTR_AESNI_ROUND(k12);
c0 = _mm_aesenc_si128(c0,k5); ZT_AES_CTR_AESNI_ROUND(k13);
c1 = _mm_aesenc_si128(c1,k5);
c2 = _mm_aesenc_si128(c2,k5);
c3 = _mm_aesenc_si128(c3,k5);
c0 = _mm_aesenc_si128(c0,k6);
c1 = _mm_aesenc_si128(c1,k6);
c2 = _mm_aesenc_si128(c2,k6);
c3 = _mm_aesenc_si128(c3,k6);
c0 = _mm_aesenc_si128(c0,k7);
c1 = _mm_aesenc_si128(c1,k7);
c2 = _mm_aesenc_si128(c2,k7);
c3 = _mm_aesenc_si128(c3,k7);
c0 = _mm_aesenc_si128(c0,k8);
c1 = _mm_aesenc_si128(c1,k8);
c2 = _mm_aesenc_si128(c2,k8);
c3 = _mm_aesenc_si128(c3,k8);
c0 = _mm_aesenc_si128(c0,k9);
c1 = _mm_aesenc_si128(c1,k9);
c2 = _mm_aesenc_si128(c2,k9);
c3 = _mm_aesenc_si128(c3,k9);
c0 = _mm_aesenc_si128(c0,k10);
c1 = _mm_aesenc_si128(c1,k10);
c2 = _mm_aesenc_si128(c2,k10);
c3 = _mm_aesenc_si128(c3,k10);
c0 = _mm_aesenc_si128(c0,k11);
c1 = _mm_aesenc_si128(c1,k11);
c2 = _mm_aesenc_si128(c2,k11);
c3 = _mm_aesenc_si128(c3,k11);
c0 = _mm_aesenc_si128(c0,k12);
c1 = _mm_aesenc_si128(c1,k12);
c2 = _mm_aesenc_si128(c2,k12);
c3 = _mm_aesenc_si128(c3,k12);
c0 = _mm_aesenc_si128(c0,k13);
c1 = _mm_aesenc_si128(c1,k13);
c2 = _mm_aesenc_si128(c2,k13);
c3 = _mm_aesenc_si128(c3,k13);
_mm_storeu_si128((__m128i *)out,_mm_xor_si128(_mm_loadu_si128((const __m128i *)in),_mm_aesenclast_si128(c0,k14))); _mm_storeu_si128((__m128i *)out,_mm_xor_si128(_mm_loadu_si128((const __m128i *)in),_mm_aesenclast_si128(c0,k14)));
_mm_storeu_si128((__m128i *)(out + 16),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 16)),_mm_aesenclast_si128(c1,k14))); _mm_storeu_si128((__m128i *)(out + 16),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 16)),_mm_aesenclast_si128(c1,k14)));
_mm_storeu_si128((__m128i *)(out + 32),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 32)),_mm_aesenclast_si128(c2,k14))); _mm_storeu_si128((__m128i *)(out + 32),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 32)),_mm_aesenclast_si128(c2,k14)));
_mm_storeu_si128((__m128i *)(out + 48),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 48)),_mm_aesenclast_si128(c3,k14))); _mm_storeu_si128((__m128i *)(out + 48),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 48)),_mm_aesenclast_si128(c3,k14)));
in += 64; _mm_storeu_si128((__m128i *)(out + 64),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 64)),_mm_aesenclast_si128(c4,k14)));
out += 64; _mm_storeu_si128((__m128i *)(out + 80),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 80)),_mm_aesenclast_si128(c5,k14)));
len -= 64; _mm_storeu_si128((__m128i *)(out + 96),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 96)),_mm_aesenclast_si128(c6,k14)));
_mm_storeu_si128((__m128i *)(out + 112),_mm_xor_si128(_mm_loadu_si128((const __m128i *)(in + 112)),_mm_aesenclast_si128(c7,k14)));
in += 128;
out += 128;
len -= 128;
} }
#undef ZT_AES_CTR_AESNI_ROUND
while (len >= 16) { while (len >= 16) {
__m128i c0 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton(ctr++),(__m64)iv0),k0); __m128i c0 = _mm_xor_si128(_mm_set_epi64((__m64)Utils::hton(ctr++),(__m64)iv0),k0);
c0 = _mm_aesenc_si128(c0,k1); c0 = _mm_aesenc_si128(c0,k1);

21
selftest.cpp
View File

@ -207,33 +207,32 @@ static int testCrypto()
} }
std::cout << "OK" ZT_EOL_S << " GMAC-AES-256 (benchmark): "; std::cout.flush(); std::cout << "OK" ZT_EOL_S << " GMAC-AES-256 (benchmark): "; std::cout.flush();
int64_t start = OSUtils::now(); int64_t start = OSUtils::now();
for(unsigned long i=0;i<200000;++i) { for(unsigned long i=0;i<500000;++i) {
tv.gmac((const uint8_t *)buf1,buf1,sizeof(buf1),(uint8_t *)buf1); tv.gmac((const uint8_t *)buf1,buf1,ZT_DEFAULT_MTU,(uint8_t *)buf1);
} }
int64_t end = OSUtils::now(); int64_t end = OSUtils::now();
*dummy = hexbuf[0]; *dummy = hexbuf[0];
std::cout << (((double)(200000 * sizeof(buf1)) / 1048576.0) / ((double)(end - start) / 1000.0)) << " MiB/second" ZT_EOL_S; std::cout << (((double)(500000 * ZT_DEFAULT_MTU) / 1048576.0) / ((double)(end - start) / 1000.0)) << " MiB/second (dummy: " << (unsigned int)*dummy << ")" ZT_EOL_S;
std::cout << " AES-256-CTR (benchmark): "; std::cout.flush(); std::cout << " AES-256-CTR (benchmark): "; std::cout.flush();
start = OSUtils::now(); start = OSUtils::now();
for(unsigned long i=0;i<200000;++i) { for(unsigned long i=0;i<500000;++i) {
tv.ctr((const uint8_t *)hexbuf,buf1,sizeof(buf1),buf2); tv.ctr((const uint8_t *)hexbuf,buf1,ZT_DEFAULT_MTU,buf1);
hexbuf[0] = buf2[0]; *dummy = buf1[0];
} }
end = OSUtils::now(); end = OSUtils::now();
*dummy = buf2[0]; std::cout << (((double)(500000 * ZT_DEFAULT_MTU) / 1048576.0) / ((double)(end - start) / 1000.0)) << " MiB/second (dummy: " << (unsigned int)*dummy << ")" ZT_EOL_S;
std::cout << (((double)(200000 * sizeof(buf1)) / 1048576.0) / ((double)(end - start) / 1000.0)) << " MiB/second" ZT_EOL_S;
} }
{ {
std::cout << " AES-256-GMAC-CTR (benchmark): "; std::cout.flush(); std::cout << " AES-256-GMAC-CTR (benchmark): "; std::cout.flush();
AES k1,k2,k3,k4; AES k1,k2,k3,k4;
AES::initGmacCtrKeys(AES_TEST_VECTOR_0_KEY,k1,k2,k3,k4); AES::initGmacCtrKeys(AES_TEST_VECTOR_0_KEY,k1,k2,k3,k4);
int64_t start = OSUtils::now(); int64_t start = OSUtils::now();
for(unsigned long i=0;i<200000;++i) { for(unsigned long i=0;i<500000;++i) {
AES::ztGmacCtrEncrypt(k1,k2,k3,k4,(const uint8_t *)hexbuf,buf1,sizeof(buf1),buf1,(uint8_t *)(hexbuf + 8)); AES::ztGmacCtrEncrypt(k1,k2,k3,k4,(const uint8_t *)hexbuf,buf1,ZT_DEFAULT_MTU,buf1,(uint8_t *)(hexbuf + 8));
*dummy = buf1[0]; *dummy = buf1[0];
} }
int64_t end = OSUtils::now(); int64_t end = OSUtils::now();
std::cout << (((double)(200000 * sizeof(buf1)) / 1048576.0) / ((double)(end - start) / 1000.0)) << " MiB/second" ZT_EOL_S; std::cout << (((double)(500000 * ZT_DEFAULT_MTU) / 1048576.0) / ((double)(end - start) / 1000.0)) << " MiB/second (dummy: " << (unsigned int)*dummy << ")" ZT_EOL_S;
} }
} }