Enable fast floating point model and fast debug linking (#11466)

This commit enables /fp:fast. This doubles the performance of the Delta E
computation in #11095 for instance. Additionally it re-enables two options for
debug builds which are normally enabled by default by Visual Studio.

## PR Checklist
* [x] I work here
* [x] Tests added/passed

## Validation Steps Performed
* No change in binary size
* No obvious change in behavior

.github/actions/spelling/allow/math.txt

@@ -8,3 +8,4 @@ LStep
 powf
 RSub
 sqrtf
+ULP

src/buffer/out/TextColor.cpp

@@ -4,6 +4,8 @@
 #include "precomp.h"
 #include "TextColor.h"
 
+#include <til/bit.h>
+
 // clang-format off
 
 // A table mapping 8-bit RGB colors, in the form RRRGGGBB,
@@ -186,7 +188,7 @@ COLORREF TextColor::GetColor(const std::array<COLORREF, 256>& colorTable, const
             //    the result will be something like 0b00100000.
             // 5. Use BitScanForward (bsf) to find the index of the most significant 1 bit.
             const auto haystack = _mm256_loadu_si256(reinterpret_cast<const __m256i*>(colorTable.data())); // 1.
-            const auto needle = _mm256_set1_epi32(__builtin_bit_cast(int, defaultColor)); // 2.
+            const auto needle = _mm256_set1_epi32(til::bit_cast<int>(defaultColor)); // 2.
             const auto result = _mm256_cmpeq_epi32(haystack, needle); // 3.
             const auto mask = _mm256_movemask_ps(_mm256_castsi256_ps(result)); // 4.
             unsigned long index;
@@ -203,7 +205,7 @@ COLORREF TextColor::GetColor(const std::array<COLORREF, 256>& colorTable, const
             //   --> the index returned by _BitScanForward must be divided by 2.
             const auto haystack1 = _mm_loadu_si128(reinterpret_cast<const __m128i*>(colorTable.data() + 0));
             const auto haystack2 = _mm_loadu_si128(reinterpret_cast<const __m128i*>(colorTable.data() + 4));
-            const auto needle = _mm_set1_epi32(__builtin_bit_cast(int, defaultColor));
+            const auto needle = _mm_set1_epi32(til::bit_cast<int>(defaultColor));
             const auto result1 = _mm_cmpeq_epi32(haystack1, needle);
             const auto result2 = _mm_cmpeq_epi32(haystack2, needle);
             const auto result = _mm_packs_epi32(result1, result2); // 3.5

src/cascadia/ut_app/ColorHelperTests.cpp

@@ -6,6 +6,9 @@
 #include "../TerminalSettingsModel/Profile.h"
 #include "../TerminalApp/ColorHelper.h"
 
+// Import some templates to compare floats using approximate matching.
+#include <consoletaeftemplates.hpp>
+
 using namespace Microsoft::Console;
 using namespace winrt::TerminalApp;
 using namespace WEX::Logging;

src/common.build.pre.props

@@ -119,6 +119,7 @@
       <LanguageStandard>stdcpp17</LanguageStandard>
       <AdditionalOptions>/utf-8 %(AdditionalOptions)</AdditionalOptions>
       <ControlFlowGuard>Guard</ControlFlowGuard>
+      <FloatingPointModel>Fast</FloatingPointModel>
     </ClCompile>
     <ResourceCompile>
       <PreprocessorDefinitions>EXTERNAL_BUILD;_UNICODE;UNICODE;%(PreprocessorDefinitions)</PreprocessorDefinitions>
@@ -145,6 +146,10 @@
       <Optimization>Disabled</Optimization>
       <PreprocessorDefinitions>_DEBUG;DBG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
     </ClCompile>
+    <Link>
+      <SetChecksum>false</SetChecksum>
+      <GenerateDebugInformation>DebugFastLink</GenerateDebugInformation>
+    </Link>
   </ItemDefinitionGroup>
 
   <!-- For Release ONLY -->

src/inc/consoletaeftemplates.hpp

@@ -17,6 +17,8 @@ Revision History:
 
 #pragma once
 
+#include <til/bit.h>
+
 // Helper for declaring a variable to store a TEST_METHOD_PROPERTY and get it's value from the test metadata
 #define INIT_TEST_PROPERTY(type, identifer, description) \
     type identifer;                                      \
@@ -43,6 +45,45 @@ Revision History:
 
 namespace WEX::TestExecution
 {
+    // Compare two floats using a ULP (unit last place) tolerance of up to 4.
+    // Allows you to compare two floats that are almost equal.
+    // Think of: 0.200000000000000 vs. 0.200000000000001.
+    template<typename T, typename U>
+    bool CompareFloats(T a, T b) noexcept
+    {
+        if (std::isnan(a))
+        {
+            return std::isnan(b);
+        }
+
+        if (a == b)
+        {
+            return true;
+        }
+
+        const auto nDiff = static_cast<std::make_signed_t<U>>(til::bit_cast<U>(a) - til::bit_cast<U>(b));
+        const auto uDiff = static_cast<U>(nDiff < 0 ? -nDiff : nDiff);
+        return uDiff <= 4;
+    }
+
+    template<>
+    struct VerifyCompareTraits<float, float>
+    {
+        static bool AreEqual(float a, float b) noexcept
+        {
+            return CompareFloats<float, uint32_t>(a, b);
+        }
+    };
+
+    template<>
+    struct VerifyCompareTraits<double, double>
+    {
+        static bool AreEqual(double a, double b) noexcept
+        {
+            return CompareFloats<double, uint64_t>(a, b);
+        }
+    };
+
     template<>
     class VerifyOutputTraits<SMALL_RECT>
     {

src/inc/til/bit.h

@@ -0,0 +1,16 @@
+// Copyright (c) Microsoft Corporation.
+// Licensed under the MIT license.
+
+#pragma once
+
+namespace til
+{
+    template<class To, class From, std::enable_if_t<std::conjunction_v<std::bool_constant<sizeof(To) == sizeof(From)>, std::is_trivially_copyable<To>, std::is_trivially_copyable<From>>, int> = 0>
+    [[nodiscard]] constexpr To bit_cast(const From& _Val) noexcept
+    {
+#ifdef __cpp_lib_bit_cast
+#warning "Replace til::bit_cast and __builtin_bit_cast with std::bit_cast"
+#endif
+        return __builtin_bit_cast(To, _Val);
+    }
+}

src/project.inc

@@ -29,7 +29,7 @@ USE_NATIVE_EH           = 1
 # -------------------------------------
 
 MSC_WARNING_LEVEL       = /W4 /WX
-USER_C_FLAGS            = $(USER_C_FLAGS) /utf-8
+USER_C_FLAGS            = $(USER_C_FLAGS) /fp:fast /utf-8
 
 # -------------------------------------
 # Common Console Includes and Libraries