dranke/upkr
1
0
Fork 0
mirror of https://github.com/exoticorn/upkr.git synced 2026-01-20 11:36:42 +01:00
Code Issues Projects Releases Wiki Activity

some more risc-v optimizations

Browse Source
This commit is contained in:
Dennis Ranke
2022-09-24 08:45:14 +02:00
parent 8c9e4311b9
commit ced6cc8c32
2 changed files with 56 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
asm_unpackers/Makefile
View File

@@ -6,10 +6,25 @@ test_riscv64: build/unpack_riscv64
qemu-riscv64 $< test_data.upk /tmp/out.bin qemu-riscv64 $< test_data.upk /tmp/out.bin
cmp test_data.bin /tmp/out.bin cmp test_data.bin /tmp/out.bin
build/unpack_riscv64.bin: unpack_riscv.S build/unpack_riscv64.o: unpack_riscv.S
mkdir -p build mkdir -p build
riscv64-linux-gnu-gcc -c -o build/unpack_riscv64.o $? riscv64-linux-gnu-gcc -c -o $@ $?
riscv64-linux-gnu-objcopy -O binary --only-section=.text build/unpack_riscv64.o $@
build/unpack_riscv64.bin: build/unpack_riscv64.o
riscv64-linux-gnu-objcopy -O binary --only-section=.text $? $@
disas-riscv64: build/unpack_riscv64.o
riscv64-linux-gnu-objdump -d $?
build/unpack_riscv32.o: unpack_riscv.S
mkdir -p build
riscv64-linux-gnu-gcc -march=rv32imc -mabi=ilp32 -c -o $@ $?
build/unpack_riscv32.bin: build/unpack_riscv32.o
riscv64-linux-gnu-objcopy -O binary --only-section=.text $? $@
disas-riscv32: build/unpack_riscv32.o
riscv64-linux-gnu-objdump -d $?
build/unpack_armv6m: ../c_unpacker/main.c unpack_armv6m.S build/unpack_armv6m: ../c_unpacker/main.c unpack_armv6m.S
mkdir -p build mkdir -p build
@@ -32,5 +47,5 @@ test_c: build/unpack_c
$< test_data.upk /tmp/out.bin $< test_data.upk /tmp/out.bin
cmp test_data.bin /tmp/out.bin cmp test_data.bin /tmp/out.bin
sizes: build/unpack_armv6m.bin build/unpack_riscv64.bin sizes: build/unpack_armv6m.bin build/unpack_riscv64.bin build/unpack_riscv32.bin
ls -l build/*.bin ls -l build/*.bin

76
asm_unpackers/unpack_riscv.S
View File

@@ -1,6 +1,6 @@
.section .text .section .text
#define FRAME_SIZE (256+64*4+4) #define FRAME_SIZE (256+32*4+4)
// x8 prob array ptr // x8 prob array ptr
// x9 prev was literal // x9 prev was literal
@@ -28,39 +28,39 @@ upkr_unpack:
jal upkr_decode_bit jal upkr_decode_bit
beqz x15, .Lliteral beqz x15, .Lliteral
li x14, 256 slli x14, x14, 8
beqz x9, .Lread_offset beqz x9, .Lread_offset_inc_x14
jal upkr_decode_bit jal upkr_decode_bit
beqz x15, .Lskip_offset bnez x15, .Lread_offset
.Lread_offset:
jal t3, upkr_decode_number
addi x12, x9, -1
beqz x12, .Ldone
.Lskip_offset: .Lfinished_offset:
li x14, 256+64 addi x14, x14, 64
jal t3, upkr_decode_number jal t3, upkr_decode_number
1: 1:
sub x15, x10, x12 add x14, x10, t0
lbu x15, (x15) lbu x14, (x14)
sb x15, (x10) .Lstore_byte:
sb x14, (x10)
addi x10, x10, 1 addi x10, x10, 1
addi x9, x9, -1 addi x9, x9, 1
bnez x9, 1b blt x9, x0, 1b
j .Lmainloop j .Lmainloop
.Lliteral: .Lliteral:
li x14, 1
1:
jal upkr_decode_bit jal upkr_decode_bit
addi x14, x14, -1
slli x14, x14, 1 slli x14, x14, 1
add x14, x14, x15 add x14, x14, x15
srli x9, x14, 8 srli x9, x14, 8
beqz x9, 1b beqz x9, .Lliteral
sb x14, 0(x10) j .Lstore_byte
addi x10, x10, 1
j .Lmainloop
.Lread_offset_inc_x14:
addi x14, x14, 1
.Lread_offset:
jal t3, upkr_decode_number
addi t0, x9, 1
bnez t0, .Lfinished_offset
.Ldone: .Ldone:
addi sp, sp, FRAME_SIZE addi sp, sp, FRAME_SIZE
mv x8, x17 mv x8, x17
@@ -68,16 +68,14 @@ upkr_unpack:
jr t4 jr t4
// x14 context index // x14 context index
// return: x9 decoded number // return: x9 negtive decoded number
upkr_decode_number: upkr_decode_number:
mv t5, x14 mv t5, x14
li x9, 0 li x9, 0
li x8, 1 li x8, -1
1: 1:
addi x14, x14, 1
jal upkr_decode_bit jal upkr_decode_bit
beqz x15, 1f beqz x15, 1f
addi x14, x14, 1
jal upkr_decode_bit jal upkr_decode_bit
beqz x15, 2f beqz x15, 2f
add x9, x9, x8 add x9, x9, x8
@@ -99,46 +97,46 @@ upkr_load_byte:
// x11 in ptr // x11 in ptr
// x13 state // x13 state
// x14 context index // x14 context index
// return: x15 decoded bit // return:
// x14 context index + 1
// x15 decoded bit
upkr_decode_bit: upkr_decode_bit:
srli x15, x13, 12 srli x15, x13, 12
beqz x15, upkr_load_byte beqz x15, upkr_load_byte
mv t0, x9
mv t1, x14 mv t1, x14
mv t2, x10 mv t2, x10
add x14, x14, sp add x14, x14, sp
lbu x9, 0(x14) lbu x12, 0(x14)
andi x10, x13, 255 andi x10, x13, 255
sltu x15, x10, x9 sltu x15, x10, x12
srli x13, x13, 8 srli x13, x13, 8
beqz x15, .Lelse beqz x15, .Lelse
mul x13, x13, x9 mul x13, x13, x12
add x13, x13, x10 add x13, x13, x10
li x10, 256 + 8 li x10, 256 + 8
sub x10, x10, x9 sub x10, x10, x12
srli x10, x10, 4 srli x10, x10, 4
add x9, x9, x10 add x12, x12, x10
j .Lendif j .Lendif
.Lelse: .Lelse:
li x16, 256 li x16, 256
sub x16, x16, x9 sub x16, x16, x12
mul x13, x13, x16 mul x13, x13, x16
add x13, x13, x10 add x13, x13, x10
sub x13, x13, x9 sub x13, x13, x12
addi x10, x9, 8 addi x10, x12, 8
srli x10, x10, 4 srli x10, x10, 4
sub x9, x9, x10 sub x12, x12, x10
.Lendif: .Lendif:
sb x9, 0(x14) sb x12, 0(x14)
mv x9, t0 addi x14, t1, 1
mv x14, t1
mv x10, t2 mv x10, t2
ret ret