Make PopCount and CountTrailingZeros more standard-like

12 files changed, 145 insertions, 113 deletions

diff --git a/al/auxeffectslot.cpp b/al/auxeffectslot.cpp
index 4af2acf2..8b31ab80 100644
--- a/al/auxeffectslot.cpp
+++ b/al/auxeffectslot.cpp
@@ -35,6 +35,7 @@
 #include "AL/alc.h"
 #include "AL/efx.h"
 
+#include "albit.h"
 #include "alcmain.h"
 #include "alcontext.h"
 #include "almalloc.h"
@@ -240,7 +241,7 @@ bool EnsureEffectSlots(ALCcontext *context, size_t needed)
     size_t count{std::accumulate(context->mEffectSlotList.cbegin(),
         context->mEffectSlotList.cend(), size_t{0},
         [](size_t cur, const EffectSlotSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<ALuint>(PopCount(sublist.FreeMask)); })};
+        { return cur + static_cast<ALuint>(al::popcount(sublist.FreeMask)); })};
 
     while(needed > count)
     {
@@ -268,7 +269,7 @@ ALeffectslot *AllocEffectSlot(ALCcontext *context)
         [](const EffectSlotSubList &entry) noexcept -> bool
         { return entry.FreeMask != 0; });
     auto lidx = static_cast<ALuint>(std::distance(context->mEffectSlotList.begin(), sublist));
-    auto slidx = static_cast<ALuint>(CountTrailingZeros(sublist->FreeMask));
+    auto slidx = static_cast<ALuint>(al::countr_zero(sublist->FreeMask));
 
     ALeffectslot *slot{::new(sublist->EffectSlots + slidx) ALeffectslot{}};
     aluInitEffectPanning(&slot->mSlot, context);
@@ -991,7 +992,7 @@ void UpdateAllEffectSlotProps(ALCcontext *context)
         uint64_t usemask{~sublist.FreeMask};
         while(usemask)
         {
-            const int idx{CountTrailingZeros(usemask)};
+            const int idx{al::countr_zero(usemask)};
             ALeffectslot *slot{sublist.EffectSlots + idx};
             usemask &= ~(1_u64 << idx);
 
@@ -1007,7 +1008,7 @@ EffectSlotSubList::~EffectSlotSubList()
     uint64_t usemask{~FreeMask};
     while(usemask)
     {
-        const ALsizei idx{CountTrailingZeros(usemask)};
+        const int idx{al::countr_zero(usemask)};
         al::destroy_at(EffectSlots+idx);
         usemask &= ~(1_u64 << idx);
     }
diff --git a/al/buffer.cpp b/al/buffer.cpp
index cdbedf70..14fae1ee 100644
--- a/al/buffer.cpp
+++ b/al/buffer.cpp
@@ -41,6 +41,7 @@
 #include "AL/alc.h"
 #include "AL/alext.h"
 
+#include "albit.h"
 #include "albyte.h"
 #include "alcmain.h"
 #include "alcontext.h"
@@ -326,7 +327,7 @@ bool EnsureBuffers(ALCdevice *device, size_t needed)
 {
     size_t count{std::accumulate(device->BufferList.cbegin(), device->BufferList.cend(), size_t{0},
         [](size_t cur, const BufferSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<ALuint>(PopCount(sublist.FreeMask)); })};
+        { return cur + static_cast<ALuint>(al::popcount(sublist.FreeMask)); })};
 
     while(needed > count)
     {
@@ -355,7 +356,7 @@ ALbuffer *AllocBuffer(ALCdevice *device)
     );
 
     auto lidx = static_cast<ALuint>(std::distance(device->BufferList.begin(), sublist));
-    auto slidx = static_cast<ALuint>(CountTrailingZeros(sublist->FreeMask));
+    auto slidx = static_cast<ALuint>(al::countr_zero(sublist->FreeMask));
 
     ALbuffer *buffer{::new (sublist->Buffers + slidx) ALbuffer{}};
 
@@ -1621,7 +1622,7 @@ BufferSubList::~BufferSubList()
     uint64_t usemask{~FreeMask};
     while(usemask)
     {
-        const ALsizei idx{CountTrailingZeros(usemask)};
+        const int idx{al::countr_zero(usemask)};
         al::destroy_at(Buffers+idx);
         usemask &= ~(1_u64 << idx);
     }
diff --git a/al/effect.cpp b/al/effect.cpp
index 02b31fac..93aa5547 100644
--- a/al/effect.cpp
+++ b/al/effect.cpp
@@ -38,6 +38,7 @@
 #include "AL/efx-presets.h"
 #include "AL/efx.h"
 
+#include "albit.h"
 #include "alcmain.h"
 #include "alcontext.h"
 #include "almalloc.h"
@@ -155,7 +156,7 @@ bool EnsureEffects(ALCdevice *device, size_t needed)
 {
     size_t count{std::accumulate(device->EffectList.cbegin(), device->EffectList.cend(), size_t{0},
         [](size_t cur, const EffectSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<ALuint>(PopCount(sublist.FreeMask)); })};
+        { return cur + static_cast<ALuint>(al::popcount(sublist.FreeMask)); })};
 
     while(needed > count)
     {
@@ -183,7 +184,7 @@ ALeffect *AllocEffect(ALCdevice *device)
         { return entry.FreeMask != 0; }
     );
     auto lidx = static_cast<ALuint>(std::distance(device->EffectList.begin(), sublist));
-    auto slidx = static_cast<ALuint>(CountTrailingZeros(sublist->FreeMask));
+    auto slidx = static_cast<ALuint>(al::countr_zero(sublist->FreeMask));
 
     ALeffect *effect{::new (sublist->Effects + slidx) ALeffect{}};
     InitEffectParams(effect, AL_EFFECT_NULL);
@@ -543,7 +544,7 @@ EffectSubList::~EffectSubList()
     uint64_t usemask{~FreeMask};
     while(usemask)
     {
-        const ALsizei idx{CountTrailingZeros(usemask)};
+        const int idx{al::countr_zero(usemask)};
         al::destroy_at(Effects+idx);
         usemask &= ~(1_u64 << idx);
     }
diff --git a/al/filter.cpp b/al/filter.cpp
index e549b1c9..0bcfe408 100644
--- a/al/filter.cpp
+++ b/al/filter.cpp
@@ -36,6 +36,7 @@
 #include "AL/alc.h"
 #include "AL/efx.h"
 
+#include "albit.h"
 #include "alcmain.h"
 #include "alcontext.h"
 #include "almalloc.h"
@@ -324,7 +325,7 @@ bool EnsureFilters(ALCdevice *device, size_t needed)
 {
     size_t count{std::accumulate(device->FilterList.cbegin(), device->FilterList.cend(), size_t{0},
         [](size_t cur, const FilterSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<ALuint>(PopCount(sublist.FreeMask)); })};
+        { return cur + static_cast<ALuint>(al::popcount(sublist.FreeMask)); })};
 
     while(needed > count)
     {
@@ -353,7 +354,7 @@ ALfilter *AllocFilter(ALCdevice *device)
         { return entry.FreeMask != 0; }
     );
     auto lidx = static_cast<ALuint>(std::distance(device->FilterList.begin(), sublist));
-    auto slidx = static_cast<ALuint>(CountTrailingZeros(sublist->FreeMask));
+    auto slidx = static_cast<ALuint>(al::countr_zero(sublist->FreeMask));
 
     ALfilter *filter{::new(sublist->Filters + slidx) ALfilter{}};
     InitFilterParams(filter, AL_FILTER_NULL);
@@ -704,7 +705,7 @@ FilterSubList::~FilterSubList()
     uint64_t usemask{~FreeMask};
     while(usemask)
     {
-        const ALsizei idx{CountTrailingZeros(usemask)};
+        const int idx{al::countr_zero(usemask)};
         al::destroy_at(Filters+idx);
         usemask &= ~(1_u64 << idx);
     }
diff --git a/al/source.cpp b/al/source.cpp
index 770e3778..a4e59e59 100644
--- a/al/source.cpp
+++ b/al/source.cpp
@@ -45,6 +45,7 @@
 #include "AL/alext.h"
 #include "AL/efx.h"
 
+#include "albit.h"
 #include "alcmain.h"
 #include "alcontext.h"
 #include "almalloc.h"
@@ -684,7 +685,7 @@ bool EnsureSources(ALCcontext *context, size_t needed)
     size_t count{std::accumulate(context->mSourceList.cbegin(), context->mSourceList.cend(),
         size_t{0},
         [](size_t cur, const SourceSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<ALuint>(PopCount(sublist.FreeMask)); })};
+        { return cur + static_cast<ALuint>(al::popcount(sublist.FreeMask)); })};
 
     while(needed > count)
     {
@@ -712,7 +713,7 @@ ALsource *AllocSource(ALCcontext *context)
         { return entry.FreeMask != 0; }
     );
     auto lidx = static_cast<ALuint>(std::distance(context->mSourceList.begin(), sublist));
-    auto slidx = static_cast<ALuint>(CountTrailingZeros(sublist->FreeMask));
+    auto slidx = static_cast<ALuint>(al::countr_zero(sublist->FreeMask));
 
     ALsource *source{::new(sublist->Sources + slidx) ALsource{}};
 
@@ -3484,7 +3485,7 @@ SourceSubList::~SourceSubList()
     uint64_t usemask{~FreeMask};
     while(usemask)
     {
-        const ALsizei idx{CountTrailingZeros(usemask)};
+        const int idx{al::countr_zero(usemask)};
         al::destroy_at(Sources+idx);
         usemask &= ~(1_u64 << idx);
     }
diff --git a/alc/alc.cpp b/alc/alc.cpp
index e7c85525..02aaaa18 100644
--- a/alc/alc.cpp
+++ b/alc/alc.cpp
@@ -64,6 +64,7 @@
 #include "al/filter.h"
 #include "al/listener.h"
 #include "al/source.h"
+#include "albit.h"
 #include "alcmain.h"
 #include "albyte.h"
 #include "alconfig.h"
@@ -2128,7 +2129,7 @@ static ALCenum UpdateDeviceParams(ALCdevice *device, const int *attrList)
             uint64_t usemask{~sublist.FreeMask};
             while(usemask)
             {
-                const int idx{CountTrailingZeros(usemask)};
+                const int idx{al::countr_zero(usemask)};
                 ALeffectslot *slot{sublist.EffectSlots + idx};
                 usemask &= ~(1_u64 << idx);
 
@@ -2149,7 +2150,7 @@ static ALCenum UpdateDeviceParams(ALCdevice *device, const int *attrList)
             uint64_t usemask{~sublist.FreeMask};
             while(usemask)
             {
-                const int idx{CountTrailingZeros(usemask)};
+                const int idx{al::countr_zero(usemask)};
                 ALsource *source{sublist.Sources + idx};
                 usemask &= ~(1_u64 << idx);
 
@@ -2292,19 +2293,19 @@ ALCdevice::~ALCdevice()
 
     size_t count{std::accumulate(BufferList.cbegin(), BufferList.cend(), size_t{0u},
         [](size_t cur, const BufferSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<uint>(PopCount(~sublist.FreeMask)); })};
+        { return cur + static_cast<uint>(al::popcount(~sublist.FreeMask)); })};
     if(count > 0)
         WARN("%zu Buffer%s not deleted\n", count, (count==1)?"":"s");
 
     count = std::accumulate(EffectList.cbegin(), EffectList.cend(), size_t{0u},
         [](size_t cur, const EffectSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<uint>(PopCount(~sublist.FreeMask)); });
+        { return cur + static_cast<uint>(al::popcount(~sublist.FreeMask)); });
     if(count > 0)
         WARN("%zu Effect%s not deleted\n", count, (count==1)?"":"s");
 
     count = std::accumulate(FilterList.cbegin(), FilterList.cend(), size_t{0u},
         [](size_t cur, const FilterSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<uint>(PopCount(~sublist.FreeMask)); });
+        { return cur + static_cast<uint>(al::popcount(~sublist.FreeMask)); });
     if(count > 0)
         WARN("%zu Filter%s not deleted\n", count, (count==1)?"":"s");
 
@@ -2356,7 +2357,7 @@ ALCcontext::~ALCcontext()
 
     count = std::accumulate(mSourceList.cbegin(), mSourceList.cend(), size_t{0u},
         [](size_t cur, const SourceSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<uint>(PopCount(~sublist.FreeMask)); });
+        { return cur + static_cast<uint>(al::popcount(~sublist.FreeMask)); });
     if(count > 0)
         WARN("%zu Source%s not deleted\n", count, (count==1)?"":"s");
     mSourceList.clear();
@@ -2381,7 +2382,7 @@ ALCcontext::~ALCcontext()
 
     count = std::accumulate(mEffectSlotList.cbegin(), mEffectSlotList.cend(), size_t{0u},
         [](size_t cur, const EffectSlotSubList &sublist) noexcept -> size_t
-        { return cur + static_cast<uint>(PopCount(~sublist.FreeMask)); });
+        { return cur + static_cast<uint>(al::popcount(~sublist.FreeMask)); });
     if(count > 0)
         WARN("%zu AuxiliaryEffectSlot%s not deleted\n", count, (count==1)?"":"s");
     mEffectSlotList.clear();
diff --git a/alc/backends/base.cpp b/alc/backends/base.cpp
index 8642f40c..c4a4abeb 100644
--- a/alc/backends/base.cpp
+++ b/alc/backends/base.cpp
@@ -12,6 +12,7 @@
 #include <mmreg.h>
 #endif
 
+#include "albit.h"
 #include "alcmain.h"
 #include "alnumeric.h"
 #include "aloptional.h"
@@ -180,7 +181,7 @@ void BackendBase::setChannelOrderFromWFXMask(uint chanmask)
     uint idx{0};
     while(chanmask)
     {
-        const int bit{CountTrailingZeros(chanmask)};
+        const int bit{al::countr_zero(chanmask)};
         const uint mask{1u << bit};
         chanmask &= ~mask;
 
diff --git a/alc/backends/wasapi.cpp b/alc/backends/wasapi.cpp
index 1d1b93c4..0786a7d7 100644
--- a/alc/backends/wasapi.cpp
+++ b/alc/backends/wasapi.cpp
@@ -55,6 +55,7 @@
 #include <thread>
 #include <vector>
 
+#include "albit.h"
 #include "alcmain.h"
 #include "alu.h"
 #include "compat.h"
@@ -1643,7 +1644,7 @@ HRESULT WasapiCapture::resetProxy()
         if((InputType.dwChannelMask&SPEAKER_LOW_FREQUENCY))
         {
             constexpr auto lfemask = MaskFromTopBits(SPEAKER_LOW_FREQUENCY);
-            const int lfeidx{PopCount(uint32_t{InputType.dwChannelMask&lfemask}) - 1};
+            const int lfeidx{al::popcount(InputType.dwChannelMask&lfemask) - 1};
             chanmask &= ~(1u << lfeidx);
         }
 
diff --git a/alc/converter.cpp b/alc/converter.cpp
index 5016b373..f7d4fc46 100644
--- a/alc/converter.cpp
+++ b/alc/converter.cpp
@@ -9,6 +9,7 @@
 #include <iterator>
 #include <limits.h>
 
+#include "albit.h"
 #include "albyte.h"
 #include "alnumeric.h"
 #include "core/fpu_ctrl.h"
@@ -338,7 +339,7 @@ void ChannelConverter::convert(const void *src, float *dst, uint frames) const
 {
     if(mDstChans == DevFmtMono)
     {
-        const float scale{std::sqrt(1.0f / static_cast<float>(PopCount(mChanMask)))};
+        const float scale{std::sqrt(1.0f / static_cast<float>(al::popcount(mChanMask)))};
         switch(mSrcType)
         {
 #define HANDLE_FMT(T) case T: Multi2Mono<T>(mChanMask, mSrcStep, scale, dst, src, frames); break
diff --git a/common/albit.h b/common/albit.h
index 225c0b89..c54bb31a 100644
--- a/common/albit.h
+++ b/common/albit.h
@@ -1,6 +1,12 @@
 #ifndef AL_BIT_H
 #define AL_BIT_H
 
+#include <limits>
+#include <type_traits>
+#if !defined(__GNUC__) && (defined(_WIN32) || defined(_WIN64))
+#include <intrin.h>
+#endif
+
 namespace al {
 
 #ifdef __BYTE_ORDER__
@@ -30,6 +36,108 @@ enum class endian {
 };
 #endif
 
+
+/* Define popcount (population count/count 1 bits) and countr_zero (count
+ * trailing zero bits, starting from the lsb) methods, for various integer
+ * types.
+ */
+#ifdef __GNUC__
+
+namespace detail_ {
+    inline int popcount(unsigned long long val) noexcept { return __builtin_popcountll(val); }
+    inline int popcount(unsigned long val) noexcept { return __builtin_popcountl(val); }
+    inline int popcount(unsigned int val) noexcept { return __builtin_popcount(val); }
+
+    inline int countr_zero(unsigned long long val) noexcept { return __builtin_ctzll(val); }
+    inline int countr_zero(unsigned long val) noexcept { return __builtin_ctzl(val); }
+    inline int countr_zero(unsigned int val) noexcept { return __builtin_ctz(val); }
+} // namespace detail_
+
+template<typename T>
+inline std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
+int> popcount(T v) noexcept { return detail_::popcount(v); }
+
+template<typename T>
+inline std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
+int> countr_zero(T val) noexcept
+{ return val ? detail_::countr_zero(val) : std::numeric_limits<T>::digits; }
+
+#else
+
+/* There be black magics here. The popcount method is derived from
+ * https://graphics.stanford.edu/~seander/bithacks.html#CountBitsSetParallel
+ * while the ctz-utilizing-popcount algorithm is shown here
+ * http://www.hackersdelight.org/hdcodetxt/ntz.c.txt
+ * as the ntz2 variant. These likely aren't the most efficient methods, but
+ * they're good enough if the GCC built-ins aren't available.
+ */
+namespace detail_ {
+    template<typename T>
+    constexpr T repbits(unsigned char bits) noexcept
+    {
+        T ret{bits};
+        for(size_t i{1};i < sizeof(T);++i)
+            ret = (ret<<8) | bits;
+        return ret;
+    }
+} // namespace detail_
+
+template<typename T>
+constexpr std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
+int> popcount(T v) noexcept
+{
+    constexpr T m55{detail_::repbits<T>(0x55)};
+    constexpr T m33{detail_::repbits<T>(0x33)};
+    constexpr T m0f{detail_::repbits<T>(0x0f)};
+    constexpr T m01{detail_::repbits<T>(0x01)};
+
+    v = v - ((v >> 1) & m55);
+    v = (v & m33) + ((v >> 2) & m33);
+    v = (v + (v >> 4)) & m0f;
+    return static_cast<int>((v * m01) >> ((sizeof(T)-1)*8));
+}
+
+#if defined(_WIN64)
+
+template<typename T>
+inline std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
+int> countr_zero(T v)
+{
+    unsigned long idx{std::numeric_limits<T>::digits};
+    if /*constexpr*/(std::numeric_limits<T>::digits <= 32)
+        _BitScanForward(&idx, static_cast<uint32_t>(v));
+    else // std::numeric_limits<T>::digits > 32
+        _BitScanForward64(&idx, v);
+    return static_cast<int>(idx);
+}
+
+#elif defined(_WIN32)
+
+template<typename T>
+inline std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
+int> countr_zero(T v)
+{
+    unsigned long idx{std::numeric_limits<T>::digits};
+    if /*constexpr*/(std::numeric_limits<T>::digits <= 32)
+        _BitScanForward(&idx, static_cast<uint32_t>(v));
+    else if(!_BitScanForward(&idx, static_cast<uint32_t>(v)))
+    {
+        if(_BitScanForward(&idx, static_cast<uint32_t>(v>>32)))
+            idx += 32;
+    }
+    return static_cast<int>(idx);
+}
+
+#else
+
+template<typename T>
+constexpr std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
+int> countr_zero(T value)
+{ return popcount(static_cast<T>(~value & (value - 1))); }
+
+#endif
+#endif
+
 } // namespace al
 
 #endif /* AL_BIT_H */
diff --git a/common/alcomplex.cpp b/common/alcomplex.cpp
index 8a823b01..de10ede2 100644
--- a/common/alcomplex.cpp
+++ b/common/alcomplex.cpp
@@ -8,6 +8,7 @@
 #include <cstddef>
 #include <utility>
 
+#include "albit.h"
 #include "alnumeric.h"
 #include "math_defs.h"
 
@@ -18,7 +19,7 @@ void complex_fft(const al::span<std::complex<double>> buffer, const double sign)
     /* Get the number of bits used for indexing. Simplifies bit-reversal and
      * the main loop count.
      */
-    const size_t log2_size{static_cast<size_t>(CountTrailingZeros(fftsize))};
+    const size_t log2_size{static_cast<size_t>(al::countr_zero(fftsize))};
 
     /* Bit-reversal permutation applied to a sequence of fftsize items. */
     for(size_t idx{1u};idx < fftsize-1;++idx)
diff --git a/common/alnumeric.h b/common/alnumeric.h
index b9384a7f..c16f3e62 100644
--- a/common/alnumeric.h
+++ b/common/alnumeric.h
@@ -103,92 +103,6 @@ inline size_t RoundUp(size_t value, size_t r) noexcept
 }
 
 
-/* Define CountTrailingZeros (count trailing zero bits, starting from the lsb)
- * and PopCount (population count/count 1 bits) methods, for 32- and 64-bit
- * integers. The CountTrailingZeros results are *UNDEFINED* if the value is 0.
- */
-#ifdef __GNUC__
-
-/* Define variations for unsigned (long (long)) int, since we don't know what
- * uint32/64_t are typedef'd to.
- */
-inline int PopCount(unsigned long long val) { return __builtin_popcountll(val); }
-inline int PopCount(unsigned long val) { return __builtin_popcountl(val); }
-inline int PopCount(unsigned int val) { return __builtin_popcount(val); }
-
-inline int CountTrailingZeros(unsigned long long val) { return __builtin_ctzll(val); }
-inline int CountTrailingZeros(unsigned long val) { return __builtin_ctzl(val); }
-inline int CountTrailingZeros(unsigned int val) { return __builtin_ctz(val); }
-
-#else
-
-/* There be black magics here. The popcnt method is derived from
- * https://graphics.stanford.edu/~seander/bithacks.html#CountBitsSetParallel
- * while the ctz-utilizing-popcnt algorithm is shown here
- * http://www.hackersdelight.org/hdcodetxt/ntz.c.txt
- * as the ntz2 variant. These likely aren't the most efficient methods, but
- * they're good enough if the GCC built-ins aren't available.
- */
-inline int PopCount(uint32_t v)
-{
-    v = v - ((v >> 1) & 0x55555555u);
-    v = (v & 0x33333333u) + ((v >> 2) & 0x33333333u);
-    v = (v + (v >> 4)) & 0x0f0f0f0fu;
-    return static_cast<int>((v * 0x01010101u) >> 24);
-}
-inline int PopCount(uint64_t v)
-{
-    v = v - ((v >> 1) & 0x5555555555555555_u64);
-    v = (v & 0x3333333333333333_u64) + ((v >> 2) & 0x3333333333333333_u64);
-    v = (v + (v >> 4)) & 0x0f0f0f0f0f0f0f0f_u64;
-    return static_cast<int>((v * 0x0101010101010101_u64) >> 56);
-}
-
-#if defined(_WIN64)
-
-inline int CountTrailingZeros(uint32_t v)
-{
-    unsigned long idx = 32;
-    _BitScanForward(&idx, v);
-    return static_cast<int>(idx);
-}
-inline int CountTrailingZeros(uint64_t v)
-{
-    unsigned long idx = 64;
-    _BitScanForward64(&idx, v);
-    return static_cast<int>(idx);
-}
-
-#elif defined(_WIN32)
-
-inline int CountTrailingZeros(uint32_t v)
-{
-    unsigned long idx = 32;
-    _BitScanForward(&idx, v);
-    return static_cast<int>(idx);
-}
-inline int CountTrailingZeros(uint64_t v)
-{
-    unsigned long idx = 64;
-    if(!_BitScanForward(&idx, static_cast<uint32_t>(v&0xffffffff)))
-    {
-        if(_BitScanForward(&idx, static_cast<uint32_t>(v>>32)))
-            idx += 32;
-    }
-    return static_cast<int>(idx);
-}
-
-#else
-
-inline int CountTrailingZeros(uint32_t value)
-{ return PopCount(~value & (value - 1)); }
-inline int CountTrailingZeros(uint64_t value)
-{ return PopCount(~value & (value - 1)); }
-
-#endif
-#endif
-
-
 /**
  * Fast float-to-int conversion. No particular rounding mode is assumed; the
  * IEEE-754 default is round-to-nearest with ties-to-even, though an app could