From d33395ac1aab90686951a5aafdfaf3a497c8a1de Mon Sep 17 00:00:00 2001
From: HedgehogInTheCPP <a.rainman@gmail.com>
Date: Fri, 14 Aug 2026 01:33:14 +0300
Subject: [PATCH] cleanup.

---
 include/fast_float/ascii_number.h | 16 +---------------
 include/fast_float/float_common.h | 19 +++++++++++++------
 2 files changed, 14 insertions(+), 21 deletions(-)

diff --git a/include/fast_float/ascii_number.h b/include/fast_float/ascii_number.h
index c7c5cfb..962a403 100644
--- a/include/fast_float/ascii_number.h
+++ b/include/fast_float/ascii_number.h
@@ -11,21 +11,7 @@
 #include "float_common.h"
 
 #if defined(FASTFLOAT_X86_SIMD)
-#include <emmintrin.h>
-#ifdef FASTFLOAT_SSE_PATCH_REWORK
-#if FASTFLOAT_X86_SIMD >= 30
-#include <pmmintrin.h>
-#endif
-#if FASTFLOAT_X86_SIMD >= 31
-#include <tmmintrin.h>
-#endif
-#if FASTFLOAT_X86_SIMD >= 41
-#include <smmintrin.h>
-#endif
-#if FASTFLOAT_X86_SIMD >= 42
-#include <nmmintrin.h>
-#endif
-#endif
+#include <immintrin.h>
 #elif defined(FASTFLOAT_ARM_NEON)
 #include <arm_neon.h>
 #endif
diff --git a/include/fast_float/float_common.h b/include/fast_float/float_common.h
index 8bbee8b..236ce7b 100644
--- a/include/fast_float/float_common.h
+++ b/include/fast_float/float_common.h
@@ -193,10 +193,15 @@ using parse_options = parse_options_t<char>;
 #endif
 
 #ifdef FASTFLOAT_X86_SIMD // user defined level
-static_assert(FASTFLOAT_X86_SIMD >= 20 && FASTFLOAT_X86_SIMD <= 42,
-                "FASTFLOAT_X86_SIMD should be between 20(SSE2) and 42(SSE4.2)");
+static_assert(FASTFLOAT_X86_SIMD >= 20 && FASTFLOAT_X86_SIMD <= 52,
+              "FASTFLOAT_X86_SIMD should be between 20(SSE2) and 42(SSE4.2) "
+              "and optionally up to 52(AVX2)");
 #else // auto detect
-#if defined(__SSE4_2__)
+#if defined(__AVX2__)
+#define FASTFLOAT_X86_SIMD 52
+#elif defined(__AVX__) // On MSVC there's no way to check for SSE4.2 specifically so check __AVX__.
+#define FASTFLOAT_X86_SIMD 50
+#elif defined(__SSE4_2__)
 #define FASTFLOAT_X86_SIMD 42
 #elif defined(__SSE4_1__)
 #define FASTFLOAT_X86_SIMD 41
@@ -517,7 +522,8 @@ template <typename T> struct span {
   }
 };
 
-struct alignas(16) value128 {
+/* alignas(16) - better data cache usage wo align */
+struct value128 {
   uint64_t low;
   uint64_t high;
 
@@ -568,7 +574,7 @@ leading_zeroes(uint64_t input_num) noexcept {
     return leading_zeroes_generic(input_num);
   }
 #ifdef FASTFLOAT_VISUAL_STUDIO
-#if defined(__AVX2__)
+#if FASTFLOAT_X86_SIMD >= 52
   // use lzcnt on MSVC only on AVX2 capable CPU's that all have this BMI
   // instruction
   return __lzcnt64(input_num);
@@ -696,7 +702,8 @@ full_multiplication(uint64_t a, uint64_t b) noexcept {
   return answer;
 }
 
-struct /* alignas(16) */ adjusted_mantissa {
+/* alignas(16) - better data cache usage witout align */
+struct adjusted_mantissa {
   am_mant_t mantissa;
   am_pow_t power2;