Rename _bsr/_bsf to bsr/bsf

Now that these functions are behind _COSMO_SOURCE there's no reason for
having the ugly underscore anymore. To use these functions, you need to
pass -mcosmo to cosmocc.

dsp/scale/gyarados.c

@@ -202,8 +202,8 @@ void *Gyarados(long dyw, long dxw, int dst[dyw][dxw], long syw, long sxw,
       CHECK_LE(sxn, sxw);
       CHECK_LE(dyn, dyw);
       CHECK_LE(dxn, dxw);
-      CHECK_LT(_bsrl(syn) + _bsrl(sxn), 32);
-      CHECK_LT(_bsrl(dyn) + _bsrl(dxn), 32);
+      CHECK_LT(bsrl(syn) + bsrl(sxn), 32);
+      CHECK_LT(bsrl(dyn) + bsrl(dxn), 32);
       CHECK_LE(dyw, 0x7fff);
       CHECK_LE(dxw, 0x7fff);
       CHECK_LE(syw, 0x7fff);

libc/calls/fstat-nt.c

@@ -71,7 +71,7 @@ static textwindows long GetSizeOfReparsePoint(int64_t fh) {
       }
     }
     if (x >= 0200) {
-      z += _bsrl(tpenc(x)) >> 3;
+      z += bsrl(tpenc(x)) >> 3;
     }
     ++z;
   }

libc/calls/pledge-linux.c

@@ -1040,7 +1040,7 @@ static privileged char *FixCpy(char p[17], uint64_t x, int k) {
 }
 
 static privileged char *HexCpy(char p[17], uint64_t x) {
-  return FixCpy(p, x, ROUNDUP(x ? _bsrl(x) + 1 : 1, 4));
+  return FixCpy(p, x, ROUNDUP(x ? bsrl(x) + 1 : 1, 4));
 }
 
 static privileged int GetPid(void) {

libc/calls/sig.c

@@ -97,7 +97,7 @@ static textwindows int __sig_getter(atomic_ulong *sigs, sigset_t masked) {
   for (;;) {
     pending = atomic_load_explicit(sigs, memory_order_acquire);
     if ((deliverable = pending & ~masked)) {
-      sig = _bsfl(deliverable) + 1;
+      sig = bsfl(deliverable) + 1;
       bit = 1ull << (sig - 1);
       if (atomic_fetch_and_explicit(sigs, ~bit, memory_order_acq_rel) & bit) {
         return sig;

libc/fmt/formatbinary64.c

@@ -45,7 +45,7 @@ char *FormatBinary64(char p[hasatleast 67], uint64_t x, char z) {
       *p++ = '0';
       *p++ = 'b';
     }
-    i = PickGoodWidth(_bsrl(x));
+    i = PickGoodWidth(bsrl(x));
     do {
       b = 1;
       b <<= i;

libc/fmt/itoa64radix16.greg.c

@@ -22,5 +22,5 @@
 #include "libc/macros.internal.h"
 
 size_t uint64toarray_radix16(uint64_t x, char b[hasatleast 17]) {
-  return uint64toarray_fixed16(x, b, ROUNDUP(x ? _bsrl(x) + 1 : 1, 4));
+  return uint64toarray_fixed16(x, b, ROUNDUP(x ? bsrl(x) + 1 : 1, 4));
 }

libc/integral/cxx.inc

@@ -18,12 +18,14 @@
 #if 1
 template <bool _P, typename _T, typename _U>
 struct __cxx_choose_expr {
-  __cxx_choose_expr(_T _a, _U _b) : _value(_a) {}
+  __cxx_choose_expr(_T _a, _U _b) : _value(_a) {
+  }
   const _T _value;
 };
 template <typename _T, typename _U>
 struct __cxx_choose_expr<false, _T, _U> {
-  __cxx_choose_expr(_T _a, _U _b) : _value(_b) {}
+  __cxx_choose_expr(_T _a, _U _b) : _value(_b) {
+  }
   const _U _value;
 };
 #define __builtin_choose_expr(X, A, B) \
@@ -32,3 +34,8 @@ struct __cxx_choose_expr<false, _T, _U> {
 #define __builtin_choose_expr(X, A, B) ((X) ? (A) : (B))
 #endif
 #endif
+
+#ifdef __aarch64__
+/* todo jart whyyyy */
+#define _Float16 __fp16
+#endif

libc/intrin/bsf.c

@@ -22,7 +22,7 @@
  * Returns position of first bit set.
  *
  *                           ctz(𝑥)         31^clz(𝑥)   clz(𝑥)
- *       uint32 𝑥  _bsf(𝑥) tzcnt(𝑥)   ffs(𝑥)  _bsr(𝑥) lzcnt(𝑥)
+ *       uint32 𝑥   bsf(𝑥) tzcnt(𝑥)   ffs(𝑥)   bsr(𝑥) lzcnt(𝑥)
  *     0x00000000      wut       32        0      wut       32
  *     0x00000001        0        0        1        0       31
  *     0x80000001        0        0        1       31        0
@@ -35,7 +35,7 @@
  * @param 𝑥 is a 64-bit integer
  * @return number in range 0..63 or undefined if 𝑥 is 0
  */
-int(_bsfl)(long x) {
+int(bsfl)(long x) {
   unsigned l, r;
   x &= -x;
   l = x | x >> 32;
@@ -52,7 +52,7 @@ int(_bsfl)(long x) {
  * Returns position of first bit set.
  *
  *                           ctz(𝑥)         31^clz(𝑥)   clz(𝑥)
- *       uint32 𝑥  _bsf(𝑥) tzcnt(𝑥)   ffs(𝑥)  _bsr(𝑥) lzcnt(𝑥)
+ *       uint32 𝑥   bsf(𝑥) tzcnt(𝑥)   ffs(𝑥)   bsr(𝑥) lzcnt(𝑥)
  *     0x00000000      wut       32        0      wut       32
  *     0x00000001        0        0        1        0       31
  *     0x80000001        0        0        1       31        0
@@ -65,8 +65,8 @@ int(_bsfl)(long x) {
  * @param x is a 32-bit integer
  * @return number in range 0..31 or undefined if 𝑥 is 0
  */
-int(_bsf)(int x) {
-  return _bsf((unsigned)x);
+int(bsf)(int x) {
+  return bsf((unsigned)x);
 }
 
-__weak_reference(_bsfl, _bsfll);
+__weak_reference(bsfl, bsfll);

libc/intrin/bsf.h

@@ -1,16 +1,18 @@
+#ifdef _COSMO_SOURCE
 #ifndef COSMOPOLITAN_LIBC_NEXGEN32E_BSF_H_
 #define COSMOPOLITAN_LIBC_NEXGEN32E_BSF_H_
 COSMOPOLITAN_C_START_
 
-libcesque int _bsf(int) pureconst;
-libcesque int _bsfl(long) pureconst;
-libcesque int _bsfll(long long) pureconst;
+libcesque int bsf(int) pureconst;
+libcesque int bsfl(long) pureconst;
+libcesque int bsfll(long long) pureconst;
 
 #if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-#define _bsf(x)   __builtin_ctz(x)
-#define _bsfl(x)  __builtin_ctzl(x)
-#define _bsfll(x) __builtin_ctzll(x)
+#define bsf(x)   __builtin_ctz(x)
+#define bsfl(x)  __builtin_ctzl(x)
+#define bsfll(x) __builtin_ctzll(x)
 #endif
 
 COSMOPOLITAN_C_END_
 #endif /* COSMOPOLITAN_LIBC_NEXGEN32E_BSF_H_ */
+#endif /* _COSMO_SOURCE */

libc/intrin/bsr.c

@@ -23,7 +23,7 @@
  * Returns binary logarithm of 𝑥.
  *
  *                           ctz(𝑥)         31^clz(𝑥)   clz(𝑥)
- *       uint32 𝑥  _bsf(𝑥) tzcnt(𝑥)   ffs(𝑥)  _bsr(𝑥) lzcnt(𝑥)
+ *       uint32 𝑥   bsf(𝑥) tzcnt(𝑥)   ffs(𝑥)   bsr(𝑥) lzcnt(𝑥)
  *     0x00000000      wut       32        0      wut       32
  *     0x00000001        0        0        1        0       31
  *     0x80000001        0        0        1       31        0
@@ -36,7 +36,7 @@
  * @param x is a 32-bit integer
  * @return number in range 0..31 or undefined if 𝑥 is 0
  */
-int(_bsr)(int x) {
+int(bsr)(int x) {
   int r = 0;
   if(x & 0xFFFF0000u) { x >>= 16; r |= 16; }
   if(x & 0xFF00) { x >>= 8; r |= 8; }

libc/intrin/bsr.h

@@ -1,16 +1,18 @@
+#ifdef _COSMO_SOURCE
 #ifndef COSMOPOLITAN_LIBC_NEXGEN32E_BSR_H_
 #define COSMOPOLITAN_LIBC_NEXGEN32E_BSR_H_
 COSMOPOLITAN_C_START_
 
-libcesque int _bsr(int) pureconst;
-libcesque int _bsrl(long) pureconst;
-libcesque int _bsrll(long long) pureconst;
+libcesque int bsr(int) pureconst;
+libcesque int bsrl(long) pureconst;
+libcesque int bsrll(long long) pureconst;
 
 #if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-#define _bsr(x)   (__builtin_clz(x) ^ (sizeof(int) * 8 - 1))
-#define _bsrl(x)  (__builtin_clzl(x) ^ (sizeof(long) * 8 - 1))
-#define _bsrll(x) (__builtin_clzll(x) ^ (sizeof(long long) * 8 - 1))
+#define bsr(x)   (__builtin_clz(x) ^ (sizeof(int) * 8 - 1))
+#define bsrl(x)  (__builtin_clzl(x) ^ (sizeof(long) * 8 - 1))
+#define bsrll(x) (__builtin_clzll(x) ^ (sizeof(long long) * 8 - 1))
 #endif
 
 COSMOPOLITAN_C_END_
 #endif /* COSMOPOLITAN_LIBC_NEXGEN32E_BSR_H_ */
+#endif /* _COSMO_SOURCE */

libc/intrin/bsrl.c

@@ -29,7 +29,7 @@ static const char kDebruijn[64] = {
  * Returns binary logarithm of 𝑥.
  *
  *                           ctz(𝑥)         31^clz(𝑥)   clz(𝑥)
- *       uint32 𝑥  _bsf(𝑥) tzcnt(𝑥)   ffs(𝑥)  _bsr(𝑥) lzcnt(𝑥)
+ *       uint32 𝑥   bsf(𝑥) tzcnt(𝑥)   ffs(𝑥)   bsr(𝑥) lzcnt(𝑥)
  *     0x00000000      wut       32        0      wut       32
  *     0x00000001        0        0        1        0       31
  *     0x80000001        0        0        1       31        0
@@ -42,7 +42,7 @@ static const char kDebruijn[64] = {
  * @param x is a 64-bit integer
  * @return number in range 0..63 or undefined if 𝑥 is 0
  */
-int(_bsrl)(long x) {
+int(bsrl)(long x) {
   x |= x >> 1;
   x |= x >> 2;
   x |= x >> 4;
@@ -52,4 +52,4 @@ int(_bsrl)(long x) {
   return kDebruijn[(x * 0x03f79d71b4cb0a89ull) >> 58];
 }
 
-__weak_reference(_bsrl, _bsrll);
+__weak_reference(bsrl, bsrll);

libc/intrin/cxaatexit.c

@@ -58,7 +58,7 @@ int __cxa_atexit(void *fp, void *arg, void *pred) {
       return enomem();
     }
   }
-  i = _bsr(~b->mask);
+  i = bsr(~b->mask);
   b->mask |= 1u << i;
   b->p[i].fp = fp;
   b->p[i].arg = arg;

libc/intrin/cxafinalize.c

@@ -43,7 +43,7 @@ StartOverLocked:
     for (;;) {
       mask = b->mask;
       while (mask) {
-        i = _bsf(mask);
+        i = bsf(mask);
         mask &= ~(1u << i);
         if (!pred || pred == b->p[i].pred) {
           b->mask &= ~(1u << i);

libc/intrin/describefdset.c

@@ -43,7 +43,7 @@ const char *(DescribeFdSet)(char buf[N], ssize_t rc, int nfds, fd_set *fds) {
   for (int fd = 0; fd < nfds; fd += 64) {
     uint64_t w = fds->fds_bits[fd >> 6];
     while (w) {
-      unsigned m = _bsr(w);
+      unsigned m = bsr(w);
       w &= ~((uint64_t)1 << m);
       if (fd + m < nfds) {
         if (!gotsome) {

libc/intrin/formathex64.c

@@ -48,7 +48,7 @@ char *FormatHex64(char p[hasatleast 19], uint64_t x, char z) {
       *p++ = '0';
       *p++ = 'x';
     }
-    i = PickGoodWidth(_bsrl(x), z);
+    i = PickGoodWidth(bsrl(x), z);
     do {
       *p++ = "0123456789abcdef"[(x >> (i -= 4)) & 15];
     } while (i);

libc/intrin/tpenc.c

@@ -34,7 +34,7 @@ uint64_t tpenc(uint32_t c) {
   int e, n;
   uint64_t w;
   if (0 <= c && c <= 127) return c;
-  e = kTpEnc[_bsr(c) - 7];
+  e = kTpEnc[bsr(c) - 7];
   n = e & 0xff;
   w = 0;
   do {

libc/log/cxaprintexits.c

@@ -41,7 +41,7 @@ void __cxa_printexits(FILE *f, void *pred) {
     do {
       mask = b->mask;
       while (mask) {
-        i = _bsf(mask);
+        i = bsf(mask);
         mask &= ~(1u << i);
         if (!pred || pred == b->p[i].pred) {
           symbol = GetSymbolByAddr((intptr_t)b->p[i].fp);

libc/proc/posix_spawn.c

@@ -587,7 +587,7 @@ errno_t posix_spawn(int *pid, const char *path,
     if (flags & POSIX_SPAWN_SETRLIMIT) {
       int rlimset = (*attrp)->rlimset;
       while (rlimset) {
-        int resource = _bsf(rlimset);
+        int resource = bsf(rlimset);
         rlimset &= ~(1u << resource);
         if (setrlimit(resource, (*attrp)->rlim + resource)) {
           // MacOS ARM64 RLIMIT_STACK always returns EINVAL

libc/runtime/mmap.c

@@ -73,7 +73,7 @@
 #define FRAME(x)   ((int)((intptr_t)(x) >> 16))
 
 static inline pureconst unsigned long __rounddown2pow(unsigned long x) {
-  return x ? 1ul << _bsrl(x) : 0;
+  return x ? 1ul << bsrl(x) : 0;
 }
 
 static wontreturn void __mmap_die(const char *s) {

libc/stdio/appendr.c

@@ -55,7 +55,7 @@ ssize_t appendr(char **b, size_t i) {
   z = appendz((p = *b));
   if (i != z.i || !p) {
     n = ROUNDUP(i + 1, 8) + W;
-    if (n > z.n || _bsrl(n) < _bsrl(z.n)) {
+    if (n > z.n || bsrl(n) < bsrl(z.n)) {
       if ((p = realloc(p, n))) {
         z.n = malloc_usable_size(p);
         unassert(!(z.n & (W - 1)));

libc/stdio/appendw.c

@@ -18,10 +18,10 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/assert.h"
 #include "libc/dce.h"
-#include "libc/serialize.h"
 #include "libc/intrin/bsr.h"
 #include "libc/macros.internal.h"
 #include "libc/mem/mem.h"
+#include "libc/serialize.h"
 #include "libc/stdio/append.h"
 
 #define W sizeof(size_t)
@@ -59,7 +59,7 @@ ssize_t appendw(char **b, uint64_t w) {
   char *p, *q;
   struct appendz z;
   z = appendz((p = *b));
-  l = w ? (_bsrl(w) >> 3) + 1 : 1;
+  l = w ? (bsrl(w) >> 3) + 1 : 1;
   n = ROUNDUP(z.i + 8 + 1, 8) + W;
   if (n > z.n) {
     if (!z.n) z.n = W * 2;

libc/stdio/fmt.c

@@ -359,14 +359,14 @@ static int __fmt_stoa_wide(out_f out, void *a, uint64_t w) {
   char buf[8];
   if (!isascii(w)) w = tpenc(w);
   WRITE64LE(buf, w);
-  return out(buf, a, w ? (_bsr(w) >> 3) + 1 : 1);
+  return out(buf, a, w ? (bsr(w) >> 3) + 1 : 1);
 }
 
 static int __fmt_stoa_bing(out_f out, void *a, uint64_t w) {
   char buf[8];
   w = tpenc(kCp437[w & 0xFF]);
   WRITE64LE(buf, w);
-  return out(buf, a, w ? (_bsr(w) >> 3) + 1 : 1);
+  return out(buf, a, w ? (bsr(w) >> 3) + 1 : 1);
 }
 
 static int __fmt_stoa_quoted(out_f out, void *a, uint64_t w) {
@@ -377,7 +377,7 @@ static int __fmt_stoa_quoted(out_f out, void *a, uint64_t w) {
     w = tpenc(w);
   }
   WRITE64LE(buf, w);
-  return out(buf, a, w ? (_bsr(w) >> 3) + 1 : 1);
+  return out(buf, a, w ? (bsr(w) >> 3) + 1 : 1);
 }
 
 /**

libc/str/thompike.h

@@ -5,7 +5,7 @@
 #define ThomPikeCont(x)     (0200 == (0300 & (x)))
 #define ThomPikeByte(x)     ((x) & (((1 << ThomPikeMsb(x)) - 1) | 3))
 #define ThomPikeLen(x)      (7 - ThomPikeMsb(x))
-#define ThomPikeMsb(x)      ((255 & (x)) < 252 ? _bsr(255 & ~(x)) : 1)
+#define ThomPikeMsb(x)      ((255 & (x)) < 252 ? bsr(255 & ~(x)) : 1)
 #define ThomPikeMerge(x, y) ((x) << 6 | (077 & (y)))
 
 #endif /* COSMOPOLITAN_LIBC_STR_THOMPIKE_H_ */

libc/str/tpdecodecb.internal.h

@@ -18,7 +18,7 @@ forceinline int tpdecodecb(wint_t *out, int first,
     if ((wc = get(arg, i++)) == -1) return -1;
   }
   if (__builtin_expect(!(0 <= wc && wc <= 0x7F), 0)) {
-    msb = wc < 252 ? _bsr(~wc & 0xff) : 1;
+    msb = wc < 252 ? bsr(~wc & 0xff) : 1;
     need = 7 - msb;
     wc &= ((1u << msb) - 1) | 0b00000011;
     for (j = 1; j < need; ++j) {

libc/x/utf16to8.c

@@ -16,7 +16,6 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/serialize.h"
 #include "libc/intrin/bsr.h"
 #include "libc/intrin/packsswb.h"
 #include "libc/intrin/pandn.h"
@@ -26,6 +25,7 @@
 #include "libc/intrin/punpckhbw.h"
 #include "libc/intrin/punpcklbw.h"
 #include "libc/mem/mem.h"
+#include "libc/serialize.h"
 #include "libc/str/str.h"
 #include "libc/str/thompike.h"
 #include "libc/str/utf16.h"
@@ -78,7 +78,7 @@ char *utf16to8(const char16_t *p, size_t n, size_t *z) {
       } else {
         w = tpenc(x);
         WRITE64LE(q, w);
-        q += _bsr(w) >> 3;
+        q += bsr(w) >> 3;
         q += 1;
       }
     }

net/finger/fingersyn.c

@@ -36,7 +36,7 @@ uint32_t FingerSyn(const char *p, size_t n) {
     // ttl<=128 Windows, OpenVMS 8+
     // ttl<=64  Mac, Linux, BSD, Solaris 8+, Tru64, HP-UX
     ttl = p[8] & 255;
-    h += _bsr(MAX(1, ttl - 1));
+    h += bsr(MAX(1, ttl - 1));
     h *= 0x9e3779b1;
     // TCP Options
     // We care about the order and presence of leading common options.

net/finger/getosname.c

@@ -21,7 +21,7 @@
 
 const char *GetOsName(int os) {
   if (os) {
-    switch (_bsr(os)) {
+    switch (bsr(os)) {
       case 0:
         return "LINUX";
       case 1:

net/turfwar/turfwar.c

@@ -1538,7 +1538,7 @@ bool GeneratePlot(struct Asset *out, long block, long cash) {
   CHECK_SQL(sqlite3_exec(db, "BEGIN TRANSACTION", 0, 0, 0));
   area = XN * YN;
   mask = area - 1;
-  clump = 32 - _bsr(area) - 8;
+  clump = 32 - bsr(area) - 8;
   while ((rc = DbStep(stmt)) != SQLITE_DONE) {
     if (rc != SQLITE_ROW) CHECK_DB(rc);
     ip = sqlite3_column_int64(stmt, 0);

test/libc/str/bsr_test.c

@@ -20,7 +20,7 @@
 #include "libc/str/str.h"
 #include "libc/testlib/testlib.h"
 
-static unsigned _Bsr(unsigned x) {
+static unsigned Bsr(unsigned x) {
   static const char kDebruijn[32] = {
       0, 9,  1,  10, 13, 21, 2,  29, 11, 14, 16, 18, 22, 25, 3, 30,
       8, 12, 20, 28, 15, 17, 24, 7,  19, 27, 23, 6,  26, 5,  4, 31,
@@ -35,8 +35,8 @@ static unsigned _Bsr(unsigned x) {
   return kDebruijn[x];
 }
 
-TEST(_bsr, test) {
+TEST(bsr, test) {
   for (int i = 1; i < 1000; ++i) {
-    ASSERT_EQ(_bsr(i), _Bsr(i));
+    ASSERT_EQ(bsr(i), Bsr(i));
   }
 }

third_party/chibicc/as.c

@@ -2028,7 +2028,7 @@ static int ParseModrm(struct As *a, int *disp) {
         if (((reg & 070) >> 3) == 2) modrm |= HASASZ;  // asz
         if (IsComma(a)) {
           ++a->i;
-          modrm |= (_bsr(GetInt(a)) & 3) << 6;
+          modrm |= (bsr(GetInt(a)) & 3) << 6;
         }
       }
       ConsumePunct(a, ')');

third_party/chibicc/asm.c

@@ -301,7 +301,7 @@ static void PickAsmRegisters(Asm *a) {
           if (!(m = a->ops[i].regmask)) break;
           if (popcnt(m) != j) break;
           if (!(m &= regset)) CouldNotAllocateRegister(&a->ops[i], "rm");
-          pick = 1 << (a->ops[i].reg = _bsf(m));
+          pick = 1 << (a->ops[i].reg = bsf(m));
           if (pick & PRECIOUS) a->regclob |= pick;
           regset &= ~pick;
           a->ops[i].regmask = 0;
@@ -309,14 +309,14 @@ static void PickAsmRegisters(Asm *a) {
         case kAsmXmm:
           if (!(m = a->ops[i].regmask)) break;
           if (!(m &= xmmset)) CouldNotAllocateRegister(&a->ops[i], "xmm");
-          xmmset &= ~(1 << (a->ops[i].reg = _bsf(m)));
+          xmmset &= ~(1 << (a->ops[i].reg = bsf(m)));
           a->ops[i].regmask = 0;
           break;
         case kAsmFpu:
           if (!(m = a->ops[i].x87mask)) break;
           if (popcnt(m) != j) break;
           if (!(m &= x87sts)) CouldNotAllocateRegister(&a->ops[i], "fpu");
-          x87sts &= ~(1 << (a->ops[i].reg = _bsf(m)));
+          x87sts &= ~(1 << (a->ops[i].reg = bsf(m)));
           a->ops[i].x87mask = 0;
           break;
         default:
@@ -540,7 +540,7 @@ static char *HandleAsmSpecifier(Asm *a, char *p) {
   if ((i = c - '0') >= a->n) {
     error_tok(a->tok, "bad asm reference at offset %d", p - a->str);
   }
-  z = _bsr(a->ops[i].node->ty->size);
+  z = bsr(a->ops[i].node->ty->size);
   if (z > 3 && a->ops[i].type == kAsmReg) {
     error_tok(a->tok, "bad asm op size");
   }
@@ -679,7 +679,7 @@ static void StoreAsmOutputs(Asm *a) {
           println("\tset%s\t(%%rax)", a->ops[i].str + a->ops[i].predicate);
           break;
         case kAsmReg:
-          z = _bsr(a->ops[i].node->ty->size);
+          z = bsr(a->ops[i].node->ty->size);
           if (a->ops[i].reg) {
             gen_addr(a->ops[i].node);
             if (z > 3) error_tok(a->tok, "bad asm out size");
@@ -734,7 +734,7 @@ static void StoreAsmOutputs(Asm *a) {
 static void PushClobbers(Asm *a) {
   int i, regs = a->regclob & PRECIOUS;
   while (regs) {
-    i = _bsf(regs);
+    i = bsf(regs);
     pushreg(kGreg[3][i]);
     regs &= ~(1 << i);
   }
@@ -743,7 +743,7 @@ static void PushClobbers(Asm *a) {
 static void PopClobbers(Asm *a) {
   int i, regs = a->regclob & PRECIOUS;
   while (regs) {
-    i = _bsr(regs);
+    i = bsr(regs);
     popreg(kGreg[3][i]);
     regs &= ~(1 << i);
   }

third_party/chibicc/dox2.c

@@ -281,7 +281,7 @@ static int CompareDoxIndexEntry(const void *p1, const void *p2, void *arg) {
 }
 
 static unsigned long roundup2pow(unsigned long x) {
-  return x > 1 ? 2ul << _bsrl(x - 1) : x ? 1 : 0;
+  return x > 1 ? 2ul << bsrl(x - 1) : x ? 1 : 0;
 }
 
 static void IndexDox(struct Dox *dox) {

third_party/chibicc/file.c

@@ -67,7 +67,7 @@ void canonicalize_newline(char *p) {
           p += 16;
           q += 16;
         } else {
-          m = _bsf(m);
+          m = bsf(m);
           memmove(q, p, m);
           p += m;
           q += m;
@@ -126,7 +126,7 @@ void remove_backslash_newline(char *p) {
           i += 16;
           j += 16;
         } else {
-          m = _bsf(m);
+          m = bsf(m);
           memmove(p + j, p + i, m);
           i += m;
           j += m;

third_party/chibicc/test/implicit_test.c

@@ -9,7 +9,7 @@ implicit_functions_are_long() {
 }
 
 external_functions_are_long() {
-  if (_bsrl(0x0001000000000000) != 48) {
+  if (bsrl(0x0001000000000000) != 48) {
     __builtin_trap();
   }
 }

third_party/chibicc/tokenize.c

@@ -680,7 +680,7 @@ static void convert_universal_chars(char *p) {
           p += 16;
           q += 16;
         } else {
-          m = _bsf(m);
+          m = bsf(m);
           memmove(q, p, m);
           p += m;
           q += m;

third_party/dlmalloc/dlmalloc.c

@@ -935,7 +935,7 @@ static void* internal_memalign(mstate m, size_t alignment, size_t bytes) {
   if (alignment <  MIN_CHUNK_SIZE) /* must be at least a minimum chunk size */
     alignment = MIN_CHUNK_SIZE;
   /* alignment is 32+ bytes rounded up to nearest two power */
-  alignment = 2ul << _bsrl(MAX(MIN_CHUNK_SIZE, alignment) - 1);
+  alignment = 2ul << bsrl(MAX(MIN_CHUNK_SIZE, alignment) - 1);
   if (bytes >= MAX_REQUEST - alignment) {
     if (m != 0)  { /* Test isn't needed but avoids compiler warning */
       MALLOC_FAILURE_ACTION;

third_party/linenoise/linenoise.c

@@ -369,7 +369,7 @@ static wint_t Capitalize(wint_t c) {
 static struct rune DecodeUtf8(int c) {
   struct rune r;
   if (c < 252) {
-    r.n = _bsr(255 & ~c);
+    r.n = bsr(255 & ~c);
     r.c = c & (((1 << r.n) - 1) | 3);
     r.n = 6 - r.n;
   } else {
@@ -645,7 +645,7 @@ static void abAppendw(struct abuf *a, unsigned long long w) {
   p[5] = (0x0000ff0000000000 & w) >> 050;
   p[6] = (0x00ff000000000000 & w) >> 060;
   p[7] = (0xff00000000000000 & w) >> 070;
-  a->len += w ? (_bsrll(w) >> 3) + 1 : 1;
+  a->len += w ? (bsrll(w) >> 3) + 1 : 1;
   p[8] = 0;
 }
 
@@ -1634,7 +1634,7 @@ static size_t linenoiseEscape(char *d, const char *s, size_t n) {
         break;
     }
     WRITE32LE(p, w);
-    p += (_bsr(w) >> 3) + 1;
+    p += (bsr(w) >> 3) + 1;
     l = w;
   }
   return p - d;

third_party/lua/lobject.h

@@ -785,7 +785,7 @@ LUAI_FUNC void luaO_chunkid (char *out, const char *source, size_t srclen);
 ** Computes ceil(log2(x))
 */
 static inline int luaO_ceillog2 (unsigned int x) {
-  return --x ? _bsr(x) + 1 : 0;
+  return --x ? bsr(x) + 1 : 0;
 }
 
 #endif

third_party/python/Parser/node.c

@@ -31,7 +31,7 @@ fancy_roundup(int x)
     /* Round up to the closest power of 2 >= n. */
     int r;
     assert(x > 128);
-    r = 1u << (_bsr(x - 1) + 1); /* hacker's delight */
+    r = 1u << (bsr(x - 1) + 1); /* hacker's delight */
     return r > 0 ? r : -1;
 }
 

third_party/xed/x86ild.greg.c

@@ -878,7 +878,7 @@ privileged static void xed_evex_scanner(struct XedDecodedInst *d) {
 }
 
 privileged static uint64_t xed_read_number(uint8_t *p, size_t n, bool s) {
-  switch (s << 2 | _bsr(n)) {
+  switch (s << 2 | bsr(n)) {
     case 0b000:
       return *p;
     case 0b100:

tool/build/lib/panel.c

@@ -32,7 +32,7 @@ static int tpdecode(const char *s, wint_t *out) {
     if ((wc = s[i++] & 255) == -1) return -1;
   }
   if (!(0 <= wc && wc <= 0x7F)) {
-    msb = wc < 252 ? _bsr(~wc & 0xff) : 1;
+    msb = wc < 252 ? bsr(~wc & 0xff) : 1;
     need = 7 - msb;
     wc &= ((1u << msb) - 1) | 0003;
     for (j = 1; j < need; ++j) {

tool/emacs/cosmo-c-builtins.el

@@ -190,6 +190,9 @@
            "__integer_pack"
            "__is_same"))
 
+        (cuda-builtin-types
+         '("__syncthreads"))
+
         (cosmopolitan-builtin-defines
          '("COSMOPOLITAN_CXX_START_"
            "COSMOPOLITAN_CXX_END_"
@@ -225,6 +228,7 @@
                                 gcc-builtin-functions-sync
                                 gcc-builtin-functions-ia32
                                 gxx-builtin-type-traits
+                                cuda-builtin-types
                                 cosmopolitan-builtin-defines
                                 cosmopolitan-builtin-functions))
             "\\_>")))

tool/emacs/cosmo-platform-constants.el

@@ -74,11 +74,13 @@
            "_OPENMP"))
 
         (cuda
-         '("__NVCC__"))
+         '("__NVCC__"
+           "__CUDA_ARCH__"))
 
         (hip
          '("__HIP__"
            "__HIPCC__"
+           "__HIP_PLATFORM_AMD__"
            "__HIP_DEVICE_COMPILE__"
            "__AMD__"
            "__AMDGCN__"

tool/net/lfuncs.c

@@ -342,7 +342,7 @@ int LuaPopcnt(lua_State *L) {
 int LuaBsr(lua_State *L) {
   long x;
   if ((x = luaL_checkinteger(L, 1))) {
-    lua_pushinteger(L, _bsrl(x));
+    lua_pushinteger(L, bsrl(x));
     return 1;
   } else {
     luaL_argerror(L, 1, "zero");
@@ -353,7 +353,7 @@ int LuaBsr(lua_State *L) {
 int LuaBsf(lua_State *L) {
   long x;
   if ((x = luaL_checkinteger(L, 1))) {
-    lua_pushinteger(L, _bsfl(x));
+    lua_pushinteger(L, bsfl(x));
     return 1;
   } else {
     luaL_argerror(L, 1, "zero");

tool/net/redbean.c

@@ -2147,7 +2147,7 @@ static void FreeStrings(struct Strings *l) {
 }
 
 static unsigned long roundup2pow(unsigned long x) {
-  return x > 1 ? 2ul << _bsrl(x - 1) : x ? 1 : 0;
+  return x > 1 ? 2ul << bsrl(x - 1) : x ? 1 : 0;
 }
 
 static void IndexAssets(void) {
@@ -2943,7 +2943,7 @@ static int GetDecimalWidth(long x) {
 }
 
 static int GetOctalWidth(int x) {
-  return !x ? 1 : x < 8 ? 2 : 1 + _bsr(x) / 3;
+  return !x ? 1 : x < 8 ? 2 : 1 + bsr(x) / 3;
 }
 
 static const char *DescribeCompressionRatio(char rb[8], uint8_t *zcf) {

tool/plinko/lib/gc.c

@@ -93,7 +93,7 @@ int Census(struct Gc *G) {
     if (!~G->M[j]) {
       l += DWBITS;
     } else {
-      l += _bsfl(~G->M[j]);
+      l += bsfl(~G->M[j]);
       break;
     }
   }
@@ -124,7 +124,7 @@ void Sweep(struct Gc *G) {
   for (; i < G->n; ++i) {
     m = G->M[i];
     if (~m) {
-      j = _bsfl(~m);
+      j = bsfl(~m);
       m >>= j;
       m <<= j;
       d -= j;
@@ -136,7 +136,7 @@ void Sweep(struct Gc *G) {
   }
   for (; i < G->n; b -= DWBITS, m = G->M[++i]) {
     for (; m; m &= ~((dword)1 << j)) {
-      a = b + ~(j = _bsfl(m));
+      a = b + ~(j = bsfl(m));
       Set(--d, MAKE(Relocate(G, LO(Get(a))), Relocate(G, HI(Get(a)))));
       SetShadow(d, MAKE(LO(GetShadow(a)), Relocate(G, HI(GetShadow(a)))));
     }

tool/plinko/lib/histo.h

@@ -7,7 +7,7 @@ COSMOPOLITAN_C_START_
 #define HISTO(H, X)                \
   do {                             \
     uint64_t x_ = X;               \
-    x_ = x_ ? _bsrl(x_) + 1 : x_;  \
+    x_ = x_ ? bsrl(x_) + 1 : x_;   \
     ++H[MIN(x_, ARRAYLEN(H) - 1)]; \
   } while (0)
 

tool/plinko/lib/printchar.c

@@ -31,7 +31,7 @@ int PrintChar(int fd, int s) {
     if (c == L'\n') Flush(fd);
   } else {
     d = c;
-    e = kTpEnc[_bsrl(d) - 7];
+    e = kTpEnc[bsrl(d) - 7];
     i = n = e & 255;
     do g_buffer[fd][bp[fd] + i--] = 0200 | (d & 077);
     while (d >>= 6, i);

tool/viz/fontspace.c

@@ -51,7 +51,7 @@ void PrintBar(unsigned char *p, size_t n) {
   }
   for (i = 0; i < j; ++i) {
     if (p[i]) {
-      fputwc(u"░░▒▒▓▓██"[_bsr(p[i])], stdout);
+      fputwc(u"░░▒▒▓▓██"[bsr(p[i])], stdout);
     } else {
       fputc(' ', stdout);
     }

tool/viz/lib/bilinearscale.c

@@ -82,8 +82,8 @@ void *BilinearScale(long dcw, long dyw, long dxw,
       DCHECK_LE(sxn, sxw);
       DCHECK_LE(dyn, dyw);
       DCHECK_LE(dxn, dxw);
-      DCHECK_LT(_bsrl(cn) + _bsrl(syn) + _bsrl(sxn), 32);
-      DCHECK_LT(_bsrl(cn) + _bsrl(dyn) + _bsrl(dxn), 32);
+      DCHECK_LT(bsrl(cn) + bsrl(syn) + bsrl(sxn), 32);
+      DCHECK_LT(bsrl(cn) + bsrl(dyn) + bsrl(dxn), 32);
       BilinearScaler(dcw, dyw, dxw, dst, scw, syw, sxw, src, c0, cn, dyn, dxn,
                      syn, sxn, ry, rx, oy, ox,
                      gc(xmemalign(64, ROUNDUP(sizeof(int) * (dyn + 1), 64))),

tool/viz/lib/ycbcr2rgb3.c

@@ -84,11 +84,11 @@ struct YCbCr {
 };
 
 static unsigned long roundup2pow(unsigned long x) {
-  return x > 1 ? 2ul << _bsrl(x - 1) : x ? 1 : 0;
+  return x > 1 ? 2ul << bsrl(x - 1) : x ? 1 : 0;
 }
 
 static unsigned long rounddown2pow(unsigned long x) {
-  return x ? 1ul << _bsrl(x) : 0;
+  return x ? 1ul << bsrl(x) : 0;
 }
 
 /**

tool/viz/memzoom.c

@@ -190,7 +190,7 @@ static void LeaveScreen(void) {
 }
 
 static unsigned long rounddown2pow(unsigned long x) {
-  return x ? 1ul << _bsrl(x) : 0;
+  return x ? 1ul << bsrl(x) : 0;
 }
 
 static void GetTtySize(void) {