start experimenting with processor grids

29 files changed, 819 insertions, 99 deletions

diff --git a/Makefile b/Makefile
index 88855fa..108c182 100644
--- a/Makefile
+++ b/Makefile
@@ -1,7 +1,7 @@
 DO	!= echo -n > deps.mk
 
 DEBUGFLAGS	!= [ $(RELEASE) ] && echo "-flto=auto -O2 -DNODEBUG" || echo "-O0 -g -DDEBUG"
-CFLAGS		= -Wall -Wextra -fopenmp -g
+CFLAGS		= -Wall -Wextra -g
 DEPFLAGS	= -MT $@ -MMD -MP -MF $@.d
 LINTFLAGS	= -fsyntax-only
 INCLUDEFLAGS	= -Iinclude
diff --git a/README.md b/README.md
index 6bf8707..a50938b 100644
--- a/README.md
+++ b/README.md
@@ -1,15 +1,53 @@
 # gran
 
-Gran is a toy granular full system simulator,
-taking heavy inspiration from [gem5](https://www.gem5.org/).
+Gran is a toy granular full system simulator, taking heavy inspiration from
+[gem5](https://www.gem5.org/).
 
-Currently there are a couple example components, including
-`simple_mem` for atomic memory accesses, `simple_bus` for non-coherent
-atomic transfers between components and `simple_riscv32`, a simple
-rv32i processor.
+Currently there are a couple example components, including `simple_mem` for
+atomic memory accesses, `simple_bus` for non-coherent atomic transfers between
+components and `simple_riscv32`, a simple rv32i processor.
 
 # Building
 
 + `make`
 
 Add `RELEASE=1` to enable optimizations.
+
+# manycore
+
+Idea: Have a grid of x/y width, where each node has three 'axes', i.e. if the
+address matches the node, pass the message to whatever is at that node (think of
+it like a miniature system I suppose?), otherwise check x, send it in that
+direction, otherwise check y, send it in that direction, otherwise check if
+message should be transcended, i.e. sent to some other cluster.
+
+Each address is 64 bits, split in half. The lower 32 bits is the offset within
+the node, and the upper 32 bits are split (again) in twine, something like x1 y1
+x2 y2 where each x or y is eight bits, specifying a node in an x/y grid. x2/y2
+specifies the lower cluster, with x1/y1 upper cluster. The idea is that a
+complete system would consist of cards that contain one cluster, clustered
+together. So a 'supercomputer' could be up to 65536 cards, with each card having
+a cluster of 65536 nodes. Each node is a compute unit, probably a single core,
+possibly with associated memory or devices like UART, whatever.
+
+Currently I'm imagining that each node also has a control region that can be
+written to with commands like 'drop all cache entries that containt address xyz'
+or 'wake up' (to enable userspace mutexes, that would be massively cool)
+
+The current focus would be to enable lightweight userspaces, full-blown
+hardware-assisted virtual memory is costly to implement (as in takes up too much
+space in hardware), but I'd still want to be able to run a command line and
+generic programs on the system. One idea would be a capability-based system,
+sort of like CHERIoT, though it has some drawbacks like having to keep track of
+the valid capabilities in separate memory or as part of ECC.
+
+Another approach that I'm currently thinking of is having a *very* simple
+region-based memory protection scheme, where instead of paging we have a TLB of,
+dunno, 64 regions that the core is allowed to access. If the core tries to
+access a region outside of those regions, the kernel is trapped to and can then
+insert the requested area into the regions or kill the process or whatever. This
+is *similar* to virtual memory in other systems (particularly MIPS), but would
+be range-based rather than page-based. If I wanted to use OpenASIP, I probably
+wouldn't be able to use interrupts, and we would either need a separate
+coprocessor to handle these kinds of exceptions or implement a tree walker in
+hardware (unsure how much extra area it would take)
diff --git a/include/gran/bus/simple_bus.h b/include/gran/bus/simple_bus.h
index 41a8d2e..62dd7d2 100644
--- a/include/gran/bus/simple_bus.h
+++ b/include/gran/bus/simple_bus.h
@@ -8,6 +8,6 @@
 
 struct component *create_simple_bus();
 stat simple_bus_add(struct component *bus, struct component *component,
-                    uintptr_t addr, size_t size);
+                    uint64_t addr, uint64_t size);
 
 #endif /* GRAN_SIMPLE_BUS_H */
diff --git a/include/gran/cpu/riscv/simple_riscv32.h b/include/gran/cpu/riscv/simple_riscv64.h
index 3cbd842..8f66013 100644
--- a/include/gran/cpu/riscv/simple_riscv32.h
+++ b/include/gran/cpu/riscv/simple_riscv64.h
@@ -1,13 +1,15 @@
 /* SPDX-License-Identifier: copyleft-next-0.3.1 */
 /* Copyright 2023 Kim Kuparinen < kimi.h.kuparinen@gmail.com > */
 
-#ifndef GRAN_SIMPLE_RISCV32
-#define GRAN_SIMPLE_RISCV32
+#ifndef GRAN_SIMPLE_RISCV64
+#define GRAN_SIMPLE_RISCV64
 
 #include <gran/component.h>
 
-struct component *create_simple_riscv32(uint32_t start_pc,
+struct component *create_simple_riscv64(uint32_t start_pc,
                                         struct component *imem,
                                         struct component *dmem);
 
+void simple_riscv64_set_reg(struct component *cpu, size_t reg, uint64_t val);
+
 #endif /* GRAN_SIMPLE_RISCV */
diff --git a/include/gran/grid/node.h b/include/gran/grid/node.h
new file mode 100644
index 0000000..637ac91
--- /dev/null
+++ b/include/gran/grid/node.h
@@ -0,0 +1,14 @@
+#ifndef GRAN_GRID_NODE_H
+#define GRAN_GRID_NODE_H
+
+#include <gran/component.h>
+#include <stdint.h>
+
+struct component *create_grid_node(uint8_t u, uint8_t v, uint8_t x, uint8_t y);
+
+stat grid_node_connect(struct component *node,
+		struct component *left, struct component *right,
+		struct component *up, struct component *down,
+		struct component *lower, struct component *ascend);
+
+#endif /* GRAN_GRID_NODE_H */
diff --git a/include/gran/grid/router.h b/include/gran/grid/router.h
new file mode 100644
index 0000000..cff0fa1
--- /dev/null
+++ b/include/gran/grid/router.h
@@ -0,0 +1,10 @@
+#ifndef GRAN_NODE_ROUTER_H
+#define GRAN_NODE_ROUTER_H
+
+#include <gran/component.h>
+
+struct component *create_node_router(uint8_t u, uint8_t v, uint8_t x, uint8_t y);
+stat node_router_add(struct component *router, struct component *component, uint32_t addr, uint32_t size);
+stat node_router_ascend(struct component *router, struct component *node);
+
+#endif /* GRAN_NODE_ROUTER_H */
diff --git a/include/gran/packet.h b/include/gran/packet.h
index d18dfd6..ab07b56 100644
--- a/include/gran/packet.h
+++ b/include/gran/packet.h
@@ -39,7 +39,7 @@ enum packet_state packet_state(struct packet *pkt);
 void packet_set_state(struct packet *pkt, enum packet_state state);
 
 size_t packet_size(struct packet *pkt);
-uintptr_t packet_addr(struct packet *pkt);
+uint64_t packet_addr(struct packet *pkt);
 void *packet_data(struct packet *pkt);
 
 void destroy_packet(struct packet *pkt);
diff --git a/include/gran/uart/simple_uart.h b/include/gran/uart/simple_uart.h
new file mode 100644
index 0000000..75906fa
--- /dev/null
+++ b/include/gran/uart/simple_uart.h
@@ -0,0 +1,8 @@
+#ifndef GRAN_SIMPLE_UART_H
+#define GRAN_SIMPLE_UART_H
+
+#include <gran/component.h>
+
+struct component *create_simple_uart();
+
+#endif /* GRAN_SIMPLE_UART_H */
diff --git a/include/gran/vec.h b/include/gran/vec.h
new file mode 100644
index 0000000..08a608a
--- /dev/null
+++ b/include/gran/vec.h
@@ -0,0 +1,46 @@
+/* SPDX-License-Identifier: copyleft-next-0.3.1 */
+
+#ifndef VEC_H
+#define VEC_H
+
+#include <stddef.h>
+
+struct vec {
+	size_t n;
+	size_t s;
+	size_t ns;
+	void *buf;
+};
+
+struct vec vec_create(size_t s);
+void vec_destroy(struct vec *v);
+void vec_reset(struct vec *v);
+
+size_t vec_len(struct vec *v);
+void *vec_at(struct vec *v, size_t i);
+void *vec_back(struct vec *v);
+void *vec_pop(struct vec *v);
+void vec_append(struct vec *v, void *n);
+
+typedef int (*vec_comp_t)(const void *, const void *);
+void vec_sort(struct vec *v, vec_comp_t comp);
+
+#define foreach_vec(iter, v) \
+	for (size_t iter = 0; iter < vec_len(&v); ++iter)
+
+#define vect_at(type, v, i) \
+	*(type *)vec_at(&v, i)
+
+#define vect_append(type, v, e) \
+	vec_append(&v, (type *)(e))
+
+#define vect_back(type, v) \
+	*(type *)vec_back(&v)
+
+#define vect_pop(type, v) \
+	*(type *)vec_pop(&v)
+
+#define vec_uninit(v) \
+	(v.buf == NULL)
+
+#endif /* VEC_H */
diff --git a/src/clock_domain.c b/src/clock_domain.c
index 65dc72a..6610d8f 100644
--- a/src/clock_domain.c
+++ b/src/clock_domain.c
@@ -12,12 +12,10 @@
 #include <assert.h>
 
 #include <gran/clock_domain.h>
-
-#define MAX_COMPONENTS 512
+#include <gran/vec.h>
 
 struct clock_domain {
-	size_t num_components;
-	struct component *components[MAX_COMPONENTS];
+	struct vec components;
 	struct clock_time time;
 	tick period;
 
@@ -44,21 +42,23 @@ struct clock_domain *create_clock_domain(tick period)
 
 	clk->period = period;
 	mtx_init(&clk->mtx, mtx_plain);
+	clk->components = vec_create(sizeof(struct component *));
 
 	return clk;
 }
 
 void destroy_clock_domain(struct clock_domain *clk)
 {
-	for (size_t i = 0; i < clk->num_components; ++i)
-		destroy(clk->components[i]);
+	for (size_t i = 0; i < vec_len(&clk->components); ++i)
+		destroy(vect_at(struct component *, clk->components, i));
 
+	vec_destroy(&clk->components);
 	free(clk);
 }
 
 stat clock_domain_add(struct clock_domain *clk, struct component *component)
 {
-	clk->components[clk->num_components++] = component;
+	vect_append(struct component *, clk->components, &component);
 	return OK;
 }
 
@@ -80,18 +80,11 @@ static void clocked_component_tick(struct clock_domain *clk,
 
 stat clock_domain_tick(struct clock_domain *clk)
 {
-	/* openmp on gcc apparently doesn't really handle if conditionals
-	 * particularly efficiently, so it's faster to do it manually */
-	if (clk->num_components == 1) {
-		clocked_component_tick(clk, clk->components[0]);
-		advance_clock(clk);
-		return clk->ret;
+	for (size_t i = 0; i < vec_len(&clk->components); ++i) {
+		clocked_component_tick(clk,
+				vect_at(struct component *, clk->components, i));
 	}
 
-#pragma omp parallel for
-	for (size_t i = 0; i < clk->num_components; ++i)
-		clocked_component_tick(clk, clk->components[i]);
-
 	advance_clock(clk);
 	return clk->ret;
 }
diff --git a/src/components/bus/simple_bus.c b/src/components/bus/simple_bus.c
index 440ff8a..218e930 100644
--- a/src/components/bus/simple_bus.c
+++ b/src/components/bus/simple_bus.c
@@ -172,7 +172,7 @@ struct component *create_simple_bus()
 }
 
 stat simple_bus_add(struct component *bus, struct component *component,
-                    uintptr_t addr, size_t size)
+                    uint64_t addr, uint64_t size)
 {
 	struct mem_region *new = calloc(1, sizeof(struct mem_region));
 	if (!new)
diff --git a/src/components/cpu/riscv/simple_riscv32.c b/src/components/cpu/riscv/simple_riscv64.c
index 51b2a42..be96e60 100644
--- a/src/components/cpu/riscv/simple_riscv32.c
+++ b/src/components/cpu/riscv/simple_riscv64.c
@@ -2,30 +2,31 @@
 /* Copyright 2023 Kim Kuparinen < kimi.h.kuparinen@gmail.com > */
 
 #include <byteswap.h>
+#include <string.h>
 #include <assert.h>
 #include <stdbool.h>
 #include <stdint.h>
 
-#include <gran/cpu/riscv/simple_riscv32.h>
+#include <gran/cpu/riscv/simple_riscv64.h>
 
-struct simple_rv32_ldst {
+struct simple_rv64_ldst {
 	struct packet *pkt;
 	uint32_t reg;
 	bool u;
 };
 
-struct simple_riscv32 {
+struct simple_riscv64 {
 	struct component component;
 
 	struct component *imem;
 	struct component *dmem;
 
-	struct simple_rv32_ldst dls;
-	struct simple_rv32_ldst ils;
+	struct simple_rv64_ldst dls;
+	struct simple_rv64_ldst ils;
 
 	/* have to be careful with x0 */
-	uint32_t regs[32];
-	uint32_t pc;
+	uint64_t regs[32];
+	uint64_t pc;
 };
 
 /* big endian format, going from smallest to highest address.
@@ -119,7 +120,7 @@ union rv_insn {
 	uint32_t val;
 };
 
-static uint32_t get_reg(struct simple_riscv32 *cpu, size_t i)
+static uint64_t get_reg(struct simple_riscv64 *cpu, size_t i)
 {
 	assert(i < 32);
 
@@ -129,7 +130,7 @@ static uint32_t get_reg(struct simple_riscv32 *cpu, size_t i)
 	return cpu->regs[i];
 }
 
-static void set_reg(struct simple_riscv32 *cpu, size_t i, uint32_t v)
+static void set_reg(struct simple_riscv64 *cpu, size_t i, uint64_t v)
 {
 	assert(i < 32);
 
@@ -143,18 +144,18 @@ static void set_reg(struct simple_riscv32 *cpu, size_t i, uint32_t v)
 #define EXTEND_IMM20(x) ((int32_t)((x) << 12) >> 12)
 #define SHAMT(x) ((x) & 0b11111)
 
-static stat op_imm(struct simple_riscv32 *cpu, union rv_insn insn)
+static stat op_imm(struct simple_riscv64 *cpu, union rv_insn insn)
 {
-	uint32_t dst = 0;
-	uint32_t src = get_reg(cpu, insn.itype.rs1);
+	uint64_t dst = 0;
+	uint64_t src = get_reg(cpu, insn.itype.rs1);
 
-	uint32_t imm = EXTEND_IMM12(insn.itype.imm);
+	uint64_t imm = EXTEND_IMM12(insn.itype.imm);
 
 	switch (insn.rtype.funct3) {
 	/* ADDI  */
 	case 0b000: dst = src + imm; break;
 	/* SLTI  */
-	case 0b010: dst = (int32_t)src < (int32_t)imm; break;
+	case 0b010: dst = (int64_t)src < (int64_t)imm; break;
 	/* SLTIU */
 	case 0b011: dst = src < imm; break;
 	/* ANDI  */
@@ -171,7 +172,7 @@ static stat op_imm(struct simple_riscv32 *cpu, union rv_insn insn)
 		if (imm & ~0b11111)     /* SRLI */
 			dst = src >> SHAMT(imm);
 		else     /* SRAI */
-			dst = (int32_t)src >> SHAMT(imm);
+			dst = (int64_t)src >> SHAMT(imm);
 		break;
 
 	default:
@@ -184,26 +185,67 @@ static stat op_imm(struct simple_riscv32 *cpu, union rv_insn insn)
 	return OK;
 }
 
-static stat lui(struct simple_riscv32 *cpu, union rv_insn insn)
+static stat op_imm_32(struct simple_riscv64 *cpu, union rv_insn insn)
+{
+	uint64_t dst = 0;
+	uint64_t src = get_reg(cpu, insn.itype.rs1);
+
+	uint64_t imm = EXTEND_IMM12(insn.itype.imm);
+
+	switch (insn.rtype.funct3) {
+	/* ADDIW  */
+	case 0b000: dst = src + imm; break;
+	/* SLTI  */
+	case 0b010: dst = (int64_t)src < (int64_t)imm; break;
+	/* SLTIU */
+	case 0b011: dst = src < imm; break;
+	/* ANDI  */
+	case 0b111: dst = src & imm; break;
+	/* ORI   */
+	case 0b110: dst = src | imm; break;
+	/* XORI  */
+	case 0b100: dst = src ^ imm; break;
+	/* SLLIW  */
+	case 0b001: dst = src << SHAMT(imm); break;
+
+	/* SRLIW / SRAIW */
+	case 0b101:
+		if (imm & ~0b11111)     /* SRLI */
+			dst = src >> SHAMT(imm);
+		else     /* SRAI */
+			dst = (int64_t)src >> SHAMT(imm);
+		break;
+
+	default:
+		error("unknown OP-IMM instruction: %x", insn.itype.funct3);
+		return ENOSUCH;
+	}
+
+	set_reg(cpu, insn.itype.rd, dst);
+	cpu->pc += 4;
+	return OK;
+}
+
+static stat lui(struct simple_riscv64 *cpu, union rv_insn insn)
 {
 	set_reg(cpu, insn.utype.rd, insn.utype.imm << 12);
 	cpu->pc += 4;
 	return OK;
 }
 
-static stat auipc(struct simple_riscv32 *cpu, union rv_insn insn)
+static stat auipc(struct simple_riscv64 *cpu, union rv_insn insn)
 {
-	uint32_t res = cpu->pc + (insn.utype.imm << 12);
+	uint64_t res = cpu->pc + (insn.utype.imm << 12);
 	set_reg(cpu, insn.utype.rd, res);
 	cpu->pc += 4;
 	return OK;
 }
 
-static stat op(struct simple_riscv32 *cpu, union rv_insn insn)
+static stat op(struct simple_riscv64 *cpu, union rv_insn insn)
 {
-	uint32_t dst = 0;
-	uint32_t src1 = get_reg(cpu, insn.rtype.rs1);
-	uint32_t src2 = get_reg(cpu, insn.rtype.rs2);
+	uint64_t dst = 0;
+	uint64_t src1 = get_reg(cpu, insn.rtype.rs1);
+	uint64_t src2 = get_reg(cpu, insn.rtype.rs2);
 
 	switch (insn.rtype.funct3) {
 	/* ADD/SUB */
@@ -215,7 +257,7 @@ static stat op(struct simple_riscv32 *cpu, union rv_insn insn)
 		break;
 
 	/* SLT */
-	case 0b010: dst = (int32_t)src1 < (int32_t)src2; break;
+	case 0b010: dst = (int64_t)src1 < (int64_t)src2; break;
 	/* SLTU */
 	case 0b011: dst = src1 < src2; break;
 	/* AND */
@@ -231,7 +273,7 @@ static stat op(struct simple_riscv32 *cpu, union rv_insn insn)
 		if (insn.rtype.funct7)     /* SRL */
 			dst = src1 >> src2;
 		else     /* SRA */
-			dst = (int32_t)src1 >> src2;
+			dst = (int64_t)src1 >> src2;
 		break;
 
 	default:
@@ -250,18 +292,18 @@ static stat op(struct simple_riscv32 *cpu, union rv_insn insn)
 	             | (insn.jtype.imm1 << 11) \
 	             | (insn.jtype.imm0 << 12))
 
-static stat jal(struct simple_riscv32 *cpu, union rv_insn insn)
+static stat jal(struct simple_riscv64 *cpu, union rv_insn insn)
 {
 	/** @todo generate exception on unaligned jumps */
-	int32_t imm = JTYPE_IMM(insn);
+	int64_t imm = JTYPE_IMM(insn);
 	set_reg(cpu, insn.jtype.rd, cpu->pc + 4);
 	cpu->pc += imm;
 	return OK;
 }
 
-static stat jalr(struct simple_riscv32 *cpu, union rv_insn insn)
+static stat jalr(struct simple_riscv64 *cpu, union rv_insn insn)
 {
-	int32_t src = get_reg(cpu, insn.itype.rs1);
+	int64_t src = get_reg(cpu, insn.itype.rs1);
 	set_reg(cpu, insn.itype.rd, cpu->pc + 4);
 	cpu->pc += src + EXTEND_IMM12(insn.itype.imm);
 	return OK;
@@ -273,11 +315,11 @@ static stat jalr(struct simple_riscv32 *cpu, union rv_insn insn)
 	             | (insn.btype.imm1 << 1) \
 	             | (insn.btype.imm0 << 11))
 
-static stat branch(struct simple_riscv32 *cpu, union rv_insn insn)
+static stat branch(struct simple_riscv64 *cpu, union rv_insn insn)
 {
-	uint32_t src1 = get_reg(cpu, insn.btype.rs1);
-	uint32_t src2 = get_reg(cpu, insn.btype.rs2);
-	int32_t offset = BTYPE_IMM(insn);
+	uint64_t src1 = get_reg(cpu, insn.btype.rs1);
+	uint64_t src2 = get_reg(cpu, insn.btype.rs2);
+	int64_t offset = BTYPE_IMM(insn);
 
 	switch (insn.btype.funct3) {
 	/* BEQ */
@@ -298,7 +340,7 @@ static stat branch(struct simple_riscv32 *cpu, union rv_insn insn)
 
 	/* BLT */
 	case 0b100:
-		if ((int32_t)src1 < (int32_t)src2) {
+		if ((int64_t)src1 < (int64_t)src2) {
 			cpu->pc += offset;
 			return OK;
 		}
@@ -314,7 +356,7 @@ static stat branch(struct simple_riscv32 *cpu, union rv_insn insn)
 
 	/* BGE */
 	case 0b101:
-		if ((int32_t)src1 >= (int32_t)src2) {
+		if ((int64_t)src1 >= (int64_t)src2) {
 			cpu->pc += offset;
 			return OK;
 		}
@@ -337,13 +379,13 @@ static stat branch(struct simple_riscv32 *cpu, union rv_insn insn)
 	return OK;
 }
 
-static stat load(struct simple_riscv32 *cpu, union rv_insn insn)
+static stat load(struct simple_riscv64 *cpu, union rv_insn insn)
 {
-	int32_t imm = EXTEND_IMM12(insn.itype.imm);
-	int32_t base = get_reg(cpu, insn.itype.rs1);
+	int64_t imm = EXTEND_IMM12(insn.itype.imm);
+	int64_t base = get_reg(cpu, insn.itype.rs1);
 
-	size_t addr = base + imm;
-	size_t size = 0;
+	int64_t addr = base + imm;
+	int64_t size = 0;
 	bool u = false;
 
 	// assume little endian for now
@@ -356,6 +398,7 @@ static stat load(struct simple_riscv32 *cpu, union rv_insn insn)
 	case 0b001: size = 2; break;
 	/* LW */
 	case 0b010: size = 4; break;
+	case 0b011: size = 8; break;
 	default:
 		error("unknown LOAD width %x", insn.btype.funct3);
 		return ENOSUCH;
@@ -365,7 +408,7 @@ static stat load(struct simple_riscv32 *cpu, union rv_insn insn)
 	if (!pkt)
 		return EMEM;
 
-	cpu->dls = (struct simple_rv32_ldst){pkt, insn.itype.rd, u};
+	cpu->dls = (struct simple_rv64_ldst){pkt, insn.itype.rd, u};
 	stat ret = read(cpu->dmem, pkt);
 	if (ret)
 		return ret;
@@ -378,13 +421,13 @@ static stat load(struct simple_riscv32 *cpu, union rv_insn insn)
 	EXTEND_IMM12((insn.stype.imm1 << 5) \
 	             | (insn.stype.imm0))
 
-static stat store(struct simple_riscv32 *cpu, union rv_insn insn)
+static stat store(struct simple_riscv64 *cpu, union rv_insn insn)
 {
-	int32_t imm = STYPE_IMM(insn);
-	int32_t base = get_reg(cpu, insn.stype.rs1);
-	size_t addr = base + imm;
+	int64_t imm = STYPE_IMM(insn);
+	int64_t base = get_reg(cpu, insn.stype.rs1);
+	int64_t addr = base + imm;
 
-	uint32_t src = get_reg(cpu, insn.stype.rs2);
+	uint64_t src = get_reg(cpu, insn.stype.rs2);
 
 	switch (insn.stype.funct3) {
 	/* SB */
@@ -405,6 +448,12 @@ static stat store(struct simple_riscv32 *cpu, union rv_insn insn)
 		*(uint32_t *)packet_data(cpu->dls.pkt) = src;
 		break;
 	}
+	/* SD */
+	case 0b011: {
+		cpu->dls.pkt = create_packet(PACKET_WRITE, addr, 8);
+		*(uint64_t *)packet_data(cpu->dls.pkt) = src;
+		break;
+	}
 	default:
 		error("unknown width of STORE %x", insn.stype.funct3);
 		return ENOSUCH;
@@ -414,11 +463,11 @@ static stat store(struct simple_riscv32 *cpu, union rv_insn insn)
 	return write(cpu->dmem, cpu->dls.pkt);
 }
 
-static void finalize_ld(struct simple_riscv32 *cpu)
+static void finalize_ld(struct simple_riscv64 *cpu)
 {
-	struct simple_rv32_ldst ld = cpu->dls;
+	struct simple_rv64_ldst ld = cpu->dls;
 
-	uint32_t val = 0;
+	uint64_t val = 0;
 	void *data = packet_data(ld.pkt);
 	switch (packet_size(ld.pkt)) {
 	case 1:
@@ -430,7 +479,11 @@ static void finalize_ld(struct simple_riscv32 *cpu)
 		else val = *(int16_t *)data;
 		break;
 
-	case 4: val = *(uint32_t *)data;
+	case 4: if (ld.u) val = *(uint32_t *)data;
+		else val = *(int32_t *)data;
+		break;
+
+	case 8: val = *(uint64_t *)data;
 		break;
 
 	default:
@@ -440,13 +493,13 @@ static void finalize_ld(struct simple_riscv32 *cpu)
 	set_reg(cpu, ld.reg, val);
 }
 
-static void finalize_st(struct simple_riscv32 *cpu)
+static void finalize_st(struct simple_riscv64 *cpu)
 {
 	(void)cpu;
 	/* nothing really to do, this is here mostly for vibe */
 }
 
-static void finalize_dls(struct simple_riscv32 *cpu)
+static void finalize_dls(struct simple_riscv64 *cpu)
 {
 	struct packet *pkt = cpu->dls.pkt;
 
@@ -455,13 +508,13 @@ static void finalize_dls(struct simple_riscv32 *cpu)
 	else if (packet_type(pkt) == PACKET_WRITE)
 		finalize_st(cpu);
 	else
-		error("unsupported packet type for simple_riscv32");
+		error("unsupported packet type for simple_riscv64");
 
 	destroy_packet(pkt);
 	cpu->dls.pkt = NULL;
 }
 
-static uint32_t finalize_ils(struct simple_riscv32 *cpu)
+static uint32_t finalize_ils(struct simple_riscv64 *cpu)
 {
 	uint32_t insn = *(uint32_t *)packet_data(cpu->ils.pkt);
 	destroy_packet(cpu->ils.pkt);
@@ -469,7 +522,7 @@ static uint32_t finalize_ils(struct simple_riscv32 *cpu)
 	return insn;
 }
 
-static stat simple_riscv32_clock(struct simple_riscv32 *cpu)
+static stat simple_riscv64_clock(struct simple_riscv64 *cpu)
 {
 	/* there's an active data transfer we should handle */
 	if (cpu->dls.pkt) {
@@ -509,6 +562,7 @@ static stat simple_riscv32_clock(struct simple_riscv32 *cpu)
 	// all formats have identical opcodes, use whatever
 	switch (i.rtype.op) {
 	case OP_IMM: ret = op_imm(cpu, i); break;
+	case OP_IMM_32: ret = op_imm_32(cpu, i); break;
 	case LUI: ret = lui(cpu, i); break;
 	case AUIPC: ret = auipc(cpu, i); break;
 	case OP: ret = op(cpu, i); break;
@@ -530,7 +584,7 @@ static stat simple_riscv32_clock(struct simple_riscv32 *cpu)
 	return ret;
 }
 
-static void simple_riscv32_destroy(struct simple_riscv32 *cpu)
+static void simple_riscv64_destroy(struct simple_riscv64 *cpu)
 {
 	/* oh yeah, will have to think about the name stuff,
 	 * i.e. how and where to free it, and where to assign it */
@@ -548,19 +602,25 @@ static void simple_riscv32_destroy(struct simple_riscv32 *cpu)
 	free(cpu);
 }
 
-struct component *create_simple_riscv32(uint32_t start_pc,
+struct component *create_simple_riscv64(uint32_t start_pc,
                                         struct component *imem,
                                         struct component *dmem)
 {
-	struct simple_riscv32 *new = calloc(1, sizeof(struct simple_riscv32));
+	struct simple_riscv64 *new = calloc(1, sizeof(struct simple_riscv64));
 	if (!new)
 		return NULL;
 
-	new->component.clock = (clock_callback)simple_riscv32_clock;
-	new->component.destroy = (destroy_callback)simple_riscv32_destroy;
+	new->component.clock = (clock_callback)simple_riscv64_clock;
+	new->component.destroy = (destroy_callback)simple_riscv64_destroy;
 
 	new->pc = start_pc;
 	new->imem = imem;
 	new->dmem = dmem;
 	return (struct component *)new;
 }
+
+void simple_riscv64_set_reg(struct component *cpu, size_t reg, uint64_t val)
+{
+	struct simple_riscv64 *rv64 = (struct simple_riscv64 *)cpu;
+	set_reg(rv64, reg, val);
+}
diff --git a/src/components/cpu/riscv/source.mk b/src/components/cpu/riscv/source.mk
index ea41f62..da1cef6 100644
--- a/src/components/cpu/riscv/source.mk
+++ b/src/components/cpu/riscv/source.mk
@@ -1 +1 @@
-SOURCES += src/components/cpu/riscv/simple_riscv32.c
+SOURCES += src/components/cpu/riscv/simple_riscv64.c
diff --git a/src/components/grid/node.c b/src/components/grid/node.c
new file mode 100644
index 0000000..0aed768
--- /dev/null
+++ b/src/components/grid/node.c
@@ -0,0 +1,106 @@
+/* very simple grid node with 32bit private region, does not currently signal
+ * being busy or anything. I think I might have to refine the message passing
+ * interface I have, but this is good enough.
+ *
+ * Each node should have a router beneath it, just to simplify my life. A router
+ * is basically a bus with a fallback ascension path.
+ */
+#include <gran/grid/node.h>
+
+struct grid_node {
+	struct component component;
+	uint8_t u, v, x, y;
+	struct component *left, *right, *up, *down, *ascend, *lower;
+};
+
+typedef read_callback callback;
+
+static stat grid_route(struct grid_node *grid, struct packet *pkt, callback op)
+{
+	uint64_t addr = packet_addr(pkt);
+	uint8_t u = (addr >> 56) & 0xff;
+	uint8_t v = (addr >> 48) & 0xff;
+	uint8_t x = (addr >> 40) & 0xff;
+	uint8_t y = (addr >> 32) & 0xff;
+
+	if (grid->u == u && grid->v == v && grid->x == x && grid->y == y)
+		return op(grid->lower, pkt);
+
+	if (grid->u != u || grid->v != v) {
+		if (!grid->ascend)
+			return EBUS;
+
+		return op(grid->ascend, pkt);
+	}
+
+	if (y < grid->y) {
+		if (!grid->down)
+			return EBUS;
+
+		return op(grid->down, pkt);
+	}
+
+	if (y > grid->y) {
+		if (!grid->up)
+			return EBUS;
+
+		return op(grid->up, pkt);
+	}
+
+	if (x < grid->x) {
+		if (!grid->left)
+			return EBUS;
+
+		return op(grid->left, pkt);
+	}
+
+	if (x > grid->x) {
+		if (!grid->right)
+			return EBUS;
+
+		return op(grid->right, pkt);
+	}
+
+	return EBUS;
+}
+
+static stat grid_write(struct grid_node *node, struct packet *pkt)
+{
+	return grid_route(node, pkt, write);
+}
+
+static stat grid_read(struct grid_node *node, struct packet *pkt)
+{
+	return grid_route(node, pkt, read);
+}
+
+struct component *create_grid_node(uint8_t u, uint8_t v, uint8_t x, uint8_t y)
+{
+	struct grid_node *node = calloc(1, sizeof(struct grid_node));
+	if (!node)
+		return NULL;
+
+	node->u = u;
+	node->v = v;
+	node->x = x;
+	node->y = y;
+	node->component.write = (write_callback)grid_write;
+	node->component.read = (read_callback)grid_read;
+
+	return (struct component *)node;
+}
+
+stat grid_node_connect(struct component *node,
+		struct component *left, struct component *right,
+		struct component *up, struct component *down,
+		struct component *lower, struct component *ascend)
+{
+	struct grid_node *n = (struct grid_node *)node;
+	n->left = left;
+	n->right = right;
+	n->up = up;
+	n->down = down;
+	n->lower = lower;
+	n->ascend = ascend;
+	return OK;
+}
diff --git a/src/components/grid/router.c b/src/components/grid/router.c
new file mode 100644
index 0000000..bdb05c8
--- /dev/null
+++ b/src/components/grid/router.c
@@ -0,0 +1,114 @@
+#include <gran/grid/router.h>
+#include <gran/vec.h>
+
+struct router_region {
+	uint32_t addr;
+	uint32_t size;
+	struct component *component;
+};
+
+struct node_router {
+	struct component component;
+	struct component *ascend;
+	struct vec regions;
+	uint8_t u, v, x, y;
+};
+
+static struct router_region *find_region(struct node_router *router, uint32_t addr)
+{
+	for (size_t i = 0; i < vec_len(&router->regions); ++i) {
+		struct router_region *region = vec_at(&router->regions, i);
+
+		if (addr >= region->addr && addr < region->addr + region->size)
+			return region;
+	}
+
+	return NULL;
+}
+
+static stat router_write(struct node_router *router, struct packet *pkt)
+{
+	uint64_t addr = packet_addr(pkt);
+	uint8_t u = (addr >> 56) & 0xff;
+	uint8_t v = (addr >> 48) & 0xff;
+	uint8_t x = (addr >> 40) & 0xff;
+	uint8_t y = (addr >> 32) & 0xff;
+
+	if (router->u != u || router->v != v || router->x != x || router->y != y)
+		return write(router->ascend, pkt);
+
+	struct router_region *region = find_region(router, addr);
+	if (!region)
+		return write(router->ascend, pkt);
+
+	return write(region->component, pkt);
+}
+
+static stat router_read(struct node_router *router, struct packet *pkt)
+{
+	uint64_t addr = packet_addr(pkt);
+	uint8_t u = (addr >> 56) & 0xff;
+	uint8_t v = (addr >> 48) & 0xff;
+	uint8_t x = (addr >> 40) & 0xff;
+	uint8_t y = (addr >> 32) & 0xff;
+
+	if (router->u != u || router->v != v || router->x != x || router->y != y)
+		return read(router->ascend, pkt);
+
+	struct router_region *region = find_region(router, addr);
+	if (!region)
+		return read(router->ascend, pkt);
+
+	return read(region->component, pkt);
+}
+
+struct component *create_node_router(uint8_t u, uint8_t v, uint8_t x, uint8_t y)
+{
+	struct node_router *router = calloc(1, sizeof(struct node_router));
+	if (!router)
+		return NULL;
+
+	router->u = u;
+	router->v = v;
+	router->x = x;
+	router->y = y;
+
+	router->component.write = (write_callback)router_write;
+	router->component.read = (read_callback)router_read;
+	router->regions = vec_create(sizeof(struct router_region));
+
+	return (struct component *)router;
+}
+
+stat node_router_add(struct component *router, struct component *component, uint32_t addr, uint32_t size)
+{
+	struct node_router *nr = (struct node_router *)router;
+	struct router_region *found = find_region(nr, addr);
+	if (!found)           found = find_region(nr, addr + size);
+
+	if (found) {
+		error("%s overlaps with %s at %x",
+				found->component->name,
+				component->name,
+				found->addr
+		     );
+
+		return EEXISTS;
+	}
+
+	struct router_region r = (struct router_region){
+		.addr = addr,
+		.size = size,
+		.component = component
+	};
+	vect_append(struct node_region, nr->regions, &r);
+
+	return OK;
+}
+
+stat node_router_ascend(struct component *router, struct component *node)
+{
+	struct node_router *nr = (struct node_router *)router;
+	nr->ascend = node;
+	return OK;
+}
diff --git a/src/components/grid/source.mk b/src/components/grid/source.mk
new file mode 100644
index 0000000..4fb44ec
--- /dev/null
+++ b/src/components/grid/source.mk
@@ -0,0 +1 @@
+SOURCES += src/components/grid/node.c src/components/grid/router.c
diff --git a/src/components/uart/simple_uart.c b/src/components/uart/simple_uart.c
new file mode 100644
index 0000000..5c97ccd
--- /dev/null
+++ b/src/components/uart/simple_uart.c
@@ -0,0 +1,28 @@
+#include <stdio.h>
+#include <gran/uart/simple_uart.h>
+
+struct simple_uart {
+	struct component component;
+};
+
+static stat simple_uart_write(struct simple_uart *uart, struct packet *pkt)
+{
+	(void)uart;
+	size_t size = packet_size(pkt);
+	if (size != 1)
+		return EBUS;
+
+	putchar(*(uint8_t *)packet_data(pkt));
+	packet_set_state(pkt, PACKET_DONE);
+	return OK;
+}
+
+struct component *create_simple_uart()
+{
+	struct simple_uart *uart = calloc(1, sizeof(struct simple_uart));
+	if (!uart)
+		return NULL;
+
+	uart->component.write = (write_callback)simple_uart_write;
+	return (struct component *)uart;
+}
diff --git a/src/components/uart/source.mk b/src/components/uart/source.mk
new file mode 100644
index 0000000..af26289
--- /dev/null
+++ b/src/components/uart/source.mk
@@ -0,0 +1 @@
+SOURCES += src/components/uart/simple_uart.c
diff --git a/src/vec.c b/src/vec.c
new file mode 100644
index 0000000..e56ac41
--- /dev/null
+++ b/src/vec.c
@@ -0,0 +1,67 @@
+/* SPDX-License-Identifier: copyleft-next-0.3.1 */
+
+#include <stdlib.h>
+#include <assert.h>
+#include <string.h>
+
+#include <gran/vec.h>
+
+struct vec vec_create(size_t ns)
+{
+	return (struct vec) {
+		       .n = 0,
+		       .s = 1,
+		       .ns = ns,
+		       .buf = malloc(ns),
+	};
+}
+
+size_t vec_len(struct vec *v)
+{
+	return v->n;
+}
+
+void *vec_at(struct vec *v, size_t i)
+{
+	assert(i < v->n && "out of vector bounds");
+	return v->buf + i * v->ns;
+}
+
+void *vec_back(struct vec *v)
+{
+	assert(v->n);
+	return v->buf + (v->n - 1) * v->ns;
+}
+
+void *vec_pop(struct vec *v)
+{
+	assert(v->n && "attempting to pop empty vector");
+	v->n--;
+	return v->buf + v->n * v->ns;
+}
+
+void vec_append(struct vec *v, void *n)
+{
+	v->n++;
+	if (v->n >= v->s) {
+		v->s *= 2;
+		v->buf = realloc(v->buf, v->s * v->ns);
+	}
+
+	void *p = vec_at(v, v->n - 1);
+	memcpy(p, n, v->ns);
+}
+
+void vec_reset(struct vec *v)
+{
+	v->n = 0;
+}
+
+void vec_destroy(struct vec *v) {
+	free(v->buf);
+}
+
+void vec_sort(struct vec *v, vec_comp_t comp)
+{
+	qsort(v->buf, v->n, v->ns, comp);
+}
diff --git a/tests/simple_grid/sim.c b/tests/simple_grid/sim.c
new file mode 100644
index 0000000..e0c485b
--- /dev/null
+++ b/tests/simple_grid/sim.c
@@ -0,0 +1,114 @@
+/* testcase for a 64x64 grid of processors, that all just spam the first memory
+ * region due to there being a 'lock' variable there, more or less the worst
+ * possible program for performance. */
+#include <assert.h>
+
+#include <gran/root.h>
+#include <gran/mem/simple_mem.h>
+#include <gran/bus/simple_bus.h>
+#include <gran/uart/simple_uart.h>
+#include <gran/grid/node.h>
+#include <gran/grid/router.h>
+#include <gran/cpu/riscv/simple_riscv64.h>
+
+unsigned char test_grid[] = {
+  0xb3, 0x67, 0xb5, 0x00, 0x63, 0x9a, 0x07, 0x0a, 0xb7, 0x17, 0x00, 0x00,
+  0x23, 0xb0, 0x07, 0x00, 0x1b, 0x57, 0x35, 0x00, 0xb7, 0x27, 0x00, 0x00,
+  0x93, 0x06, 0x80, 0x02, 0x13, 0x77, 0x77, 0x00, 0x23, 0x80, 0xd7, 0x00,
+  0x13, 0x07, 0x07, 0x03, 0x93, 0x76, 0x75, 0x00, 0x23, 0x80, 0xe7, 0x00,
+  0x13, 0x87, 0x06, 0x03, 0x23, 0x80, 0xe7, 0x00, 0x93, 0x06, 0xc0, 0x02,
+  0x1b, 0xd7, 0x35, 0x00, 0x23, 0x80, 0xd7, 0x00, 0x13, 0x77, 0x77, 0x00,
+  0x93, 0x06, 0x00, 0x02, 0x23, 0x80, 0xd7, 0x00, 0x13, 0x07, 0x07, 0x03,
+  0x93, 0xf6, 0x75, 0x00, 0x23, 0x80, 0xe7, 0x00, 0x13, 0x87, 0x06, 0x03,
+  0x23, 0x80, 0xe7, 0x00, 0x13, 0x07, 0x90, 0x02, 0x23, 0x80, 0xe7, 0x00,
+  0x93, 0x06, 0xa0, 0x00, 0x23, 0x80, 0xd7, 0x00, 0x1b, 0x87, 0x15, 0x00,
+  0x93, 0x06, 0xf0, 0x03, 0x93, 0x77, 0xf7, 0x0f, 0x63, 0x98, 0xd5, 0x00,
+  0x1b, 0x05, 0x15, 0x00, 0x13, 0x75, 0xf5, 0x0f, 0x93, 0x07, 0x00, 0x00,
+  0x13, 0x15, 0x85, 0x00, 0x33, 0x65, 0xf5, 0x00, 0xb7, 0x17, 0x00, 0x00,
+  0x23, 0xb0, 0xa7, 0x00, 0xb7, 0x16, 0x00, 0x00, 0x37, 0x47, 0x00, 0x00,
+  0x83, 0xb7, 0x06, 0x00, 0xe3, 0x9e, 0xe7, 0xfe, 0x73, 0x00, 0x10, 0x00,
+  0x67, 0x80, 0x00, 0x00, 0x13, 0x17, 0x85, 0x00, 0x33, 0x67, 0xb7, 0x00,
+  0xb7, 0x16, 0x00, 0x00, 0x83, 0xb7, 0x06, 0x00, 0xe3, 0x9e, 0xe7, 0xfe,
+  0x6f, 0xf0, 0x5f, 0xf4
+};
+unsigned int test_grid_len = 208;
+
+static struct component *get_grid(struct component **grid, int i, int j, uint8_t x, uint8_t y)
+{
+	if (i < 0)
+		return NULL;
+
+	if (j < 0)
+		return NULL;
+
+	if (i >= x)
+		return NULL;
+
+	if (j >= y)
+		return NULL;
+
+	return grid[i * x + j];
+}
+
+static stat build_grid(struct clock_domain *clk, uint8_t x, uint8_t y)
+{
+	struct component **grid = calloc(x * y, sizeof(struct component *));
+	assert(grid);
+
+	struct component **routers = calloc(x * y, sizeof(struct component *));
+	assert(routers);
+
+	for (size_t i = 0; i < x; ++i)
+	for (size_t j = 0; j < y; ++j) {
+		struct component *imem = create_simple_mem(4096);
+		init_simple_mem(imem, 0, test_grid_len, test_grid);
+
+		struct component *dmem = create_simple_mem(4096);
+		struct component *router = create_node_router(0, 0, i, j);
+		node_router_add(router, dmem, 4096, 4096);
+
+		struct component *rv64 = create_simple_riscv64(0, imem, router);
+		simple_riscv64_set_reg(rv64, 10, i); /* a0 */
+		simple_riscv64_set_reg(rv64, 11, j); /* a1 */
+
+		clock_domain_add(clk, rv64);
+
+		routers[i * x + j] = router;
+		grid[i * x + j] = create_grid_node(0, 0, i, j);
+	}
+
+	struct component *uart = create_simple_uart();
+	node_router_add(routers[0], uart, 8192, 1);
+
+	for (int i = 0; i < x; ++i)
+	for (int j = 0; j < y; ++j) {
+		struct component *node  = grid[i * x + j];
+		struct component *lower = routers[i * x + j];
+		struct component *left  = get_grid(grid, i - 1, j    , x, y);
+		struct component *right = get_grid(grid, i + 1, j    , x, y);
+		struct component *up    = get_grid(grid, i    , j + 1, x, y);
+		struct component *down  = get_grid(grid, i    , j - 1, x, y);
+		grid_node_connect(node, left, right, up, down, lower, NULL);
+		node_router_ascend(lower, node);
+	}
+
+	free(routers);
+	free(grid);
+	return OK;
+}
+
+int main()
+{
+	struct clock_domain *clk = create_clock_domain(NS(1));
+
+	stat r = build_grid(clk, 64, 64);
+	assert(r == OK);
+
+	struct gran_root *root = create_root();
+	root_add_clock(root, clk);
+
+	r = root_run(root);
+	assert(r == OK);
+
+	destroy_root(root);
+}
diff --git a/tests/simple_grid/source.mk b/tests/simple_grid/source.mk
new file mode 100644
index 0000000..e2dd442
--- /dev/null
+++ b/tests/simple_grid/source.mk
@@ -0,0 +1,6 @@
+GRID_TEST_OBJ	!= ./scripts/gen-deps --sources "tests/simple_grid/sim.c"
+
+TEST_PROGS	+= build/tests/simple_grid/sim
+
+build/tests/simple_grid/sim: $(GRID_TEST_OBJ) $(OBJS)
+	$(COMPILE) $(GRID_TEST_OBJ) $(OBJS) -o $@
diff --git a/tests/simple_grid/test.c b/tests/simple_grid/test.c
new file mode 100644
index 0000000..782062d
--- /dev/null
+++ b/tests/simple_grid/test.c
@@ -0,0 +1,43 @@
+#define X 64
+#define Y 64
+
+void _start(unsigned char x, unsigned char y)
+{
+	volatile unsigned long *counter = (unsigned long *)4096;
+	volatile char *uart = (char *)8192;
+
+	/* very hacky, not recommended but good enough for testing */
+	if (x == 0 && y == 0)
+		*counter = 0;
+	else
+		while (*counter != ((x << 8) | y))
+			;
+
+	*uart = '(';
+	/* [0 - 63] as two octal numbers */
+	*uart = ((x >> 3) & 0x7) + '0';
+	*uart = ((x >> 0) & 0x7) + '0';
+
+	*uart = ',';
+	*uart = ' ';
+
+	*uart = ((y >> 3) & 0x7) + '0';
+	*uart = ((y >> 0) & 0x7) + '0';
+
+	*uart = ')';
+	*uart = '\n';
+
+	if (y == 63) {
+		x++;
+		y = 0;
+	}
+	else
+		y++;
+
+	*counter = (x << 8) | y;
+
+	while (*counter != (64 << 8))
+		;
+
+	asm ("ebreak\n");
+}
diff --git a/tests/simple_riscv32/source.mk b/tests/simple_riscv32/source.mk
deleted file mode 100644
index 736c111..0000000
--- a/tests/simple_riscv32/source.mk
+++ /dev/null
@@ -1,6 +0,0 @@
-RV32_TEST_OBJ	!= ./scripts/gen-deps --sources "tests/simple_riscv32/sim.c"
-
-TEST_PROGS	+= build/tests/simple_riscv32/sim
-
-build/tests/simple_riscv32/sim: $(RV32_TEST_OBJ) $(TEST_OBJS) $(OBJS)
-	$(COMPILE) $(RV32_TEST_OBJ) $(TEST_OBJS) $(OBJS) -o $@
diff --git a/tests/simple_riscv32/s.asm b/tests/simple_riscv64/s.asm
index d423dd9..d423dd9 100644
--- a/tests/simple_riscv32/s.asm
+++ b/tests/simple_riscv64/s.asm
diff --git a/tests/simple_riscv32/sim.c b/tests/simple_riscv64/sim.c
index fcf19a1..ae3cad4 100644
--- a/tests/simple_riscv32/sim.c
+++ b/tests/simple_riscv64/sim.c
@@ -5,7 +5,7 @@
 
 #include <gran/root.h>
 #include <gran/mem/simple_mem.h>
-#include <gran/cpu/riscv/simple_riscv32.h>
+#include <gran/cpu/riscv/simple_riscv64.h>
 
 unsigned char simple_sum[] = {
 	0x13, 0x05, 0x00, 0x00, 0xb7, 0x45, 0x0f, 0x00, 0x93, 0x85, 0x05, 0x24,
@@ -24,10 +24,10 @@ int main()
 	init_simple_mem(imem, 0, simple_sum_len, simple_sum);
 
 	struct component *dmem = create_simple_mem(size);
-	struct component *rv32 = create_simple_riscv32(0, imem, dmem);
+	struct component *rv64 = create_simple_riscv64(0, imem, dmem);
 
 	struct clock_domain *clk = create_clock_domain(NS(1));
-	clock_domain_add(clk, rv32);
+	clock_domain_add(clk, rv64);
 
 	struct gran_root *root = create_root();
 	root_add_clock(root, clk);
diff --git a/tests/simple_riscv64/source.mk b/tests/simple_riscv64/source.mk
new file mode 100644
index 0000000..fbc541d
--- /dev/null
+++ b/tests/simple_riscv64/source.mk
@@ -0,0 +1,6 @@
+RV64_TEST_OBJ	!= ./scripts/gen-deps --sources "tests/simple_riscv64/sim.c"
+
+TEST_PROGS	+= build/tests/simple_riscv64/sim
+
+build/tests/simple_riscv64/sim: $(RV64_TEST_OBJ) $(TEST_OBJS) $(OBJS)
+	$(COMPILE) $(RV64_TEST_OBJ) $(TEST_OBJS) $(OBJS) -o $@
diff --git a/tests/simple_uart/hello.c b/tests/simple_uart/hello.c
new file mode 100644
index 0000000..30719e0
--- /dev/null
+++ b/tests/simple_uart/hello.c
@@ -0,0 +1,18 @@
+/* compile with
+ *	riscv64-unknown-elf-gcc -ffreestanding -nostdlib -Wl,-Ttext=0 -march=rv64i -mabi=lp64 -O2 hello.c -o simple_hello
+ *	xxd --include simple_hello
+ */
+void _start()
+{
+	/* relies on hard-coded system, good enough for this simple test */
+	volatile char *uart = (char *)4096;
+	*uart = 'H';
+	*uart = 'e';
+	*uart = 'l';
+	*uart = 'l';
+	*uart = 'o';
+	*uart = '!';
+	*uart = '\n';
+
+	asm ("ebreak\n");
+}
diff --git a/tests/simple_uart/sim.c b/tests/simple_uart/sim.c
new file mode 100644
index 0000000..762b591
--- /dev/null
+++ b/tests/simple_uart/sim.c
@@ -0,0 +1,45 @@
+/* SPDX-License-Identifier: copyleft-next-0.3.1 */
+/* Copyright 2023 Kim Kuparinen < kimi.h.kuparinen@gmail.com > */
+
+#include <assert.h>
+
+#include <gran/root.h>
+#include <gran/mem/simple_mem.h>
+#include <gran/bus/simple_bus.h>
+#include <gran/uart/simple_uart.h>
+#include <gran/cpu/riscv/simple_riscv64.h>
+
+unsigned char simple_hello[] = {
+  0xb7, 0x17, 0x00, 0x00, 0x13, 0x07, 0x80, 0x04, 0x23, 0x80, 0xe7, 0x00,
+  0x13, 0x07, 0x50, 0x06, 0x23, 0x80, 0xe7, 0x00, 0x13, 0x07, 0xc0, 0x06,
+  0x23, 0x80, 0xe7, 0x00, 0x23, 0x80, 0xe7, 0x00, 0x13, 0x07, 0xf0, 0x06,
+  0x23, 0x80, 0xe7, 0x00, 0x13, 0x07, 0x10, 0x02, 0x23, 0x80, 0xe7, 0x00,
+  0x13, 0x07, 0xa0, 0x00, 0x23, 0x80, 0xe7, 0x00, 0x73, 0x00, 0x10, 0x00,
+  0x67, 0x80, 0x00, 0x00
+};
+unsigned int simple_hello_len = 64;
+
+int main()
+{
+	const size_t size = 4096;
+	struct component *imem = create_simple_mem(size);
+	init_simple_mem(imem, 0, simple_hello_len, simple_hello);
+
+	struct component *dmem = create_simple_mem(size);
+	struct component *uart = create_simple_uart();
+	struct component *bus = create_simple_bus();
+	simple_bus_add(bus, dmem, 0, size);
+	simple_bus_add(bus, uart, size, 1);
+
+	struct component *rv64 = create_simple_riscv64(0, imem, bus);
+
+	struct clock_domain *clk = create_clock_domain(NS(1));
+	clock_domain_add(clk, rv64);
+
+	struct gran_root *root = create_root();
+	root_add_clock(root, clk);
+
+	assert(root_run(root) == OK);
+
+	destroy_root(root);
+}
diff --git a/tests/simple_uart/source.mk b/tests/simple_uart/source.mk
new file mode 100644
index 0000000..a42a447
--- /dev/null
+++ b/tests/simple_uart/source.mk
@@ -0,0 +1,6 @@
+UART_TEST_OBJ	!= ./scripts/gen-deps --sources "tests/simple_uart/sim.c"
+
+TEST_PROGS	+= build/tests/simple_uart/sim
+
+build/tests/simple_uart/sim: $(UART_TEST_OBJ) $(OBJS)
+	$(COMPILE) $(UART_TEST_OBJ) $(TEST_OBJS) $(OBJS) -o $@