Add POPCOUNT function (#2166)

include/helpers.hpp

@@ -42,12 +42,13 @@ static inline void unreachable_() {
 
 // Ideally we'd use `std::bit_width`, but it has insufficient compiler support
 #ifdef __GNUC__ // GCC or compatible
-	#define ctz __builtin_ctz
-	#define clz __builtin_clz
+	#define ctz      __builtin_ctz
+	#define clz      __builtin_clz
+	#define popcount __builtin_popcount
 
 #elif defined(_MSC_VER)
 	#include <intrin.h>
-	#pragma intrinsic(_BitScanReverse, _BitScanForward)
+	#pragma intrinsic(_BitScanReverse, _BitScanForward, __popcnt)
 
 static inline int ctz(unsigned int x) {
 	assume(x != 0);
@@ -63,6 +64,10 @@ static inline int clz(unsigned int x) {
 	return 31 - count;
 }
 
+static inline int popcount(unsigned int x) {
+	return __popcnt(x);
+}
+
 #else
 	#include <limits.h>
 
@@ -85,6 +90,15 @@ static inline int clz(unsigned int x) {
 	}
 	return count;
 }
+
+static inline int popcount(unsigned int x) {
+	int count = 0;
+	while (x) {
+		x &= x - 1;
+		++count;
+	}
+	return count;
+}
 #endif
 
 // Macros for stringification

include/linkdefs.hpp

@@ -57,6 +57,7 @@ enum RPNCommand {
 	RPN_LOW = 0x71,
 	RPN_BITWIDTH = 0x72,
 	RPN_TZCOUNT = 0x73,
+	RPN_POPCOUNT = 0x74,
 
 	RPN_CONST = 0x80,
 	RPN_SYM = 0x81

include/opmath.hpp

@@ -23,5 +23,6 @@ int32_t op_low(int32_t value);
 
 int32_t op_bitwidth(int32_t value);
 int32_t op_tzcount(int32_t value);
+int32_t op_popcount(int32_t value);
 
 #endif // RGBDS_OP_MATH_HPP

man/rgbasm.5

@@ -436,6 +436,8 @@ equals $roman clz ( n )$, the count of leading zero bits in the binary represent
 .Ar n .
 .It Fn TZCOUNT n Ta Returns $roman ctz ( n )$, the count of trailing zero bits in the binary representation of
 .Ar n .
+.It Fn POPCOUNT n Ta Returns how many one bits are in the binary representation of
+.Ar n .
 .El
 .EQ
 delim off

man/rgbds.5

@@ -411,6 +411,8 @@ byte.
 value.
 .It Li $73 Ta Cm TZCOUNT
 value.
+.It Li $74 Ta Cm POPCOUNT
+value.
 .It Li $80 Ta Integer literal; followed by the
 .Cm LONG
 integer.

src/asm/lexer.cpp

@@ -173,6 +173,7 @@ static UpperMap<int> const keywords{
     {"ISCONST",       T_(OP_ISCONST)       },
 
     {"BITWIDTH",      T_(OP_BITWIDTH)      },
+    {"POPCOUNT",      T_(OP_POPCOUNT)      },
     {"TZCOUNT",       T_(OP_TZCOUNT)       },
 
     {"BYTELEN",       T_(OP_BYTELEN)       },

src/asm/parser.y

@@ -286,6 +286,7 @@
 %token OP_ISCONST "ISCONST"
 %token OP_LOG "LOG"
 %token OP_LOW "LOW"
+%token OP_POPCOUNT "POPCOUNT"
 %token OP_POW "POW"
 %token OP_READFILE "READFILE"
 %token OP_REVCHAR "REVCHAR"
@@ -1396,6 +1397,9 @@ relocexpr_no_str:
 	| OP_BITWIDTH LPAREN relocexpr RPAREN {
 		$$.makeUnaryOp(RPN_BITWIDTH, std::move($3));
 	}
+	| OP_POPCOUNT LPAREN relocexpr RPAREN {
+		$$.makeUnaryOp(RPN_POPCOUNT, std::move($3));
+	}
 	| OP_TZCOUNT LPAREN relocexpr RPAREN {
 		$$.makeUnaryOp(RPN_TZCOUNT, std::move($3));
 	}

src/asm/rpn.cpp

@@ -279,6 +279,9 @@ void Expression::makeUnaryOp(RPNCommand op, Expression &&src) {
 		case RPN_TZCOUNT:
 			data = op_tzcount(val);
 			break;
+		case RPN_POPCOUNT:
+			data = op_popcount(val);
+			break;
 		// LCOV_EXCL_START
 		default:
 			// `makeUnaryOp` should never be called with a non-unary operator!

src/link/patch.cpp

@@ -182,6 +182,9 @@ static int32_t computeRPNExpr(Patch const &patch, std::vector<Symbol> const &fil
 		case RPN_TZCOUNT:
 			value = op_tzcount(popRPN(patch));
 			break;
+		case RPN_POPCOUNT:
+			value = op_popcount(popRPN(patch));
+			break;
 
 		case RPN_OR:
 			value = popRPN(patch) | popRPN(patch);

src/opmath.cpp

@@ -141,3 +141,7 @@ int32_t op_bitwidth(int32_t value) {
 int32_t op_tzcount(int32_t value) {
 	return value != 0 ? ctz(static_cast<uint32_t>(value)) : 32;
 }
+
+int32_t op_popcount(int32_t value) {
+	return popcount(static_cast<uint32_t>(value));
+}

test/asm/bit-functions.asm

@@ -9,3 +9,9 @@ assert TZCOUNT(-1) == 0
 assert TZCOUNT($80000000) == 31
 
 assert TZCOUNT(1.0) == 16
+
+assert POPCOUNT(0) == 0
+assert POPCOUNT(1) == 1
+assert POPCOUNT(-1) == 32
+assert POPCOUNT(%10101100) == 4
+assert POPCOUNT(-%10101100) == 27

test/link/operators.asm

@@ -30,3 +30,4 @@ dw @ >>> 1
 
 db BITWIDTH(@)
 db TZCOUNT(@)
+db POPCOUNT(@)