Merge branch 'main' of https://github.com/fastfloat/fast_float

.github/workflows/s390x.yml

@@ -22,9 +22,9 @@ jobs:
           distro: ubuntu_latest
           install: |
             apt-get update -q -y
-            apt-get install -y cmake make g++
+            apt-get install -y cmake make g++ git
           run: |
-            cmake -DCMAKE_BUILD_TYPE=Release -B build
+            cmake -DCMAKE_BUILD_TYPE=Release -B build  -DFASTFLOAT_TEST=ON
             cmake --build build -j=2
             ctest --output-on-failure --test-dir build
 

CMakeLists.txt

@@ -1,7 +1,7 @@
 cmake_minimum_required(VERSION 3.14)
 
 
-project(fast_float VERSION 8.2.1 LANGUAGES CXX)
+project(fast_float VERSION 8.2.2 LANGUAGES CXX)
 set(FASTFLOAT_CXX_STANDARD 11 CACHE STRING "the C++ standard to use for fastfloat")
 set(CMAKE_CXX_STANDARD ${FASTFLOAT_CXX_STANDARD})
 option(FASTFLOAT_TEST "Enable tests" OFF)

README.md

@@ -569,7 +569,7 @@ sufficiently recent version of CMake (3.11 or better at least):
 FetchContent_Declare(
   fast_float
   GIT_REPOSITORY https://github.com/fastfloat/fast_float.git
-  GIT_TAG tags/v8.2.1
+  GIT_TAG tags/v8.2.2
   GIT_SHALLOW TRUE)
 
 FetchContent_MakeAvailable(fast_float)
@@ -585,7 +585,7 @@ You may also use [CPM](https://github.com/cpm-cmake/CPM.cmake), like so:
 CPMAddPackage(
   NAME fast_float
   GITHUB_REPOSITORY "fastfloat/fast_float"
-  GIT_TAG v8.2.1)
+  GIT_TAG v8.2.2)
 ```
 
 ## Using as single header
@@ -597,7 +597,7 @@ if desired as described in the command line help.
 
 You may directly download automatically generated single-header files:
 
-<https://github.com/fastfloat/fast_float/releases/download/v8.2.1/fast_float.h>
+<https://github.com/fastfloat/fast_float/releases/download/v8.2.2/fast_float.h>
 
 ## Benchmarking
 

benchmarks/CMakeLists.txt

@@ -11,7 +11,9 @@ FetchContent_MakeAvailable(counters)
 add_executable(realbenchmark benchmark.cpp)
 target_link_libraries(realbenchmark PRIVATE counters::counters)
 add_executable(bench_ip bench_ip.cpp)
+add_executable(bench_uint16 bench_uint16.cpp)
 target_link_libraries(bench_ip PRIVATE counters::counters)
+target_link_libraries(bench_uint16 PRIVATE counters::counters)
 
 set_property(
     TARGET realbenchmark
@@ -19,8 +21,12 @@ set_property(
 set_property(
     TARGET bench_ip
     PROPERTY CXX_STANDARD 17)
+set_property(
+    TARGET bench_uint16
+    PROPERTY CXX_STANDARD 17)
 target_link_libraries(realbenchmark PUBLIC fast_float)
 target_link_libraries(bench_ip PUBLIC fast_float)
+target_link_libraries(bench_uint16 PUBLIC fast_float)
 
 include(ExternalProject)
 

benchmarks/bench_uint16.cpp

@@ -0,0 +1,139 @@
+#include "counters/bench.h"
+#include "fast_float/fast_float.h"
+#include <charconv>
+#include <cstdint>
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+#include <random>
+#include <atomic>
+#include <string>
+#include <vector>
+
+void pretty_print(size_t volume, size_t bytes, std::string name,
+                  counters::event_aggregate agg) {
+  if (agg.inner_count > 1) {
+    printf("# (inner count: %d)\n", agg.inner_count);
+  }
+  printf("%-40s : ", name.c_str());
+  printf(" %5.2f GB/s ", bytes / agg.fastest_elapsed_ns());
+  printf(" %5.1f Mip/s ", volume * 1000.0 / agg.fastest_elapsed_ns());
+  printf(" %5.2f ns/ip ", agg.fastest_elapsed_ns() / volume);
+  if (counters::event_collector().has_events()) {
+    printf(" %5.2f GHz ", agg.fastest_cycles() / agg.fastest_elapsed_ns());
+    printf(" %5.2f c/ip ", agg.fastest_cycles() / volume);
+    printf(" %5.2f i/ip ", agg.fastest_instructions() / volume);
+    printf(" %5.2f c/b ", agg.fastest_cycles() / bytes);
+    printf(" %5.2f i/b ", agg.fastest_instructions() / bytes);
+    printf(" %5.2f i/c ", agg.fastest_instructions() / agg.fastest_cycles());
+  }
+  printf("\n");
+}
+
+enum class parse_method { standard, fast_float };
+
+void validate(const std::string &buffer, const std::vector<uint16_t> &expected,
+              char delimiter) {
+  const char *p = buffer.data();
+  const char *pend = p + buffer.size();
+
+  for (size_t i = 0; i < expected.size(); i++) {
+    uint16_t val;
+    auto r = fast_float::from_chars(p, pend, val);
+    if (r.ec != std::errc() || val != expected[i]) {
+      printf("Validation failed at index %zu: expected %u, got %u\n", i,
+             expected[i], val);
+      std::abort();
+    }
+    p = r.ptr;
+    if (i + 1 < expected.size()) {
+      if (p >= pend || *p != delimiter) {
+        printf("Validation failed at index %zu: delimiter mismatch\n", i);
+        std::abort();
+      }
+      ++p;
+    }
+  }
+
+  if (p != pend) {
+    printf("Validation failed: trailing bytes remain\n");
+    std::abort();
+  }
+  printf("Validation passed!\n");
+}
+
+int main() {
+  constexpr size_t N = 500000;
+  constexpr char delimiter = ',';
+  std::mt19937 rng(1234);
+  std::uniform_int_distribution<int> dist(0, 65535);
+
+  std::vector<uint16_t> expected;
+  expected.reserve(N);
+
+  std::string buffer;
+  buffer.reserve(N * 6); // up to 5 digits + delimiter
+
+  for (size_t i = 0; i < N; ++i) {
+    uint16_t val = (uint16_t)dist(rng);
+    expected.push_back(val);
+    std::string s = std::to_string(val);
+    buffer.append(s);
+    if (i + 1 < N) {
+      buffer.push_back(delimiter);
+    }
+  }
+
+  size_t total_bytes = buffer.size();
+
+  validate(buffer, expected, delimiter);
+
+  volatile uint64_t sink = 0;
+
+  pretty_print(N, total_bytes, "parse_uint16_std_fromchars",
+               counters::bench([&]() {
+                 uint64_t sum = 0;
+                 const char *p = buffer.data();
+                 const char *pend = p + buffer.size();
+                 for (size_t i = 0; i < N; ++i) {
+                   uint16_t value = 0;
+                   auto r = std::from_chars(p, pend, value);
+                   if (r.ec != std::errc())
+                     std::abort();
+                   sum += value;
+                   p = r.ptr;
+                   if (i + 1 < N) {
+                     if (p >= pend || *p != delimiter)
+                       std::abort();
+                     ++p;
+                   }
+                 }
+                 if (p != pend)
+                   std::abort();
+                 sink += sum;
+               }));
+
+  pretty_print(N, total_bytes, "parse_uint16_fastfloat", counters::bench([&]() {
+                 uint64_t sum = 0;
+                 const char *p = buffer.data();
+                 const char *pend = p + buffer.size();
+                 for (size_t i = 0; i < N; ++i) {
+                   uint16_t value = 0;
+                   auto r = fast_float::from_chars(p, pend, value);
+                   if (r.ec != std::errc())
+                     std::abort();
+                   sum += value;
+                   p = r.ptr;
+                   if (i + 1 < N) {
+                     if (p >= pend || *p != delimiter)
+                       std::abort();
+                     ++p;
+                   }
+                 }
+                 if (p != pend)
+                   std::abort();
+                 sink += sum;
+               }));
+
+  return EXIT_SUCCESS;
+}

include/fast_float/ascii_number.h

@@ -31,7 +31,7 @@ fastfloat_really_inline constexpr bool has_simd_opt() noexcept {
 // able to optimize it well.
 template <typename UC>
 fastfloat_really_inline constexpr bool is_integer(UC c) noexcept {
-  return !(c > UC('9') || c < UC('0'));
+  return (unsigned)(c - UC('0')) <= 9u;
 }
 
 fastfloat_really_inline constexpr uint64_t byteswap(uint64_t val) noexcept {
@@ -41,6 +41,11 @@ fastfloat_really_inline constexpr uint64_t byteswap(uint64_t val) noexcept {
          (val & 0x000000000000FF00) << 40 | (val & 0x00000000000000FF) << 56;
 }
 
+fastfloat_really_inline constexpr uint32_t byteswap_32(uint32_t val) {
+  return (val >> 24) | ((val >> 8) & 0x0000FF00u) | ((val << 8) & 0x00FF0000u) |
+         (val << 24);
+}
+
 // Read 8 UC into a u64. Truncates UC if not char.
 template <typename UC>
 fastfloat_really_inline FASTFLOAT_CONSTEXPR20 uint64_t
@@ -62,6 +67,25 @@ read8_to_u64(UC const *chars) noexcept {
   return val;
 }
 
+// Read 4 UC into a u32. Truncates UC if not char.
+template <typename UC>
+fastfloat_really_inline FASTFLOAT_CONSTEXPR20 uint32_t
+read4_to_u32(UC const *chars) {
+  if (cpp20_and_in_constexpr() || !std::is_same<UC, char>::value) {
+    uint32_t val = 0;
+    for (int i = 0; i < 4; ++i) {
+      val |= uint32_t(uint8_t(*chars)) << (i * 8);
+      ++chars;
+    }
+    return val;
+  }
+  uint32_t val;
+  ::memcpy(&val, chars, sizeof(uint32_t));
+#if FASTFLOAT_IS_BIG_ENDIAN == 1
+  val = byteswap_32(val);
+#endif
+  return val;
+}
 #ifdef FASTFLOAT_SSE2
 
 fastfloat_really_inline uint64_t simd_read8_to_u64(__m128i const &data) {
@@ -140,6 +164,18 @@ is_made_of_eight_digits_fast(uint64_t val) noexcept {
             0x8080808080808080));
 }
 
+fastfloat_really_inline constexpr bool
+is_made_of_four_digits_fast(uint32_t val) noexcept {
+  return !((((val + 0x46464646) | (val - 0x30303030)) & 0x80808080));
+}
+
+fastfloat_really_inline FASTFLOAT_CONSTEXPR14 uint32_t
+parse_four_digits_unrolled(uint32_t val) noexcept {
+  val -= 0x30303030;
+  val = (val * 10) + (val >> 8);
+  return (((val & 0x00FF00FF) * 0x00640001) >> 16) & 0xFFFF;
+}
+
 #ifdef FASTFLOAT_HAS_SIMD
 
 // Call this if chars might not be 8 digits.
@@ -558,97 +594,145 @@ parse_int_string(UC const *p, UC const *pend, T &value,
   auto const *const start_digits = p;
 
   FASTFLOAT_IF_CONSTEXPR17((std::is_same<T, std::uint8_t>::value)) {
-    const auto len = static_cast<am_digits>(pend - p);
-    if (len == 0) {
-      if (has_leading_zeros) {
-        value = 0;
-        answer.ec = std::errc();
-        answer.ptr = p;
-      } else {
-        answer.ec = std::errc::invalid_argument;
-        answer.ptr = first;
-      }
-      return answer;
-    }
-
-    uint32_t digits;
-
-#if FASTFLOAT_HAS_IS_CONSTANT_EVALUATED && FASTFLOAT_HAS_BIT_CAST
-    if (std::is_constant_evaluated()) {
-      uint8_t str[4];
-      for (uint_fast8_t j = 0; j != 4 && j != len; ++j) {
-        str[j] = static_cast<uint8_t>(p[j]);
-      }
-      digits = bit_cast<uint32_t>(str);
-#if FASTFLOAT_IS_BIG_ENDIAN
-      digits = byteswap(digits);
-#endif
-    }
-#else
-    if (false) {
-    }
-#endif
-    else if (len >= 4) {
-      std::memcpy(&digits, p, 4);
-#if FASTFLOAT_IS_BIG_ENDIAN
-      digits = byteswap(digits);
-#endif
-    } else {
-      uint32_t b0 = static_cast<uint8_t>(p[0]);
-      uint32_t b1 = (len > 1) ? static_cast<uint8_t>(p[1]) : 0xFFu;
-      uint32_t b2 = (len > 2) ? static_cast<uint8_t>(p[2]) : 0xFFu;
-      uint32_t b3 = 0xFFu;
-#if FASTFLOAT_IS_BIG_ENDIAN
-      digits = (b0 << 24) | (b1 << 16) | (b2 << 8) | b3;
-#else
-      digits = b0 | (b1 << 8) | (b2 << 16) | (b3 << 24);
-#endif
-    }
-
-    const uint32_t magic =
-        ((digits + 0x46464646u) | (digits - 0x30303030u)) & 0x80808080u;
-    const auto tz = countr_zero_32(magic); // 7, 15, 23, 31, or 32
-    am_digits nd = (tz >> 3);
-    nd = std::min(nd, len);
-    if (nd == 0) {
-      if (has_leading_zeros) {
-        value = 0;
-        answer.ec = std::errc();
-        answer.ptr = p;
+    if (options.base == 10) {
+      const auto len = static_cast<am_digits>(pend - p);
+      if (len == 0) {
+        if (has_leading_zeros) {
+          value = 0;
+          answer.ec = std::errc();
+          answer.ptr = p;
+        } else {
+          answer.ec = std::errc::invalid_argument;
+          answer.ptr = first;
+        }
         return answer;
       }
-      answer.ec = std::errc::invalid_argument;
-      answer.ptr = first;
-      return answer;
-    }
-    if (nd > 3) {
-      const UC *q = p + nd;
-      am_digits rem = len - nd;
-      while (rem) {
-        if (*q < UC('0') || *q > UC('9'))
-          break;
-        ++q;
-        --rem;
+
+      uint32_t digits;
+
+#if FASTFLOAT_HAS_IS_CONSTANT_EVALUATED && FASTFLOAT_HAS_BIT_CAST
+      if (std::is_constant_evaluated()) {
+        uint8_t str[4];
+        for (uint_fast8_t j = 0; j != 4 && j != len; ++j) {
+          str[j] = static_cast<uint8_t>(p[j]);
+        }
+        digits = bit_cast<uint32_t>(str);
+#if FASTFLOAT_IS_BIG_ENDIAN
+        digits = byteswap_32(digits);
+#endif
+      }
+#else
+      if (false) {
+      }
+#endif
+      else if (len >= 4) {
+        ::memcpy(&digits, p, 4);
+#if FASTFLOAT_IS_BIG_ENDIAN
+        digits = byteswap_32(digits);
+#endif
+      } else {
+        uint32_t b0 = static_cast<uint8_t>(p[0]);
+        uint32_t b1 = (len > 1) ? static_cast<uint8_t>(p[1]) : 0xFFu;
+        uint32_t b2 = (len > 2) ? static_cast<uint8_t>(p[2]) : 0xFFu;
+        uint32_t b3 = 0xFFu;
+        digits = b0 | (b1 << 8) | (b2 << 16) | (b3 << 24);
       }
-      answer.ec = std::errc::result_out_of_range;
-      answer.ptr = q;
-      return answer;
-    }
 
-    digits ^= 0x30303030u;
-    digits <<= ((4 - nd) * 8);
+      uint32_t magic =
+          ((digits + 0x46464646u) | (digits - 0x30303030u)) & 0x80808080u;
+      uint32_t tz = (uint32_t)countr_zero_32(magic); // 7, 15, 23, 31, or 32
+      uint32_t nd = (tz == 32) ? 4 : (tz >> 3);
+      nd = (uint32_t)std::min((size_t)nd, len);
+      if (nd == 0) {
+        if (has_leading_zeros) {
+          value = 0;
+          answer.ec = std::errc();
+          answer.ptr = p;
+          return answer;
+        }
+        answer.ec = std::errc::invalid_argument;
+        answer.ptr = first;
+        return answer;
+      }
+      if (nd > 3) {
+        const UC *q = p + nd;
+        size_t rem = len - nd;
+        while (rem) {
+          if (*q < UC('0') || *q > UC('9'))
+            break;
+          ++q;
+          --rem;
+        }
+        answer.ec = std::errc::result_out_of_range;
+        answer.ptr = q;
+        return answer;
+      }
 
-    const uint32_t check = ((digits >> 24) & 0xff) | ((digits >> 8) & 0xff00) |
-                           ((digits << 8) & 0xff0000);
-    if (check > 0x00020505) {
-      answer.ec = std::errc::result_out_of_range;
+      digits ^= 0x30303030u;
+      digits <<= ((4 - nd) * 8);
+
+      uint32_t check = ((digits >> 24) & 0xff) | ((digits >> 8) & 0xff00) |
+                       ((digits << 8) & 0xff0000);
+      if (check > 0x00020505) {
+        answer.ec = std::errc::result_out_of_range;
+        answer.ptr = p + nd;
+        return answer;
+      }
+      value = (uint8_t)((0x640a01 * digits) >> 24);
+      answer.ec = std::errc();
       answer.ptr = p + nd;
       return answer;
     }
-    value = static_cast<uint8_t>((0x640a01 * digits) >> 24);
-    answer.ec = std::errc();
-    answer.ptr = p + nd;
-    return answer;
+  }
+
+  FASTFLOAT_IF_CONSTEXPR17((std::is_same<T, std::uint16_t>::value)) {
+    if (options.base == 10) {
+      const size_t len = size_t(pend - p);
+      if (len == 0) {
+        if (has_leading_zeros) {
+          value = 0;
+          answer.ec = std::errc();
+          answer.ptr = p;
+        } else {
+          answer.ec = std::errc::invalid_argument;
+          answer.ptr = first;
+        }
+        return answer;
+      }
+
+      if (len >= 4) {
+        uint32_t digits = read4_to_u32(p);
+        if (is_made_of_four_digits_fast(digits)) {
+          uint32_t v = parse_four_digits_unrolled(digits);
+          if (len >= 5 && is_integer(p[4])) {
+            v = v * 10 + uint32_t(p[4] - '0');
+            if (len >= 6 && is_integer(p[5])) {
+              answer.ec = std::errc::result_out_of_range;
+              const UC *q = p + 5;
+              while (q != pend && is_integer(*q)) {
+                q++;
+              }
+              answer.ptr = q;
+              return answer;
+            }
+            if (v > 65535) {
+              answer.ec = std::errc::result_out_of_range;
+              answer.ptr = p + 5;
+              return answer;
+            }
+            value = uint16_t(v);
+            answer.ec = std::errc();
+            answer.ptr = p + 5;
+            return answer;
+          }
+          // 4 digits
+          value = uint16_t(v);
+          answer.ec = std::errc();
+          answer.ptr = p + 4;
+          return answer;
+        }
+      }
+    }
   }
 
   // Parse digits

include/fast_float/float_common.h

@@ -17,7 +17,7 @@
 
 #define FASTFLOAT_VERSION_MAJOR 8
 #define FASTFLOAT_VERSION_MINOR 2
-#define FASTFLOAT_VERSION_PATCH 1
+#define FASTFLOAT_VERSION_PATCH 2
 
 #define FASTFLOAT_STRINGIZE_IMPL(x) #x
 #define FASTFLOAT_STRINGIZE(x) FASTFLOAT_STRINGIZE_IMPL(x)