Introduce powerOf2Floor() and popCount8() functions

This is a prerequisite for the feature "Graph meter coloring (with
GraphData structure rework)".

powerOf2Floor() will utilize __builtin_clz() or stdc_bit_floor_ui()
(__builtin_clz() is preferred) if either is supported.

popCount8() will utilize ARM NEON instructions and x86 POPCNT
instruction if the machine supports either of them.

I am not adopting the C23 standard interface stdc_count_ones_uc() yet,
as I am not sure C libraries would implement it as fast as our version.

Signed-off-by: Kang-Che Sung <explorer09@gmail.com>

Author: Explorer09

XUtils.c

@@ -12,6 +12,7 @@ in the source distribution for its full text.
 #include <assert.h>
 #include <errno.h>
 #include <fcntl.h>
+#include <limits.h> // IWYU pragma: keep
 #include <math.h>
 #include <stdarg.h>
 #include <stdint.h>
@@ -387,3 +388,14 @@ unsigned int countTrailingZeros(unsigned int x) {
    return mod37BitPosition[(-x & x) % 37];
 }
 #endif
+
+#if !defined(HAVE_BUILTIN_CLZ) && !defined(HAVE_STDC_BIT_FLOOR)
+/* Returns the nearest power of two that is not greater than x.
+   If x is 0, returns 0. */
+unsigned int powerOf2Floor(unsigned int x) {
+   for (unsigned int shift = 1; shift < sizeof(x) * CHAR_BIT; shift <<= 1)
+      x |= x >> shift;
+
+   return x - (x >> 1);
+}
+#endif

XUtils.h

@@ -15,15 +15,27 @@ in the source distribution for its full text.
 #endif
 
 #include <dirent.h>
+#include <limits.h> // IWYU pragma: keep
 #include <stdbool.h>
 #include <stddef.h> // IWYU pragma: keep
+#include <stdint.h> // IWYU pragma: keep
 #include <stdio.h>
 #include <stdlib.h> // IWYU pragma: keep
 #include <string.h> // IWYU pragma: keep
 
 #include "Compat.h"
 #include "Macros.h"
 
+#ifdef HAVE_STDBIT_H
+#include <stdbit.h>
+#endif
+
+#if defined(HAVE_ARM_NEON_H) && defined(__ARM_NEON)
+// ARM C Language Extensions (ACLE) recommends us to check __ARM_NEON before
+// including <arm_neon.h>
+#include <arm_neon.h>
+#endif
+
 
 ATTR_NORETURN
 void fail(void);
@@ -150,6 +162,49 @@ static inline unsigned int countTrailingZeros(unsigned int x) {
 unsigned int countTrailingZeros(unsigned int x);
 #endif
 
+/* Returns the nearest power of two that is not greater than x.
+   If x is 0, returns 0. */
+#if defined(HAVE_BUILTIN_CLZ)
+static inline unsigned int powerOf2Floor(unsigned int x) {
+   if (x == 0)
+      return 0;
+
+   return 1U << ((int)(sizeof(x) * CHAR_BIT) - 1 - __builtin_clz(x));
+}
+#elif defined(HAVE_STDC_BIT_FLOOR)
+static inline unsigned int powerOf2Floor(unsigned int x) {
+   return stdc_bit_floor_ui(x);
+}
+#else
+unsigned int powerOf2Floor(unsigned int x);
+#endif
+
+static inline unsigned int popCount8(uint8_t x) {
+#if defined(HAVE_ARM_NEON_H) && defined(__ARM_NEON)
+   // With ARM Advanced SIMD extension (NEON), this generates smaller code than
+   // __builtin_popcount.
+   //
+   // Initialize the vector register. Set all lanes at once so that the
+   // compiler will not emit instruction to zero-initialize other lanes.
+   uint8x8_t v = vdup_n_u8(x);
+   // Count the number of set bits for each lane (8-bit) in the vector.
+   v = vcnt_u8(v);
+   // Get lane 0 and discard lanes 1 to 7. (Return type was uint8_t)
+   return vget_lane_u8(v, 0);
+#elif defined(HAVE_BUILTIN_POPCOUNT) && defined(__POPCNT__)
+   // x86 POPCNT instruction. __builtin_popcount translates to it when it is
+   // enabled ("-mpopcnt"). (Return type was int)
+   return (unsigned int)__builtin_popcount(x);
+#else
+   // This code is optimized for uint8_t input and smaller than the subroutine
+   // call of the compiler __builtin_popcount (which is tuned for
+   // unsigned int input type and not uint8_t).
+   uint32_t n = (uint32_t)(x * 0x08040201U);
+   n = (uint32_t)(((n >> 3) & 0x11111111U) * 0x11111111U) >> 28;
+   return n;
+#endif
+}
+
 /* IEC unit prefixes */
 static const char unitPrefixes[] = { 'K', 'M', 'G', 'T', 'P', 'E', 'Z', 'Y', 'R', 'Q' };
 

configure.ac

@@ -182,7 +182,10 @@ fi])
 
 # Optional Section
 
-AC_CHECK_HEADERS([execinfo.h])
+AC_CHECK_HEADERS([ \
+   execinfo.h \
+   stdbit.h \
+   ])
 
 if test "$my_htop_platform" = darwin; then
    AC_CHECK_HEADERS([mach/mach_time.h])
@@ -308,11 +311,47 @@ AC_LINK_IFELSE([
 
 AC_MSG_CHECKING(for __builtin_ctz)
 AC_COMPILE_IFELSE([
-   AC_LANG_PROGRAM([], [[__builtin_ctz(1); /* Supported in GCC 3.4 or later */]])],
+   AC_LANG_PROGRAM([], [[return __builtin_ctz(1U); /* Supported in GCC 3.4 or later */]])],
    [AC_DEFINE([HAVE_BUILTIN_CTZ], 1, [Define to 1 if the compiler supports '__builtin_ctz' function.])
    AC_MSG_RESULT(yes)],
    AC_MSG_RESULT(no))
 
+AC_MSG_CHECKING(for __builtin_clz)
+AC_COMPILE_IFELSE([
+   AC_LANG_PROGRAM([], [[return __builtin_clz(-1U); /* Supported in GCC 3.4 or later */]])],
+   [AC_DEFINE([HAVE_BUILTIN_CLZ], 1, [Define to 1 if the compiler supports '__builtin_clz' function.])
+   AC_MSG_RESULT(yes)],
+   AC_MSG_RESULT(no))
+
+AC_MSG_CHECKING(for __builtin_popcount)
+AC_COMPILE_IFELSE([
+   AC_LANG_PROGRAM([], [[return __builtin_popcount(0U); /* Supported in GCC 3.4 or later */]])],
+   [AC_DEFINE([HAVE_BUILTIN_POPCOUNT], 1, [Define to 1 if the compiler supports '__builtin_popcount' function.])
+   AC_MSG_RESULT(yes)],
+   AC_MSG_RESULT(no))
+
+AC_MSG_CHECKING(for stdc_bit_floor)
+AC_LINK_IFELSE([
+   AC_LANG_PROGRAM(
+      [[
+#include <stdbit.h>
+      ]],
+      [[
+         /* Both the type-generic and type-specific versions should exist.
+            htop uses the type-specific version. */
+         return stdc_bit_floor(0U) || stdc_bit_floor_ui(0U);
+      ]])],
+   [AC_DEFINE([HAVE_STDC_BIT_FLOOR], 1, [Define to 1 if stdc_bit_floor functions are supported.])
+   AC_MSG_RESULT(yes)],
+   AC_MSG_RESULT(no))
+
+case "$host_cpu" in
+   arm*|aarch64*)
+      dnl ARM NEON intrinsics
+      AC_CHECK_HEADERS([arm_neon.h])
+      ;;
+esac
+
 # ----------------------------------------------------------------------