Use __builtin_ctz and __builtin_clz in dlmalloc

[binji]

These operations map to the WebAssembly instructions i32.ctz and i32.clz. Without this, the various macros fall back to a C implementation which does not seem to be detected as ctz or clz operations by clang at -O3.

[kripken]

Nice!

[kleisauke]

Could the compute_tree_index macro use something similar? IIUC, that would ensure the use of i32.clz.

emscripten/system/lib/dlmalloc.c

Lines 2920 to 2932 in 39a7a63

	#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
	#define compute_tree_index(S, I)\
	{\
	unsigned int X = S >> TREEBIN_SHIFT;\
	if (X == 0)\
	I = 0;\
	else if (X > 0xFFFF)\
	I = NTREEBINS-1;\
	else {\
	unsigned int K = (unsigned) sizeof(X)*__CHAR_BIT__ - 1 - (unsigned) __builtin_clz(X); \
	I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
	}\
	}

Otherwise, it falls back on this variant:

emscripten/system/lib/dlmalloc.c

Lines 2964 to 2980 in 39a7a63

	#define compute_tree_index(S, I)\
	{\
	size_t X = S >> TREEBIN_SHIFT;\
	if (X == 0)\
	I = 0;\
	else if (X > 0xFFFF)\
	I = NTREEBINS-1;\
	else {\
	unsigned int Y = (unsigned int)X;\
	unsigned int N = ((Y - 0x100) >> 16) & 8;\
	unsigned int K = (((Y <<= N) - 0x1000) >> 16) & 4;\
	N += K;\
	N += K = (((Y <<= K) - 0x4000) >> 16) & 2;\
	K = 14 - N + ((Y <<= K) >> 15);\
	I = (K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1));\
	}\
	}

(though, perhaps the compiler might already optimize this, untested)

[binji]

@kleisauke ah, good catch! I'll add that in to this commit

[kripken]

Wow, nice improvements!

[sbc100]

I wonder this doesn't use #if __has_builtin(__builtin_ctz) .. I guess maybe dlmalloc was written before such a macro existed?

I also wonder if we should use __wasm__ here instead of __EMSCRIPTEN__? I guess it doesn't matter much since if there is almost no chance of upstreaming these chagnes.

[binji]

Yeah, __wasm__ probably is a better choice; I used __EMSCRIPTEN__ since that was used by other modifications in that file.