diff --git a/backends/cxxrtl/cxxrtl.h b/backends/cxxrtl/cxxrtl.h
index 5d0596f0d..6d955fe9b 100644
--- a/backends/cxxrtl/cxxrtl.h
+++ b/backends/cxxrtl/cxxrtl.h
@@ -518,6 +518,14 @@ struct value : public expr_base<value<Bits>> {
 		return count;
 	}
 
+	size_t chunks_used() const {
+		for (size_t n = chunks; n > 0; n--) {
+			if (data[n - 1] != 0)
+				return n;
+		}
+		return 0;
+	}
+
 	template<bool Invert, bool CarryIn>
 	std::pair<value<Bits>, bool /*CarryOut*/> alu(const value<Bits> &other) const {
 		value<Bits> result;
@@ -575,6 +583,140 @@ struct value : public expr_base<value<Bits>> {
 		result.data[result.chunks - 1] &= result.msb_mask;
 		return result;
 	}
+
+	// parallel to BigUnsigned::divideWithRemainder; quotient is stored in q,
+	// *this is left with the remainder.  See that function for commentary describing
+	// how/why this works.
+	void divideWithRemainder(const value<Bits> &b, value<Bits> &q) {
+		assert(this != &q);
+
+		if (this == &b || &q == &b) {
+			value<Bits> tmpB(b);
+			divideWithRemainder(tmpB, q);
+			return;
+		}
+
+		q = value<Bits> {0u};
+
+		size_t blen = b.chunks_used();
+		if (blen == 0) {
+			return;
+		}
+
+		size_t len = chunks_used();
+		if (len < blen) {
+			return;
+		}
+
+		size_t i, j, k;
+		size_t i2;
+		chunk_t temp;
+		bool borrowIn, borrowOut;
+
+		size_t origLen = len;
+		len++;
+		chunk::type blk[len];
+		std::copy(data, data + origLen, blk);
+		blk[origLen] = 0;
+		chunk::type subtractBuf[len];
+		std::fill(subtractBuf, subtractBuf + len, 0);
+
+		size_t qlen = origLen - blen + 1;
+
+		i = qlen;
+		while (i > 0) {
+			i--;
+			i2 = chunk::bits;
+			while (i2 > 0) {
+				i2--;
+				for (j = 0, k = i, borrowIn = false; j <= blen; j++, k++) {
+					temp = blk[k] - getShiftedBlock(b, j, i2);
+					borrowOut = (temp > blk[k]);
+					if (borrowIn) {
+						borrowOut |= (temp == 0);
+						temp--;
+					}
+					subtractBuf[k] = temp;
+					borrowIn = borrowOut;
+				}
+				for (; k < origLen && borrowIn; k++) {
+					borrowIn = (blk[k] == 0);
+					subtractBuf[k] = blk[k] - 1;
+				}
+				if (!borrowIn) {
+					q.data[i] |= (chunk::type(1) << i2);
+					while (k > i) {
+						k--;
+						blk[k] = subtractBuf[k];
+					}
+				}
+			}
+		}
+
+		std::copy(blk, blk + origLen, data);
+		std::fill(blk + origLen, blk + chunks, 0);
+	}
+
+	static chunk::type getShiftedBlock(const value<Bits> &num, size_t x, size_t y) {
+		chunk::type part1 = (x == 0 || y == 0) ? 0 : (num.data[x - 1] >> (chunk::bits - y));
+		chunk::type part2 = (x == num.chunks) ? 0 : (num.data[x] << y);
+		return part1 | part2;
+	}
+
+	// parallel to BigInteger::divideWithRemainder; quotient is stored in q,
+	// *this is left with the remainder.  See that function for commentary describing
+	// how/why this works.
+	void signedDivideWithRemainder(const value<Bits> &b, value<Bits> &q) {
+		assert(this != &q);
+
+		if (this == &b || &q == &b) {
+			value<Bits> tmpB(b);
+			signedDivideWithRemainder(tmpB, q);
+			return;
+		}
+
+		if (b.is_zero()) {
+			q = value<Bits>{0u};
+			return;
+		}
+
+		if (is_zero()) {
+			q = value<Bits>{0u};
+			return;
+		}
+
+		// BigInteger has a separate 'mag' to its sign.  We don't, so the lazy
+		// approach is to improvise said.
+		auto mag = *this;
+		bool neg = mag.is_neg();
+		if (neg)
+			mag = mag.neg();
+
+		auto bmag = b;
+		bool bneg = bmag.is_neg();
+		if (bneg)
+			bmag = bmag.neg();
+
+		bool qneg = false;
+		if (neg != b.is_neg()) {
+			qneg = true;
+			mag = mag.sub(value<Bits>{1u});
+		}
+
+		mag.divideWithRemainder(bmag, q);
+
+		if (neg != bneg) {
+			q = q.add(value<Bits>{1u});
+			mag = bmag.sub(mag);
+			mag = mag.sub(value<Bits>{1u});
+		}
+
+		neg = bneg;
+
+		*this = neg ? mag.neg() : mag;
+		if (qneg)
+			q = q.neg();
+	}
 };
 
 // Expression template for a slice, usable as lvalue or rvalue, and composable with other expression templates here.
@@ -707,6 +849,101 @@ std::ostream &operator<<(std::ostream &os, const value<Bits> &val) {
 	return os;
 }
 
+template<size_t Bits>
+struct value_formatted {
+	const value<Bits> &val;
+	bool character;
+	bool justify_left;
+	char padding;
+	int width;
+	int base;
+	bool signed_;
+	bool lzero;
+	bool plus;
+
+	value_formatted(const value<Bits> &val, bool character, bool justify_left, char padding, int width, int base, bool signed_, bool lzero, bool plus) :
+		val(val), character(character), justify_left(justify_left), padding(padding), width(width), base(base), signed_(signed_), lzero(lzero), plus(plus) {}
+	value_formatted(const value_formatted<Bits> &) = delete;
+	value_formatted<Bits> &operator=(const value_formatted<Bits> &rhs) = delete;
+};
+
+template<size_t Bits>
+std::ostream &operator<<(std::ostream &os, const value_formatted<Bits> &vf)
+{
+	value<Bits> val = vf.val;
+
+	std::string buf;
+
+	// We might want to replace some of these bit() calls with direct
+	// chunk access if it turns out to be slow enough to matter.
+
+	if (!vf.character) {
+		size_t width = Bits;
+		if (!vf.lzero && vf.base != 10) {
+			width = 0;
+			for (size_t index = 0; index < Bits; index++)
+				if (val.bit(index))
+					width = index + 1;
+		}
+
+		if (vf.base == 2) {
+			for (size_t i = width; i > 0; i--)
+				buf += (val.bit(i - 1) ? '1' : '0');
+		} else if (vf.base == 8 || vf.base == 16) {
+			size_t step = (vf.base == 16) ? 4 : 3;
+			for (size_t index = 0; index < width; index += step) {
+				uint8_t value = val.bit(index) | (val.bit(index + 1) << 1) | (val.bit(index + 2) << 2);
+				if (step == 4)
+					value |= val.bit(index + 3) << 3;
+				buf += "0123456789abcdef"[value];
+			}
+			std::reverse(buf.begin(), buf.end());
+		} else if (vf.base == 10) {
+			bool negative = vf.signed_ && val.is_neg();
+			if (negative)
+				val = val.neg();
+			if (val.is_zero())
+				buf += '0';
+			// TODO: rigorously check our signed behaviour here
+			while (!val.is_zero()) {
+				value<Bits> quotient;
+				val.signedDivideWithRemainder(value<Bits>{10u}, quotient);
+				buf += '0' + val.template slice<3, 0>().val().template get<uint8_t>();
+				val = quotient;
+			}
+			if (negative || vf.plus)
+				buf += negative ? '-' : '+';
+			std::reverse(buf.begin(), buf.end());
+		} else assert(false);
+	} else {
+		buf.reserve(Bits/8);
+		for (int i = 0; i < Bits; i += 8) {
+			char ch = 0;
+			for (int j = 0; j < 8 && i + j < int(Bits); j++)
+				if (val.bit(i + j))
+					ch |= 1 << j;
+			if (ch != 0)
+				buf.append({ch});
+		}
+		std::reverse(buf.begin(), buf.end());
+	}
+
+	assert(vf.width == 0 || vf.padding != '\0');
+	if (!vf.justify_left && buf.size() < vf.width) {
+		size_t pad_width = vf.width - buf.size();
+		if (vf.padding == '0' && (buf.front() == '+' || buf.front() == '-')) {
+			os << buf.front();
+			buf.erase(0, 1);
+		}
+		os << std::string(pad_width, vf.padding);
+	}
+	os << buf;
+	if (vf.justify_left && buf.size() < vf.width)
+		os << std::string(vf.width - buf.size(), vf.padding);
+
+	return os;
+}
+
 template<size_t Bits>
 struct wire {
 	static constexpr size_t bits = Bits;
diff --git a/backends/cxxrtl/cxxrtl_backend.cc b/backends/cxxrtl/cxxrtl_backend.cc
index 1b13985ab..4986836b9 100644
--- a/backends/cxxrtl/cxxrtl_backend.cc
+++ b/backends/cxxrtl/cxxrtl_backend.cc
@@ -24,6 +24,7 @@
 #include "kernel/celltypes.h"
 #include "kernel/mem.h"
 #include "kernel/log.h"
+#include "kernel/fmt.h"
 
 USING_YOSYS_NAMESPACE
 PRIVATE_NAMESPACE_BEGIN
@@ -217,7 +218,7 @@ bool is_internal_cell(RTLIL::IdString type)
 
 bool is_effectful_cell(RTLIL::IdString type)
 {
-	return type.isPublic();
+	return type.isPublic() || type == ID($print);
 }
 
 bool is_cxxrtl_blackbox_cell(const RTLIL::Cell *cell)
@@ -1036,6 +1037,17 @@ struct CxxrtlWorker {
 			f << ".val()";
 	}
 
+	void dump_print(const RTLIL::Cell *cell)
+	{
+		Fmt fmt = {};
+		fmt.parse_rtlil(cell);
+
+		// TODO: we may want to configure the output stream
+		f << indent << "std::cout";
+		fmt.emit_cxxrtl(f, [this](const RTLIL::SigSpec &sig) { dump_sigspec_rhs(sig); });
+		f << ";\n";
+	}
+
 	void dump_inlined_cells(const std::vector<const RTLIL::Cell*> &cells)
 	{
 		if (cells.empty()) {
@@ -1202,6 +1214,34 @@ struct CxxrtlWorker {
 			f << " = ";
 			dump_cell_expr(cell, for_debug);
 			f << ";\n";
+		// $print cell
+		} else if (cell->type == ID($print)) {
+			log_assert(!for_debug);
+			f << indent << "if (";
+			if (cell->getParam(ID::TRG_ENABLE).as_bool()) {
+				f << '(';
+				for (size_t i = 0; i < (size_t)cell->getParam(ID::TRG_WIDTH).as_int(); i++) {
+					RTLIL::SigBit trg_bit = cell->getPort(ID::TRG)[i];
+					trg_bit = sigmaps[trg_bit.wire->module](trg_bit);
+					log_assert(trg_bit.wire);
+
+					if (i != 0)
+						f << " || ";
+
+					if (cell->getParam(ID::TRG_POLARITY)[i] == State::S1)
+						f << "posedge_";
+					else
+						f << "negedge_";
+					f << mangle(trg_bit);
+				}
+				f << ") && ";
+			}
+			dump_sigspec_rhs(cell->getPort(ID::EN));
+			f << " == value<1>{1u}) {\n";
+			inc_indent();
+				dump_print(cell);
+			dec_indent();
+			f << indent << "}\n";
 		// Flip-flops
 		} else if (is_ff_cell(cell->type)) {
 			log_assert(!for_debug);
@@ -2601,6 +2641,16 @@ struct CxxrtlWorker {
 						register_edge_signal(sigmap, cell->getPort(ID::CLK),
 							cell->parameters[ID::CLK_POLARITY].as_bool() ? RTLIL::STp : RTLIL::STn);
 				}
+
+				// $print cells may be triggered on posedge/negedge events.
+				if (cell->type == ID($print) && cell->getParam(ID::TRG_ENABLE).as_bool()) {
+					for (size_t i = 0; i < (size_t)cell->getParam(ID::TRG_WIDTH).as_int(); i++) {
+						RTLIL::SigBit trg = cell->getPort(ID::TRG).extract(i, 1);
+						if (is_valid_clock(trg))
+							register_edge_signal(sigmap, trg,
+								cell->parameters[ID::TRG_POLARITY][i] == RTLIL::S1 ? RTLIL::STp : RTLIL::STn);
+					}
+				}
 			}
 
 			for (auto &mem : memories) {
diff --git a/kernel/fmt.cc b/kernel/fmt.cc
index 41de49392..074ec08ca 100644
--- a/kernel/fmt.cc
+++ b/kernel/fmt.cc
@@ -29,7 +29,7 @@ void Fmt::append_string(const std::string &str) {
 	parts.push_back(part);
 }
 
-void Fmt::parse_rtlil(RTLIL::Cell *cell) {
+void Fmt::parse_rtlil(const RTLIL::Cell *cell) {
 	std::string fmt = cell->getParam(ID(FORMAT)).decode_string();
 	RTLIL::SigSpec args = cell->getPort(ID(ARGS));
 	parts.clear();
@@ -465,6 +465,67 @@ std::vector<VerilogFmtArg> Fmt::emit_verilog() const
 	return args;
 }
 
+void Fmt::emit_cxxrtl(std::ostream &f, std::function<void(const RTLIL::SigSpec &)> emit_sig) const
+{
+	for (auto &part : parts) {
+		switch (part.type) {
+			case FmtPart::STRING:
+				f << " << \"";
+				for (char c : part.str) {
+					switch (c) {
+						case '\\':
+							YS_FALLTHROUGH
+						case '"':
+							f << '\\' << c;
+							break;
+						case '\a':
+							f << "\\a";
+							break;
+						case '\b':
+							f << "\\b";
+							break;
+						case '\f':
+							f << "\\f";
+							break;
+						case '\n':
+							f << "\\n";
+							break;
+						case '\r':
+							f << "\\r";
+							break;
+						case '\t':
+							f << "\\t";
+							break;
+						case '\v':
+							f << "\\v";
+							break;
+						default:
+							f << c;
+							break;
+					}
+				}
+				f << '"';
+				break;
+
+			case FmtPart::INTEGER:
+			case FmtPart::CHARACTER: {
+				f << " << value_formatted<" << part.sig.size() << ">(";
+				emit_sig(part.sig);
+				f << ", " << (part.type == FmtPart::CHARACTER);
+				f << ", " << (part.justify == FmtPart::LEFT);
+				f << ", (char)" << (int)part.padding;
+				f << ", " << part.width;
+				f << ", " << part.base;
+				f << ", " << part.signed_;
+				f << ", " << part.lzero;
+				f << ", " << part.plus;
+				f << ')';
+				break;
+			}
+		}
+	}
+}
+
 std::string Fmt::render() const
 {
 	std::string str;
diff --git a/kernel/fmt.h b/kernel/fmt.h
index c85da19c0..00b8d9b59 100644
--- a/kernel/fmt.h
+++ b/kernel/fmt.h
@@ -76,12 +76,14 @@ struct Fmt {
 
 	void append_string(const std::string &str);
 
-	void parse_rtlil(RTLIL::Cell *cell);
+	void parse_rtlil(const RTLIL::Cell *cell);
 	void emit_rtlil(RTLIL::Cell *cell) const;
 
 	void parse_verilog(const std::vector<VerilogFmtArg> &args, bool sformat_like, int default_base, RTLIL::IdString task_name, RTLIL::IdString module_name);
 	std::vector<VerilogFmtArg> emit_verilog() const;
 
+	void emit_cxxrtl(std::ostream &f, std::function<void(const RTLIL::SigSpec &)> emit_sig) const;
+
 	std::string render() const;
 };