void speed_ed25519()
{
unsigned long long t[NTIMINGS];
int i;
unsigned char pk[32];
unsigned char sk[64];
// Benchmarking keypair
for(i=0;i<NTIMINGS;i++)
{
t[i] = cpucycles();
crypto_sign_keypair(pk, sk);
}
print_bench("sign_keypair",t,NTIMINGS);
// Benchmarking sign
for(i=0;i<NTIMINGS;i++)
{
t[i] = cpucycles();
crypto_sign(sm, &smlen, msg, MLEN, sk);
}
print_bench("sign (59 bytes)",t,NTIMINGS);
// Benchmarking sign_open
for(i=0;i<NTIMINGS;i++)
{
t[i] = cpucycles();
crypto_sign_open(mo, &mlen, sm, smlen, pk);
}
print_bench("sign_open (59 bytes)",t,NTIMINGS);
}
int main(int argc, char *argv[])
{
int ret = 0;
int i;
if( argc > 1 && !strncmp( argv[1], "--bench", 7 ) )
{
#if !defined(ARCH_X86) && !defined(ARCH_X86_64)
fprintf( stderr, "no --bench for your cpu until you port rdtsc\n" );
return 1;
#endif
do_bench = 1;
if( argv[1][7] == '=' )
{
bench_pattern = argv[1]+8;
bench_pattern_len = strlen(bench_pattern);
}
argc--;
argv++;
}
i = ( argc > 1 ) ? atoi(argv[1]) : x264_mdate();
fprintf( stderr, "x264: using random seed %u\n", i );
srand( i );
buf1 = x264_malloc( 0x3e00 + 16*BENCH_ALIGNS );
buf2 = buf1 + 0xf00;
buf3 = buf2 + 0xf00;
buf4 = buf3 + 0x1000;
for( i=0; i<0x1e00; i++ )
buf1[i] = rand() & 0xFF;
memset( buf1+0x1e00, 0, 0x2000 );
/* 16-byte alignment is guaranteed whenever it's useful, but some functions also vary in speed depending on %64 */
if( do_bench )
for( i=0; i<BENCH_ALIGNS && !ret; i++ )
{
buf2 = buf1 + 0xf00;
buf3 = buf2 + 0xf00;
buf4 = buf3 + 0x1000;
ret |= x264_stack_pagealign( check_all_flags, i*16 );
buf1 += 16;
quiet = 1;
fprintf( stderr, "%d/%d\r", i+1, BENCH_ALIGNS );
}
else
ret = check_all_flags();
if( ret )
{
fprintf( stderr, "x264: at least one test has failed. Go and fix that Right Now!\n" );
return -1;
}
fprintf( stderr, "x264: All tests passed Yeah :)\n" );
if( do_bench )
print_bench();
return 0;
}
void speed_nothing()
{
unsigned long long t[NTIMINGS];
int i;
// Benchmarking scalarmult_base
for(i=0;i<NTIMINGS;i++)
t[i] = cpucycles();
print_bench("nothing",t,NTIMINGS);
}
void speed_poly1305()
{
unsigned long long t[NTIMINGS];
unsigned char m[1024];
unsigned char out[16];
int i;
for(i=0;i<NTIMINGS;i++)
{
t[i] = cpucycles();
crypto_onetimeauth(out,m,8,key);
}
print_bench("poly1305 (8 bytes)",t,NTIMINGS);
for(i=0;i<NTIMINGS;i++)
{
t[i] = cpucycles();
crypto_onetimeauth(out,m,64,key);
}
print_bench("poly1305 (64 bytes)",t,NTIMINGS);
for(i=0;i<NTIMINGS;i++)
{
t[i] = cpucycles();
crypto_onetimeauth(out,m,576,key);
}
print_bench("poly1305 (576 bytes)",t,NTIMINGS);
for(i=0;i<NTIMINGS;i++)
{
t[i] = cpucycles();
crypto_onetimeauth(out,m,1024,key);
}
print_bench("poly1305 (1024 bytes)",t,NTIMINGS);
for(i=0;i<NTIMINGS;i++)
{
t[i] = cpucycles();
crypto_onetimeauth(out,m,2048,key);
}
print_bench("poly1305 (2048 bytes)",t,NTIMINGS);
}
void bench_allocators(Iter begin, std::size_t n)
{
// The different sets we will benchmark.
typedef Set<int, std::less<int>, std::allocator<int>> set_type1;
typedef Set<int, std::less<int>, rt::node_allocator<int>> set_type2; // Uses a vector as buffer.
typedef Set<int, std::less<int>, __gnu_cxx::__pool_alloc<int>> set_type3;
typedef Set<int, std::less<int>, __gnu_cxx::bitmap_allocator<int>> set_type4;
typedef Set<int, std::less<int>, __gnu_cxx::__mt_alloc<int>> set_type5;
#ifdef Boost_FOUND
typedef Set<int, std::less<int>, boost::container::node_allocator<int, 100000, 1>> set_type6;
typedef Set<int, std::less<int>, boost::container::node_allocator<int, 100000, 2>> set_type7;
#endif
std::cout << n << " ";
{ // (1)
set_type1 s;
print_bench(s, begin, n);
}
{ // (2)
std::vector<char> buffer((n + 2) * 40, 0);
rt::node_allocator<int> alloc(buffer);
set_type2 s(std::less<int>(), alloc); // Uses a vector as buffer.
print_bench(s, begin, n);
}
{ // (3)
set_type3 s;
print_bench(s, begin, n);
}
{ // (4)
set_type4 s;
print_bench(s, begin, n);
}
{ // (5)
set_type5 s;
print_bench(s, begin, n);
}
#ifdef Boost_FOUND
{ // (6)
set_type6 s;
print_bench(s, begin, n);
}
{ // (7)
set_type7 s;
print_bench(s, begin, n);
}
#endif
}
