Use different hash algorithms for strings and file hashing

For hash map, using fnv1a is faster as it is a much simpler algorithm
we can afford to inline. For files murmur3 should win as it processes
bytes 4 by 4.
This commit is contained in:
Maxime Coste 2024-02-06 21:57:17 +11:00
parent 53d9b9b676
commit 04a96b059f
6 changed files with 22 additions and 10 deletions

View File

@ -140,7 +140,7 @@ decltype(auto) parse_file(StringView filename, Func&& func)
const bool crlf = has_crlf and not has_lf; const bool crlf = has_crlf and not has_lf;
auto eolformat = crlf ? EolFormat::Crlf : EolFormat::Lf; auto eolformat = crlf ? EolFormat::Crlf : EolFormat::Lf;
FsStatus fs_status{file.st.st_mtim, file.st.st_size, hash_data(file.data, file.st.st_size)}; FsStatus fs_status{file.st.st_mtim, file.st.st_size, murmur3(file.data, file.st.st_size)};
return func(parse_lines(pos, end, eolformat), bom, eolformat, fs_status); return func(parse_lines(pos, end, eolformat), bom, eolformat, fs_status);
} }

View File

@ -394,7 +394,7 @@ void Client::check_if_buffer_needs_reloading()
return; return;
if (MappedFile fd{filename}; if (MappedFile fd{filename};
fd.st.st_size == status.file_size and hash_data(fd.data, fd.st.st_size) == status.hash) fd.st.st_size == status.file_size and murmur3(fd.data, fd.st.st_size) == status.hash)
return; return;
if (reload == Autoreload::Ask) if (reload == Autoreload::Ask)

View File

@ -626,7 +626,7 @@ FsStatus get_fs_status(StringView filename)
{ {
MappedFile fd{filename}; MappedFile fd{filename};
return {fd.st.st_mtim, fd.st.st_size, hash_data(fd.data, fd.st.st_size)}; return {fd.st.st_mtim, fd.st.st_size, murmur3(fd.data, fd.st.st_size)};
} }
String get_kak_binary_path() String get_kak_binary_path()

View File

@ -27,8 +27,8 @@ static inline uint32_t fmix(uint32_t h)
return h; return h;
} }
// murmur3 hash, based on https://github.com/PeterScott/murmur3 // based on https://github.com/PeterScott/murmur3
size_t hash_data(const char* input, size_t len) size_t murmur3(const char* input, size_t len)
{ {
const uint8_t* data = reinterpret_cast<const uint8_t*>(input); const uint8_t* data = reinterpret_cast<const uint8_t*>(input);
uint32_t hash = 0x1235678; uint32_t hash = 0x1235678;
@ -73,13 +73,13 @@ size_t hash_data(const char* input, size_t len)
UnitTest test_murmur_hash{[] { UnitTest test_murmur_hash{[] {
{ {
constexpr char data[] = "Hello, World!"; constexpr char data[] = "Hello, World!";
kak_assert(hash_data(data, strlen(data)) == 0xf816f95b); kak_assert(murmur3(data, strlen(data)) == 0xf816f95b);
} }
{ {
constexpr char data[] = "xxxxxxxxxxxxxxxxxxxxxxxxxxxx"; constexpr char data[] = "xxxxxxxxxxxxxxxxxxxxxxxxxxxx";
kak_assert(hash_data(data, strlen(data)) == 3551113186); kak_assert(murmur3(data, strlen(data)) == 3551113186);
} }
kak_assert(hash_data("", 0) == 2572747774); kak_assert(murmur3("", 0) == 2572747774);
}}; }};
} }

View File

@ -5,11 +5,23 @@
#include <utility> #include <utility>
#include <cstddef> #include <cstddef>
#include <cstdint>
namespace Kakoune namespace Kakoune
{ {
size_t hash_data(const char* data, size_t len); inline size_t fnv1a(const char* data, size_t len)
{
constexpr uint32_t FNV_prime_32 = 16777619;
constexpr uint32_t offset_basis_32 = 2166136261;
uint32_t hash_value = offset_basis_32;
for (size_t i = 0; i < len; ++i)
hash_value = (hash_value ^ data[i]) * FNV_prime_32;
return hash_value;
}
size_t murmur3(const char* input, size_t len);
template<typename Type> requires std::is_integral_v<Type> template<typename Type> requires std::is_integral_v<Type>
constexpr size_t hash_value(const Type& val) constexpr size_t hash_value(const Type& val)

View File

@ -21,7 +21,7 @@ public:
friend constexpr size_t hash_value(const Type& str) friend constexpr size_t hash_value(const Type& str)
{ {
return hash_data(str.data(), (int)str.length()); return fnv1a(str.data(), (int)str.length());
} }
using iterator = CharType*; using iterator = CharType*;