The RISC-V Assembly Programmer's Manual is
© 2017 Palmer Dabbelt palmer@dabbelt.com © 2017 Michael Clark michaeljclark@mac.com © 2017 Alex Bradbury asb@lowrisc.org
It is licensed under the Creative Commons Attribution 4.0 International License (CC-BY 4.0). The full license text is available at https://creativecommons.org/licenses/by/4.0/.
I think it's probably better to beef up the binutils documentation rather than duplicating it here.
Registers are the most important part of any processor. RISC-V defines various types, depending on which extensions are included: The general registers (with the program counter), control registers, floating point registers (F extension), and vector registers (V extension).
The RV32I base integer ISA includes 32 registers, named x0
to x31
. The
program counter PC
is separate from these registers, in contrast to other
processors such as the ARM-32. The first register, x0
, has a special function:
Reading it always returns 0 and writes to it are ignored. As we will see later,
this allows various tricks and simplifications.
In practice, the programmer doesn't use this notation for the registers. Though
x1
to x31
are all equally general-use registers as far as the processor is
concerned, by convention certain registers are used for special tasks. In
assembler, they are given standardized names as part of the RISC-V application
binary interface (ABI). This is what you will usually see in code listings. If
you really want to see the numeric register names, the -M
argument to objdump
will provide them.
Register | ABI | Use by convention | Preserved? |
---|---|---|---|
x0 | zero | hardwired to 0, ignores writes | n/a |
x1 | ra | return address for jumps | no |
x2 | sp | stack pointer | yes |
x3 | gp | global pointer | n/a |
x4 | tp | thread pointer | n/a |
x5 | t0 | temporary register 0 | no |
x6 | t1 | temporary register 1 | no |
x7 | t2 | temporary register 2 | no |
x8 | s0 or fp | saved register 0 or frame pointer | yes |
x9 | s1 | saved register 1 | yes |
x10 | a0 | return value or function argument 0 | no |
x11 | a1 | return value or function argument 1 | no |
x12 | a2 | function argument 2 | no |
x13 | a3 | function argument 3 | no |
x14 | a4 | function argument 4 | no |
x15 | a5 | function argument 5 | no |
x16 | a6 | function argument 6 | no |
x17 | a7 | function argument 7 | no |
x18 | s2 | saved register 2 | yes |
x19 | s3 | saved register 3 | yes |
x20 | s4 | saved register 4 | yes |
x21 | s5 | saved register 5 | yes |
x22 | s6 | saved register 6 | yes |
x23 | s7 | saved register 7 | yes |
x24 | s8 | saved register 8 | yes |
x25 | s9 | saved register 9 | yes |
x26 | s10 | saved register 10 | yes |
x27 | s11 | saved register 11 | yes |
x28 | t3 | temporary register 3 | no |
x29 | t4 | temporary register 4 | no |
x30 | t5 | temporary register 5 | no |
x31 | t6 | temporary register 6 | no |
pc | (none) | program counter | n/a |
Registers of the RV32I. Based on RISC-V documentation and Patterson and Waterman "The RISC-V Reader" (2017)
As a general rule, the saved registers s0
to s11
are preserved across
function calls, while the argument registers a0
to a7
and the
temporary registers t0
to t6
are not. The use of the various
specialized registers such as sp
by convention will be discussed later in more
detail.
(TBA)
(TBA)
(TBA)
Addressing formats like %pcrel_lo(). We can just link to the RISC-V PS ABI document to describe what the relocations actually do.
Official Specifications webpage:
Latest Specifications draft repository:
https://riscv.org/specifications/
ALIAS line from opcodes/riscv-opc.c
To better diagnose situations where the program flow reaches an unexpected
location, you might want to emit there an instruction that's known to trap. You
can use an UNIMP
pseudo-instruction, which should trap in nearly all systems.
The de facto standard implementation of this instruction is:
-
C.UNIMP
:0000
. The all-zeroes pattern is not a valid instruction. Any system which traps on invalid instructions will thus trap on thisUNIMP
instruction form. Despite not being a valid instruction, it still fits the 16-bit (compressed) instruction format, and so0000 0000
is interpreted as being two 16-bitUNIMP
instructions. -
UNIMP
:C0001073
. This is an alias forCSRRW x0, cycle, x0
. Sincecycle
is a read-only CSR, then (whether this CSR exists or not) an attempt to write into it will generate an illegal instruction exception. This 32-bit form ofUNIMP
is emitted when targeting a system without the C extension, or when the.option norvc
directive is used.
Both the RISC-V-specific and GNU .-prefixed options.
The following table lists assembler directives:
Directive | Arguments | Description |
---|---|---|
.align | integer | align to power of 2 (alias for .p2align) |
.file | "filename" | emit filename FILE LOCAL symbol table |
.globl | symbol_name | emit symbol_name to symbol table (scope GLOBAL) |
.local | symbol_name | emit symbol_name to symbol table (scope LOCAL) |
.comm | symbol_name,size,align | emit common object to .bss section |
.common | symbol_name,size,align | emit common object to .bss section |
.ident | "string" | accepted for source compatibility |
.section | [{.text,.data,.rodata,.bss}] | emit section (if not present, default .text) and make current |
.size | symbol, symbol | accepted for source compatibility |
.text | emit .text section (if not present) and make current | |
.data | emit .data section (if not present) and make current | |
.rodata | emit .rodata section (if not present) and make current | |
.bss | emit .bss section (if not present) and make current | |
.string | "string" | emit string |
.asciz | "string" | emit string (alias for .string) |
.equ | name, value | constant definition |
.macro | name arg1 [, argn] | begin macro definition \argname to substitute |
.endm | end macro definition | |
.type | symbol, @function | accepted for source compatibility |
.option | {rvc,norvc,pic,nopic,relax,norelax,push,pop} | RISC-V options. Refer to .option for a more detailed description. |
.byte | expression [, expression]* | 8-bit comma separated words |
.2byte | expression [, expression]* | 16-bit comma separated words |
.half | expression [, expression]* | 16-bit comma separated words |
.short | expression [, expression]* | 16-bit comma separated words |
.4byte | expression [, expression]* | 32-bit comma separated words |
.word | expression [, expression]* | 32-bit comma separated words |
.long | expression [, expression]* | 32-bit comma separated words |
.8byte | expression [, expression]* | 64-bit comma separated words |
.dword | expression [, expression]* | 64-bit comma separated words |
.quad | expression [, expression]* | 64-bit comma separated words |
.dtprelword | expression [, expression]* | 32-bit thread local word |
.dtpreldword | expression [, expression]* | 64-bit thread local word |
.sleb128 | expression | signed little endian base 128, DWARF |
.uleb128 | expression | unsigned little endian base 128, DWARF |
.p2align | p2,[pad_val=0],max | align to power of 2 |
.balign | b,[pad_val=0] | byte align |
.zero | integer | zero bytes |
.variant_cc | symbol_name | annotate the symbol with variant calling convention |
Enable/disable the C-extension for the following code region.
Set the code model to PIC (position independent code) or non-PIC. This will
affect the expansion of the la
pseudoinstruction, refer to
listing of standard RISC-V pseudoinstructions.
Enable/disable linker relaxation for the following code region.
NOTE: Code region follows by .option relax
will emit
R_RISCV_RELAX
/R_RISCV_ALIGN
even linker unsupport relaxation, suggested
usage is using .option norelax
with .option push
/.option pop
if
you want to disable linker relaxation on specific code region.
NOTE: Recommended way to disable linker relaxation of specific code region is
use .option push
, .option norelax
and .option pop
, that prevent enabled
linker relaxation accidentally if user already disable linker relaxation.
Push/pop current options to/from the options stack.
The following table lists assembler relocation expansions:
Assembler Notation | Description | Instruction / Macro |
---|---|---|
%hi(symbol) | Absolute (HI20) | lui |
%lo(symbol) | Absolute (LO12) | load, store, add |
%pcrel_hi(symbol) | PC-relative (HI20) | auipc |
%pcrel_lo(label) | PC-relative (LO12) | load, store, add |
%tprel_hi(symbol) | TLS LE "Local Exec" | lui |
%tprel_lo(symbol) | TLS LE "Local Exec" | load, store, add |
%tprel_add(symbol) | TLS LE "Local Exec" | add |
%tls_ie_pcrel_hi(symbol) * | TLS IE "Initial Exec" (HI20) | auipc |
%tls_gd_pcrel_hi(symbol) * | TLS GD "Global Dynamic" (HI20) | auipc |
%got_pcrel_hi(symbol) * | GOT PC-relative (HI20) | auipc |
* These reuse %pcrel_lo(label) for their lower half
Text labels are used as branch, unconditional jump targets and symbol offsets. Text labels are added to the symbol table of the compiled module.
loop:
j loop
Numeric labels are used for local references. References to local labels are suffixed with 'f' for a forward reference or 'b' for a backwards reference.
1:
j 1b
The following example shows how to load an absolute address:
lui a0, %hi(msg + 1)
addi a0, a0, %lo(msg + 1)
Which generates the following assembler output and relocations
as seen by objdump
:
0000000000000000 <.text>:
0: 00000537 lui a0,0x0
0: R_RISCV_HI20 msg+0x1
4: 00150513 addi a0,a0,1 # 0x1
4: R_RISCV_LO12_I msg+0x1
The following example shows how to load a PC-relative address:
1:
auipc a0, %pcrel_hi(msg + 1)
addi a0, a0, %pcrel_lo(1b)
Which generates the following assembler output and relocations
as seen by objdump
:
0000000000000000 <.text>:
0: 00000517 auipc a0,0x0
0: R_RISCV_PCREL_HI20 msg+0x1
4: 00050513 mv a0,a0
4: R_RISCV_PCREL_LO12_I .L1
The following example shows how to load an address from the GOT:
1:
auipc a0, %got_pcrel_hi(msg + 1)
ld a0, %pcrel_lo(1b)(a0)
Which generates the following assembler output and relocations
as seen by objdump
:
0000000000000000 <.text>:
0: 00000517 auipc a0,0x0
0: R_RISCV_GOT_HI20 msg+0x1
4: 00050513 mv a0,a0
4: R_RISCV_PCREL_LO12_I .L1
The following example shows the li
pseudo instruction which
is used to load immediate values:
.equ CONSTANT, 0xdeadbeef
li a0, CONSTANT
Which, for RV32I, generates the following assembler output, as seen by objdump
:
00000000 <.text>:
0: deadc537 lui a0,0xdeadc
4: eef50513 addi a0,a0,-273 # deadbeef <CONSTANT+0x0>
The immediate argument to lui
is an integer in the interval [0x0, 0xfffff].
Its compressed form, c.lui
, accepts only those in the subintervals [0x1, 0x1f] and [0xfffe0, 0xfffff].
The following example shows the la
pseudo instruction which is used to load
symbol addresses using the correct sequence based on whether the code is being
assembled as PIC:
la a0, msg + 1
For non-PIC this is an alias for the lla
pseudo instruction documented below.
For PIC this is an alias for the lga
pseudo instruction documented below.
The la
pseudo instruction is the preferred way for getting the address of
variables in assembly unless explicit control over PC-relative or GOT-indirect
addressing is required.
The following example shows the lla
pseudo instruction which is used to load
local symbol addresses:
lla a0, msg + 1
This generates the following instructions and relocations as seen by objdump
:
0000000000000000 <.text>:
0: 00000517 auipc a0,0x0
0: R_RISCV_PCREL_HI20 msg+0x1
4: 00050513 mv a0,a0
4: R_RISCV_PCREL_LO12_I .L0
The following example shows the lga
pseudo instruction which is used to load
global symbol addresses:
lga a0, msg + 1
This generates the following instructions and relocations as seen by objdump
(for RV64; RV32 will use lw
instead of ld
):
0000000000000000 <.text>:
0: 00000517 auipc a0,0x0
0: R_RISCV_GOT_HI20 msg+0x1
4: 00053503 ld a0,0(a0) # 0 <.text>
4: R_RISCV_PCREL_LO12_I .L0
The following pseudo instructions are available to load from and store to global objects:
l{b|h|w|d} <rd>, <symbol>
: load byte, half word, word or double word from global1s{b|h|w|d} <rd>, <symbol>, <rt>
: store byte, half word, word or double word to global2fl{h|w|d|q} <rd>, <symbol>, <rt>
: load half, float, double or quad precision from global2fs{h|w|d|q} <rd>, <symbol>, <rt>
: store half, float, double or quad precision to global2
The following example shows how these pseudo instructions are used:
lw a0, var1
fld fa0, var2, t0
sw a0, var3, t0
fsd fa0, var4, t0
Which generates the following assembler output and relocations
as seen by objdump
:
0000000000000000 <.text>:
0: 00000517 auipc a0,0x0
0: R_RISCV_PCREL_HI20 var1
4: 00052503 lw a0,0(a0) # 0 <.text>
4: R_RISCV_PCREL_LO12_I .L0
8: 00000297 auipc t0,0x0
8: R_RISCV_PCREL_HI20 var2
c: 0002b507 fld fa0,0(t0) # 8 <.text+0x8>
c: R_RISCV_PCREL_LO12_I .L0
10: 00000297 auipc t0,0x0
10: R_RISCV_PCREL_HI20 var3
14: 00a2a023 sw a0,0(t0) # 10 <.text+0x10>
14: R_RISCV_PCREL_LO12_S .L0
18: 00000297 auipc t0,0x0
18: R_RISCV_PCREL_HI20 var4
1c: 00a2b027 fsd fa0,0(t0) # 18 <.text+0x18>
1c: R_RISCV_PCREL_LO12_S .L0
The following example shows loading a constant using the %hi
and
%lo
assembler functions.
.equ UART_BASE, 0x40003080
lui a0, %hi(UART_BASE)
addi a0, a0, %lo(UART_BASE)
Which generates the following assembler output
as seen by objdump
:
0000000000000000 <.text>:
0: 40003537 lui a0,0x40003
4: 08050513 addi a0,a0,128 # 40003080 <UART_BASE>
The following pseudo instructions are available to call subroutines far from the current position:
call <symbol>
: call away subroutine1call <rd>, <symbol>
: call away subroutine2tail <symbol>
: tail call away subroutine3jump <symbol>, <rt>
: jump to away routine4
The following example shows how these pseudo instructions are used:
call func1
tail func2
jump func3, t0
Which generates the following assembler output and relocations
as seen by objdump
:
0000000000000000 <.text>:
0: 00000097 auipc ra,0x0
0: R_RISCV_CALL func1
4: 000080e7 jalr ra # 0x0
8: 00000317 auipc t1,0x0
8: R_RISCV_CALL func2
c: 00030067 jr t1 # 0x8
10: 00000297 auipc t0,0x0
10: R_RISCV_CALL func3
14: 00028067 jr t0 # 0x10
For floating-point instructions with a rounding mode field, the rounding mode
can be specified by adding an additional operand. e.g. fcvt.w.s
with
round-to-zero can be written as fcvt.w.s a0, fa0, rtz
. If unspecified, the
default dyn
rounding mode will be used.
Supported rounding modes are as follows (must be specified in lowercase):
rne
: round to nearest, ties to evenrtz
: round towards zerordn
: round downrup
: round uprmm
: round to nearest, ties to max magnitudedyn
: dynamic rounding mode (the rounding mode specified in thefrm
field of thefcsr
register is used)
The following code sample shows how to enable timer interrupts, set and wait for a timer interrupt to occur:
.equ RTC_BASE, 0x40000000
.equ TIMER_BASE, 0x40004000
# setup machine trap vector
1: auipc t0, %pcrel_hi(mtvec) # load mtvec(hi)
addi t0, t0, %pcrel_lo(1b) # load mtvec(lo)
csrrw zero, mtvec, t0
# set mstatus.MIE=1 (enable M mode interrupt)
li t0, 8
csrrs zero, mstatus, t0
# set mie.MTIE=1 (enable M mode timer interrupts)
li t0, 128
csrrs zero, mie, t0
# read from mtime
li a0, RTC_BASE
ld a1, 0(a0)
# write to mtimecmp
li a0, TIMER_BASE
li t0, 1000000000
add a1, a1, t0
sd a1, 0(a0)
# loop
loop:
wfi
j loop
# break on interrupt
mtvec:
csrrc t0, mcause, zero
bgez t0, fail # interrupt causes are less than zero
slli t0, t0, 1 # shift off high bit
srli t0, t0, 1
li t1, 7 # check this is an m_timer interrupt
bne t0, t1, fail
j pass
pass:
la a0, pass_msg
jal puts
j shutdown
fail:
la a0, fail_msg
jal puts
j shutdown
.section .rodata
pass_msg:
.string "PASS\n"
fail_msg:
.string "FAIL\n"
Pseudoinstruction | Base Instruction(s) | Meaning | Comment |
---|---|---|---|
la rd, symbol | auipc rd, symbol[31:12]; addi rd, rd, symbol[11:0] | Load address | With .option nopic (Default) |
la rd, symbol | auipc rd, symbol@GOT[31:12]; l{w|d} rd, symbol@GOT[11:0](rd) | Load address | With .option pic |
lla rd, symbol | auipc rd, symbol[31:12]; addi rd, rd, symbol[11:0] | Load local address | |
lga rd, symbol | auipc rd, symbol@GOT[31:12]; l{w|d} rd, symbol@GOT[11:0](rd) | Load global address | |
l{b|h|w|d} rd, symbol | auipc rd, symbol[31:12]; l{b|h|w|d} rd, symbol[11:0](rd) | Load global | |
s{b|h|w|d} rd, symbol, rt | auipc rt, symbol[31:12]; s{b|h|w|d} rd, symbol[11:0](rt) | Store global | |
fl{w|d} rd, symbol, rt | auipc rt, symbol[31:12]; fl{w|d} rd, symbol[11:0](rt) | Floating-point load global | |
fs{w|d} rd, symbol, rt | auipc rt, symbol[31:12]; fs{w|d} rd, symbol[11:0](rt) | Floating-point store global | |
nop | addi x0, x0, 0 | No operation | |
li rd, immediate | Myriad sequences | Load immediate | |
mv rd, rs | addi rd, rs, 0 | Copy register | |
not rd, rs | xori rd, rs, -1 | Ones’ complement | |
neg rd, rs | sub rd, x0, rs | Two’s complement | |
negw rd, rs | subw rd, x0, rs | Two’s complement word | |
sext.b rd, rs | slli rd, rs, XLEN - 8; srai rd, rd, XLEN - 8 | Sign extend byte | It will expand to another instruction sequence when B extension is available*[1] |
sext.h rd, rs | slli rd, rs, XLEN - 16; srai rd, rd, XLEN - 16 | Sign extend half word | It will expand to another instruction sequence when B extension is available*[1] |
sext.w rd, rs | addiw rd, rs, 0 | Sign extend word | |
zext.b rd, rs | andi rd, rs, 255 | Zero extend byte | |
zext.h rd, rs | slli rd, rs, XLEN - 16; srli rd, rd, XLEN - 16 | Zero extend half word | It will expand to another instruction sequence when B extension is available*[1] |
zext.w rd, rs | slli rd, rs, XLEN - 32; srli rd, rd, XLEN - 32 | Zero extend word | It will expand to another instruction sequence when B extension is available*[1] |
seqz rd, rs | sltiu rd, rs, 1 | Set if = zero | |
snez rd, rs | sltu rd, x0, rs | Set if != zero | |
sltz rd, rs | slt rd, rs, x0 | Set if < zero | |
sgtz rd, rs | slt rd, x0, rs | Set if > zero | |
fmv.s rd, rs | fsgnj.s rd, rs, rs | Copy single-precision register | |
fabs.s rd, rs | fsgnjx.s rd, rs, rs | Single-precision absolute value | |
fneg.s rd, rs | fsgnjn.s rd, rs, rs | Single-precision negate | |
fmv.d rd, rs | fsgnj.d rd, rs, rs | Copy double-precision register | |
fabs.d rd, rs | fsgnjx.d rd, rs, rs | Double-precision absolute value | |
fneg.d rd, rs | fsgnjn.d rd, rs, rs | Double-precision negate | |
beqz rs, offset | beq rs, x0, offset | Branch if = zero | |
bnez rs, offset | bne rs, x0, offset | Branch if != zero | |
blez rs, offset | bge x0, rs, offset | Branch if ≤ zero | |
bgez rs, offset | bge rs, x0, offset | Branch if ≥ zero | |
bltz rs, offset | blt rs, x0, offset | Branch if < zero | |
bgtz rs, offset | blt x0, rs, offset | Branch if > zero | |
bgt rs, rt, offset | blt rt, rs, offset | Branch if > | |
ble rs, rt, offset | bge rt, rs, offset | Branch if ≤ | |
bgtu rs, rt, offset | bltu rt, rs, offset | Branch if >, unsigned | |
bleu rs, rt, offset | bgeu rt, rs, offset | Branch if ≤, unsigned | |
j offset | jal x0, offset | Jump | |
jal offset | jal x1, offset | Jump and link | |
jr rs | jalr x0, rs, 0 | Jump register | |
jalr rs | jalr x1, rs, 0 | Jump and link register | |
ret | jalr x0, x1, 0 | Return from subroutine | |
call offset | auipc x6, offset[31:12]; jalr x1, x6, offset[11:0] | Call far-away subroutine | |
tail offset | auipc x6, offset[31:12]; jalr x0, x6, offset[11:0] | Tail call far-away subroutine | |
fence | fence iorw, iorw | Fence on all memory and I/O |
- [1] We don't specify the code sequence when the B-extension is present, since B-extension still not ratified or frozen. We will specify the expansion sequence once it's frozen.
Pseudoinstruction | Base Instruction(s) | Meaning |
---|---|---|
rdinstret[h] rd | csrrs rd, instret[h], x0 | Read instructions-retired counter |
rdcycle[h] rd | csrrs rd, cycle[h], x0 | Read cycle counter |
rdtime[h] rd | csrrs rd, time[h], x0 | Read real-time clock |
csrr rd, csr | csrrs rd, csr, x0 | Read CSR |
csrw csr, rs | csrrw x0, csr, rs | Write CSR |
csrs csr, rs | csrrs x0, csr, rs | Set bits in CSR |
csrc csr, rs | csrrc x0, csr, rs | Clear bits in CSR |
csrwi csr, imm | csrrwi x0, csr, imm | Write CSR, immediate |
csrsi csr, imm | csrrsi x0, csr, imm | Set bits in CSR, immediate |
csrci csr, imm | csrrci x0, csr, imm | Clear bits in CSR, immediate |
frcsr rd | csrrs rd, fcsr, x0 | Read FP control/status register |
fscsr rd, rs | csrrw rd, fcsr, rs | Swap FP control/status register |
fscsr rs | csrrw x0, fcsr, rs | Write FP control/status register |
frrm rd | csrrs rd, frm, x0 | Read FP rounding mode |
fsrm rd, rs | csrrw rd, frm, rs | Swap FP rounding mode |
fsrm rs | csrrw x0, frm, rs | Write FP rounding mode |
fsrmi rd, imm | csrrwi rd, frm, imm | Swap FP rounding mode, immediate |
fsrmi imm | csrrwi x0, frm, imm | Write FP rounding mode, immediate |
frflags rd | csrrs rd, fflags, x0 | Read FP exception flags |
fsflags rd, rs | csrrw rd, fflags, rs | Swap FP exception flags |
fsflags rs | csrrw x0, fflags, rs | Write FP exception flags |
fsflagsi rd, imm | csrrwi rd, fflags, imm | Swap FP exception flags, immediate |
fsflagsi imm | csrrwi x0, fflags, imm | Write FP exception flags, immediate |