def gen_mem_image():
# text section
text = """
# load array pointers
csrr x1, mngr2proc < 200
csrr x2, mngr2proc < 0x2000
csrr x3, mngr2proc < 0x3000
csrr x4, mngr2proc < 0x4000
addi x5, x0, 0
loop:
lw x6, 0(x2) # src0_real
lw x8, 0(x3) # src1_real
lw x7, 4(x2) # src0_imag
lw x9, 4(x3) # src1_imag
mul x10, x6, x8 # real * real
mul x11, x7, x9 # imag * imag
mul x12, x7, x8 # imag * real
mul x13, x6, x9 # real * imag
sub x14, x10,x11 # dest_real
add x15, x12,x13 # dest_imag
addi x5, x5, 2
addi x2, x2, 8
addi x3, x3, 8
sw x14, 0(x4)
sw x15, 4(x4)
addi x4, x4, 8
bne x5, x1, loop
# end of the program
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
"""
mem_image = assemble(text)
# load data by manually create data sections using binutils
src0_section = mk_section(".data", c_cmplx_mult_src0_ptr, src0)
mem_image.add_section(src0_section)
src1_section = mk_section(".data", c_cmplx_mult_src1_ptr, src1)
mem_image.add_section(src1_section)
return mem_image
def gen_mem_image():
# text section
text = \
"""
# load array pointers
csrr x1, mngr2proc < 100
csrr x2, mngr2proc < 0x2000
csrr x3, mngr2proc < 0x3000
csrr x4, mngr2proc < 0x4000
add x5, x0, x1
loop:
lw x6, 0(x2)
lw x7, 0(x3)
add x8, x6, x7
sw x8, 0(x4)
addi x2, x2, 4
addi x3, x3, 4
addi x4, x4, 4
addi x5, x5, -1
bne x5, x0, loop
# end of the program
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
"""
mem_image = assemble(text)
# load data by manually create data sections using binutils
src0_section = mk_section(".data", c_vvadd_src0_ptr, src0)
src1_section = mk_section(".data", c_vvadd_src1_ptr, src1)
# load data
mem_image.add_section(src0_section)
mem_image.add_section(src1_section)
return mem_image
def gen_mem_image():
# text section
text = """
csrr x4, mngr2proc < 0x2000
csrr x5, mngr2proc < 0x3000
csrr x6, mngr2proc < 0x4000
csrr x7, mngr2proc < 10
csrr x8, mngr2proc < 10
addi x24, x0, 64 # coeff0
addi x25, x0, 48 # coeff1
# Assume that nrows and ncols are positive and otherwise well-behaved
addi x2, x7, -1 # end condition nrows
addi x3, x8, -1 # end condition ncols
addi x9, x0, 1 # ridx starts at 1
# zero: row loop
zero:
addi x10, x0, 1
# one: col loop
# Calculate mask index
one:
mul x11, x8, x9 # ridx*ncols
add x11, x11, x10 # ridx*ncols + cidx
slli x11, x11, 2 # ridx*ncols + cidx (pointer)
add x12, x5, x11 # ridx*ncols + cidx (pointer) for mask
lw x12, 0(x12) # mask[ridx*ncols + cidx]
# If block
# if ( !mask[ridx*ncols + cidx] ) goto two:
beq x12, x0, two
add x12, x6, x11 # ridx*ncols + cidx (pointer) for src
lw x13, 0(x12) # src[ridx*ncols + cidx]
mul x13, x13, x24 # src[ridx*ncols + cidx] * coeff0
add x23, x13, x0 # out = src[ridx*ncols + cidx] * coeff0
lw x13, 4(x12) # src[ridx*ncols + (cidx+1)]
mul x13, x13, x25 # src[ridx*ncols + (cidx+1)] * coeff1
add x23, x23, x13 # out += src[ridx*ncols + (cidx+1)] * coeff1
lw x13, -4(x12) # src[ridx*ncols + (cidx-1)]
mul x13, x13, x25 # src[ridx*ncols + (cidx-1)] * coeff1
add x23, x23, x13 # out += src[ridx*ncols + (cidx-1)] * coeff1
addi x22, x9, 1 # ridx+1
mul x12, x8, x22 # (ridx+1)*ncols
add x12, x12, x10 # (ridx+1)*ncols + cidx
slli x12, x12, 2 # (ridx+1)*ncols + cidx (pointer)
add x13, x6, x12 # (ridx+1)*ncols + cidx (pointer) for src
lw x13, 0(x13) # src[(ridx+1)*ncols + cidx]
mul x14, x13, x25 # src[(ridx+1)*ncols + cidx] * coeff1
add x23, x23, x14 # out += src[(ridx+1)*ncols + cidx] * coeff1
addi x22, x9, -1 # ridx-1
mul x12, x8, x22 # (ridx-1)*ncols
add x12, x12, x10 # (ridx-1)*ncols + cidx
slli x12, x12, 2 # (ridx-1)*ncols + cidx (pointer)
add x13, x6, x12 # (ridx-1)*ncols + cidx (pointer) for src
lw x13, 0(x13) # src[(ridx-1)*ncols + cidx]
mul x14, x13, x25 # src[(ridx-1)*ncols + cidx] * coeff1
add x23, x23, x14 # out += src[(ridx-1)*ncols + cidx] * coeff1
add x12, x4, x11 # ridx*ncols + cidx (pointer) for dest
srai x23, x23, 8 # out >>= shamt
sw x23, 0(x12) # dest[ridx*ncols + cidx] = out
jal x0, three # End of if block, goto three:
# Else block
two:
add x12, x6, x11 # ridx*ncols + cidx (pointer) for src
lw x13, 0(x12) # src[ridx*ncols + cidx]
add x14, x4, x11 # ridx*ncols + cidx (pointer) for dest
sw x13, 0(x14) # dest[ridx*ncols + cidx] = src[ridx*ncols + cidx]
three:
addi x10, x10, 1 # cidx++
bne x10, x3, one # if ( cidx != ncols - 1 ) goto one:
addi x9, x9, 1 # ridx++
bne x9, x2, zero # if ( ridx != nrows - 1 ) goto zero:
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
"""
mem_image = assemble(text)
# load data by manually create data sections using binutils
src_section = mk_section(".data", c_masked_filter_src_ptr, src)
mem_image.add_section(src_section)
mask_section = mk_section(".data", c_masked_filter_mask_ptr, mask)
mem_image.add_section(mask_section)
return mem_image
def gen_mem_image():
# text section
text = \
"""
csrr x25, mngr2proc < 10
csrr x26, mngr2proc < 0x2000
csrr x27, mngr2proc < 0x3000
csrr x28, mngr2proc < 0x4000
add x29, x0, x25
#addi x2, x2, 120 # adjust stack size
#li x10, 3904692793
# lui x10 953294
# addi x10 x10 569
addi x2, x2, 512 # adjust stack size
lw x10,0(x26)
jal x1, Start
Start:
addi x2, x2, -28 # Adjust stack
sw x1, 24(x2) # Save return address
sw x8, 20(x2) # Save current fx1me pointer
addi x8, x2, 28 # Adjust fx1me pointer
lw x12, 12(x27) # Load last value from array
sw x12, -8(x8) # Save it for future use in loop
lw x12, 8(x27) # Load third value from array
sw x12, -12(x8) # Save it for future use in loop
lw x12, 4(x27) # Load second value from array
sw x12, -16(x8) # Save it for future use in loop
lw x11, 0(x27) # Load first value from array
sw x11, -20(x8) # Save it for future use
addi x16, x0, 0 # int out = 0
addi x11, x11, -1 # tabs[0]-1
srl x12, x10, x11 # seed>>(tabs[0]-1)
andi x12, x12, 1 # feedback = bit read from seed
addi x13, x0, 1 # int i=1
jal x0, FOR_FEEDBACK
FOR_FEEDBACK: # for(int i=1;i<4;i++)
addi x14, x0, 4
blt x13, x14, CNT_FEEDBACK # if i<4 count feedback value
jal x0, CNT_SEED # else go to counting seed value
#x10 seed x11 tab[0] x12 feedback x13 i x14 4
CNT_FEEDBACK:
slli x14, x13, 2 # i*4 - used for addressing
addi x17, x8, -20 # -20(x8) = address of first array element
add x14, x14, x17 # decode address of array element
lw x14, 0(x14) # tabs[i]
addi x14, x14, -1 # (tabs[i]-1)
srl x15, x10, x14 # seed>>(tabs[1]-1))
andi x15, x15, 1 # seed>>(tabs[1]-1))&1
xor x12, x12, x15 # feedback = feedback ^ ((seed>>(tabs[i]-1))&1);
addi x13, x13, 1 # i++
jal x0, FOR_FEEDBACK
#x10 seed x11 tab[0] x12 feedback x13 i x14 tabs[i]-1 x15 seed>>(tabs[1]-1))&1 x17 -28
CNT_SEED:
slli x10, x10, 1 # seed = (seed<<1);
or x10, x10, x12 # seed |= feedback;
addi x13, x0, 31 # int i = 31
jal x0, FOR_OUT
FOR_OUT:
blt x13, x0, END # if i<0 end function
jal x0, CNT_OUT # else count out value
CNT_OUT:
slli x14, x16, 1 # (out << 1)
srl x15, x10, x13 # (seed>>i)
andi x15, x15, 1 # ((seed>>i)&1)
or x16, x14, x15 # out = (out << 1) | ((seed>>i)&1);
addi x13, x13, -1 # i--
jal x0, FOR_OUT
END:
sw x16, 0(x28)
addi x28, x28, 4
addi x10, x16 0 # start again with init value = out
addi x29, x29, -1
beq x29, x0, final
jalr x0, x1, 0
final:
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
"""
mem_image = assemble(text)
# load data by manually create data sections using binutils
src0_section = mk_section(".data", c_vvadd_src0_ptr, src0)
src1_section = mk_section(".data", c_vvadd_src1_ptr, src1)
# load data
mem_image.add_section(src0_section)
mem_image.add_section(src1_section)
return mem_image
def gen_mem_image():
# text section
text = ("""
# load array pointers
csrr x1, mngr2proc < {} # size
csrr x2, mngr2proc < 0x2000 # src pointer
csrr x3, mngr2proc < 0x3000 # mask pointer
csrr x4, mngr2proc < 0x4000 # dst pointer
addi x5, x0, 16 # shift amount
lw x6, 0(x3) # mask = 0xffff
addi x11, x0, 1 # Go bit
add x7, x0, x0 # cksum = 0
loop_i:
lw x8, 0(x2) # load first 32-bit word
lw x9, 4(x2) # load second 32-bit word
lw x10, 8(x2) # load third 32-bit word
# transfer data to accelerator
csrw 0x7E0, x7
csrw 0x7E1, x8
csrw 0x7E2, x9
csrw 0x7E3, x10
csrw 0x7E4, x11
# read back the result
csrr x7, 0x7E5
addi x2, x2, 12 # increment word address
addi x1, x1, -6 # decrement loop counter i
bne x1, x0, loop_i
sw x7, 0(x4)
# End of program
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
""".format(c_cksum_size))
mem_image = assemble(text)
# load data by manually create data sections using binutils
src_section = mk_section(".data", c_cksum_src_ptr, src)
msk_section = mk_section(".data", c_cksum_msk_ptr, mask)
# load data
mem_image.add_section(src_section)
mem_image.add_section(msk_section)
return mem_image
def gen_mem_image():
# text section
text = ("""
# load array pointers
csrr x1, mngr2proc < {} # size
csrr x2, mngr2proc < 0x2000 # src pointer
csrr x14, mngr2proc < 0x3000 # mask pointer
csrr x3, mngr2proc < 0x4000 # dst pointer
add x4, x0, x1 # loop var i
addi x5, x0, 16 # shift amount
lw x13, 0(x14) # mask = 0xffff
loop_i:
add x6, x0, x0 # sum1 = 0
add x7, x0, x0 # sum2 = 0
addi x10, x0, 4
loop_j:
lw x8, 0(x2) # load 2 16-bit ints
and x9, x8, x13 # src[j]
srl x8, x8, x5 # src[j+1]
add x6, x6, x9 # sum1 += src[j]
and x6, x6, x13 # sum1 &= 0xffff
add x7, x7, x6 # sum2 += sum1
and x7, x7, x13 # sum2 &= 0xffff
add x6, x6, x8 # sum1 += src[j+1]
and x6, x6, x13 # sum1 &= 0xffff
add x7, x7, x6 # sum2 += sum1
and x7, x7, x13 # sum2 &= 0xffff
addi x2, x2, 4 # j+=2
addi x10, x10, -1 # decrement loop counter
bne x10, x0, loop_j
add x11, x0, x7 # ret = sum2
sll x11, x11, x5 # ret = ret << 16
add x11, x11, x6 # ret |= sum1
sw x11 0(x3) # src[i] = ret
addi x3, x3, 4
addi x4, x4, -1
bne x4, x0, loop_i
# End of program
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
""".format(c_cksum_size))
mem_image = assemble(text)
# load data by manually create data sections using binutils
src_section = mk_section(".data", c_cksum_src_ptr, src)
msk_section = mk_section(".data", c_cksum_msk_ptr, mask)
# load data
mem_image.add_section(src_section)
mem_image.add_section(msk_section)
return mem_image
def gen_mem_image():
# text section
text = \
"""
csrr x1, mngr2proc < 50
csrr x2, mngr2proc < 0x2000
csrr x3, mngr2proc < 0x3000
csrr x4, mngr2proc < 0x4000
add x9, x0, x1
jal x0, start
start:
lw x6, 0(x2)
lw x7, 0(x3)
addi x2, x2, 4
addi x3, x3, 4
jal x0, loop
loop:
bge x6, x7, elseif
addi x28, x6, 0
addi x6, x7, 0
addi x7, x28, 0
jal x0, loop
elseif:
beq x7, x0, else
sub x6, x6, x7
jal x0, loop
else:
addi x5, x6, 0
sw x5, 0(x4)
addi x4, x4, 4
addi x9, x9, -1
bne x9, x0, start
jal x0, end
end:
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
"""
mem_image = assemble( text )
# load data by manually create data sections using binutils
src0_section = mk_section( ".data", c_vvadd_src0_ptr, src0 )
src1_section = mk_section( ".data", c_vvadd_src1_ptr, src1 )
# load data
mem_image.add_section( src0_section )
mem_image.add_section( src1_section )
return mem_image
def gen_mem_image():
# text section
text = \
"""
csrr x25, mngr2proc < 10
csrr x26, mngr2proc < 0x2000
csrr x27, mngr2proc < 0x3000
add x28, x0, x25
addi x2, x2, 120 # adjust stack size
jal x0, main
main:
lw x10,0(x26)
addi x26, x26, 4
jal x1, fib #call the fib function
sw x10, 0(x2) #result on stack
sw x10, 0(x27) #save to array
addi x27, x27, 4
addi x28, x28, -1
bne x28, x0, main
jal x0, final
fib:
addi x2, x2,-12 #save all values and return adress
sw x1, 0(x2)
sw x8, 4(x2)
sw x9, 8(x2)
addi x8, x10, 0
beq x0,x8, done #get out of this loop if current one is equal to zero
addi x5, x0, 1
beq x5,x8, done #same if current one is equal to one
addi x10, x8, -1 #do the loop with n-1
jal x1, fib
addi x9,x18, 0 #when we gout out of this loop we put the result of the loop in s1
addi x10,x8, -2 #do the loop with n-2
jal x1, fib
add x18,x18,x9 #when we get out, we add the final term of this (n-1)+(n-2) iteration
jal x0, end
done:
addi x18,x8, 0 #value of the term is 0 or 1
jal x0, end
end:
addi x10,x18, 0 #load result in a0
lw x9,8(x2) #restore all values and stack pointer
lw x8,4(x2)
lw x1,0(x2)
addi x2,x2,12
jalr x0, x1, 0
final:
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
"""
mem_image = assemble(text)
# load data by manually create data sections using binutils
src0_section = mk_section(".data", c_vvadd_src0_ptr, src0)
# load data
mem_image.add_section(src0_section)
return mem_image
def gen_mem_image():
# text section
text = ("""
# load array pointers
csrr x1, mngr2proc < {} # size
csrr x2, mngr2proc < 0x2000 # src pointer
csrr x3, mngr2proc < 0x3000 # mask pointer
csrr x4, mngr2proc < 0x4000 # dst pointer
addi x5, x0, 16 # shift amount
lw x6, 0(x3) # mask = 0xffff
add x7, x0, x0 # sum1 = 0
add x8, x0, x0 # sum2 = 0
add x9, x0, x0 # sum1_prev = 0
add x10, x0, x0 # sum2_prev = 0
loop_i:
and x7, x9, x6 # sum1 = sum1_prev & 0xffff
and x8, x7, x6 # sum2 = sum1 & 0xffff
add x7, x7, x10 # sum1 += sum2_prev
and x7, x7, x6 # sum1 &= 0xffff
add x8, x8, x7 # sum2 += sum1
and x8, x8, x6 # sum2 &= 0xffff
addi x11, x0, 3 # j = 3
loop_j:
lw x12, 0(x2) # load 2 16-bit ints
and x13, x12, x6 # src[i+j] = word & 0xffff
srl x12, x12, x5 # src[i+j+1] = word >> 16
add x7, x7, x13 # sum1 += src[i+j]
and x7, x7, x6 # sum1 &= 0xffff
add x8, x8, x7 # sum2 += sum1
and x8, x8, x6 # sum2 &= 0xffff
add x7, x7, x12 # sum1 += src[i+j+1]
and x7, x7, x6 # sum1 &= 0xffff
add x8, x8, x7 # sum2 += sum1
and x8, x8, x6 # sum2 &= 0xffff
addi x2, x2, 4 # increment word address
addi x11, x11, -1 # decrement loop counter
bne x11, x0, loop_j
add x9, x0, x7 # sum1_prev = sum1
add x10, x0, x8 # sum2_prev = sum2
addi x1, x1, -6 # decrement loop counter i
bne x1, x0, loop_i
add x14, x0, x8 # ret = sum2
sll x14, x14, x5 # ret = sum2 << 16
add x14, x14, x7 # ret |= sum1
sw x14, 0(x4)
# End of program
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
""".format(c_cksum_size))
mem_image = assemble(text)
# load data by manually create data sections using binutils
src_section = mk_section(".data", c_cksum_src_ptr, src)
msk_section = mk_section(".data", c_cksum_msk_ptr, mask)
# load data
mem_image.add_section(src_section)
mem_image.add_section(msk_section)
return mem_image
def gen_mem_image():
# text section
text = \
"""
csrr x25, mngr2proc < 3
csrr x26, mngr2proc < 0x2000
csrr x27, mngr2proc < 0x3000
add x28, x0, x25 # Load number of tests
addi x2, x2, 256
jal x0, Start
############# BRESENHAM
# only values from range 1-32 !!!
Start:
lw x10,0(x26) # Load value to x0
lw x11,4(x26) # Load value to y0
lw x12,8(x26) # Load value to x1
lw x13,12(x26) # Load value to y1
addi x26, x26, 16
addi x2, x2, -156 # Adjust stack
sw x1, 172(x2) # Save return address
sw x8, 168(x2) # Save previous frame pointer
addi x8, x2, 140 # Adjust frame pointer
sw x10, 0(x9) # Save x0
sw x11, 4(x9) # Save y0
sw x12, 8(x9) # Save x1
sw x13, 12(x9) # Save y1
sub x10, x10, x12 # dx = x0 - x1
sw x10, 16(x9) # save dx
blt x10, x0, .ABS_DX # if dx < 0 count abs(dx)
jal x0, .CNT_DY # else go to dy init
.ABS_DX:
sub x10, x0, x10 # Reverse value
sw x10, 16(x9) # save abs(dx)
jal x0, .CNT_DY # go to dy init
.CNT_DY:
sub x11, x11, x13 # dy = y0 - y1
sw x11, 20(x9) # dy = y0 - y1
blt x11, x0, .ABS_DY
jal x0, .CNT_SX
.ABS_DY:
sub x11, x0, x11 # Reverse value
sw x11, 20(x9) # save abs(dy)
jal x0, .CNT_SX
.CNT_SX:
lw x10, 0(x9) # load x0
bge x10, x12, .SX_NEG # if (x0 < x1) sx = -1
addi x10, x0, 1 # else sx = 1
sw x10, 24(x9) # save sx
jal x0, .CNT_SY
.SX_NEG:
addi x10, x0, -1 # sx = -1
sw x10, 24(x9) # save sx
jal x0, .CNT_SY
.CNT_SY:
lw x10, 4(x9) # Load y0
bge x10, x13, .SY_NEG # if (y0 < y1) sy = -1
addi x10, x0, 1 # else sy = 1
sw x10, 28(x9) # save sx
jal x0, .DX_VS_DY
.SY_NEG:
addi x10, x0, -1 # sy = -1
sw x10, 28(x9) # save sy
jal x0, .DX_VS_DY
.DX_VS_DY:
lw x10, 16(x9) # load dx
blt x11, x10, .SET_ERR # if dy < dx then err = (dx>>1)
srai x11, x11, 1 # else (dy>>1)
sub x11, x0, x11 # err = - (dy>>1)
sw x11, 32(x9) # save err
jal x0, .LOOP
.SET_ERR:
srai x10, x10, 1 # err = (dx>>1)
sw x10, 32(x9) # save err
jal x0, .LOOP
.LOOP:
lw x10, 0(x9) # Load x0
addi x10, x10, -1 # x0-1
addi x11, x0, 1 # prepare value 1
sll x10, x11, x10 # shift 1 left (x0-1) times
lw x11, 4(x9) # Load y0
slli x11, x11, 2 # multiply by 4 to use as address
addi x12, x8, -152 # get proper address in stack
add x11, x11, x12 # count targeted address in stack
lw x12, -4(x11) # get value of arr[y0-1]
or x10, x10, x12 # or it with 1<<(x0-1)
sw x10, -4(x11) # save the result in the same place
lw x10, 0(x9) # Load x0
lw x11, 8(x9) # Load x1
bne x10, x11, .CONTINUE # if(x0!=x1) continue loop
lw x10, 4(x9) # Load y0 (y1 is still in x13)
bne x10, x13, .CONTINUE # if(y0!=y1) continue loop
jal x0, .BREAK # else BREAK LOOP!!!!!!!!
.CONTINUE:
lw x10, 32(x9) # load err
sw x10, 36(x9) # save it as e2 for future use
lw x11, 16(x9) # load dx
sub x11, x0, x11 # -dx
blt x11, x10, .CONDITION_1 # if(-dx < e2)
jal x0, .NEXT_COND
.CONDITION_1:
lw x11, 20(x9) # Load dy
sub x10, x10, x11 # err -= dy
sw x10, 32(x9) # save err
lw x10, 24(x9) # load sx
lw x11, 0(x9) # load x0
add x11, x11, x10 # x0 += sx
sw x11, 0(x9) # save x0
jal x0, .NEXT_COND
.NEXT_COND:
lw x10, 36(x9) # load e2
lw x11, 20(x9) # load dy
blt x10, x11, .CONDITION_2 # if(e2 < dy)
jal x0, .LOOP # else do another LOOP itex1tion
.CONDITION_2:
lw x10, 16(x9) # Load dx
lw x11, 32(x9) # Load err
add x11, x11, x10 # err += dx;
sw x11, 32(x9) # save err
lw x10, 28(x9) # load sy
lw x11, 4(x9) # load y0
add x11, x11, x10 # y0 += sy;
sw x11, 4(x9) # save y0
jal x0, .LOOP
.BREAK:
lw x11 -152(x8)
sw x11 0(x27)
lw x11 -148(x8)
sw x11 4(x27)
lw x11 -144(x8)
sw x11 8(x27)
lw x11 -140(x8)
sw x11 12(x27)
lw x11 -136(x8)
sw x11 16(x27)
lw x11 -132(x8)
sw x11 20(x27)
lw x11 -128(x8)
sw x11 24(x27)
lw x11 -124(x8)
sw x11 28(x27)
lw x11 -120(x8)
sw x11 32(x27)
lw x11 -116(x8)
sw x11 36(x27)
lw x11 -112(x8)
sw x11 40(x27)
lw x11 -108(x8)
sw x11 44(x27)
lw x11 -104(x8)
sw x11 48(x27)
lw x11 -100(x8)
sw x11 52(x27)
lw x11 -96(x8)
sw x11 56(x27)
lw x11 -92(x8)
sw x11 60(x27)
lw x11 -88(x8)
sw x11 64(x27)
lw x11 -84(x8)
sw x11 68(x27)
lw x11 -80(x8)
sw x11 72(x27)
lw x11 -76(x8)
sw x11 76(x27)
lw x11 -72(x8)
sw x11 80(x27)
lw x11 -68(x8)
sw x11 84(x27)
lw x11 -64(x8)
sw x11 88(x27)
lw x11 -60(x8)
sw x11 92(x27)
lw x11 -56(x8)
sw x11 96(x27)
lw x11 -52(x8)
sw x11 100(x27)
lw x11 -48(x8)
sw x11 104(x27)
lw x11 -44(x8)
sw x11 108(x27)
lw x11 -40(x8)
sw x11 112(x27)
lw x11 -36(x8)
sw x11 116(x27)
lw x11 -32(x8)
sw x11 120(x27)
lw x11 -28(x8)
sw x11 124(x27)
addi x27, x27, 128 # test counter
addi x28, x28, -1
sw x0 -152(x8) # clear memory
sw x0 -148(x8)
sw x0 -144(x8)
sw x0 -140(x8)
sw x0 -136(x8)
sw x0 -132(x8)
sw x0 -128(x8)
sw x0 -124(x8)
sw x0 -120(x8)
sw x0 -116(x8)
sw x0 -112(x8)
sw x0 -108(x8)
sw x0 -104(x8)
sw x0 -100(x8)
sw x0 -96(x8)
sw x0 -92(x8)
sw x0 -88(x8)
sw x0 -84(x8)
sw x0 -80(x8)
sw x0 -76(x8)
sw x0 -72(x8)
sw x0 -68(x8)
sw x0 -64(x8)
sw x0 -60(x8)
sw x0 -56(x8)
sw x0 -52(x8)
sw x0 -48(x8)
sw x0 -44(x8)
sw x0 -40(x8)
sw x0 -36(x8)
sw x0 -32(x8)
sw x0 -28(x8)
addi x2, x2, 156 # move back stack
bne x28, x0, Start #start again
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
"""
mem_image = assemble(text)
# load data by manually create data sections using binutils
src0_section = mk_section(".data", c_vvadd_src0_ptr, src0)
# load data
mem_image.add_section(src0_section)
return mem_image
def gen_mem_image():
# text section
text = """
# load array pointers
csrr x1, mngr2proc < 0x2000
csrr x2, mngr2proc < 0x3000
csrr x3, mngr2proc < 20
csrr x4, mngr2proc < 0x4000
csrr x5, mngr2proc < 0x5000
csrr x6, mngr2proc < 50
addi x7, x0, 0 # loop counter i is in x7
zero:
slli x25, x7, 2 # multiply by 4 to get i in the index form
add x9, x1, x25 # pointer to i in srch_keys
lw x9, 0(x9) # key = srch_keys[i]
addi x10, x0, 0 # idx_min
srai x11, x6, 1 # idx_mid = dict_sz/2
addi x12, x6, -1 # idx_max = (dict_sz-1)
addi x13, x0, 0 # done = false
addi x14, x0, -1 # -1
add x15, x2, x25 # i pointer in srch_values
sw x14, 0(x15) # srch_values[i] = -1
one:
slli x24, x11, 2 # idx_mid in pointer form
add x16, x4, x24 # idx_mid pointer in dict_keys
lw x17, 0(x16) # midkey = dict_keys[idx_mid]
bne x9, x17, two # if ( key == midkey ) goto two:
# if block starts
add x16, x5, x24 # idx_mid pointer in dict_values
lw x18, 0(x16) # dict_values[idx_mid]
add x15, x2, x25 # i pointer in srch_values
sw x18, 0(x15) # srch_values[i] = dict_values[idx_mid]
addi x13, x0, 1 # done = true
# if block ends
two:
slt x18, x17, x9 # midkey < key
beq x18, x0, three # if ( midkey < key ) goto three
# if block for midkey < key
addi x10, x11, 1 # idx_min = idx_mid + 1
jal x0, four # if ( midkey > key ) goto four:
# end of if block
# else block
three:
slt x18, x9, x17 # if midkey > key
beq x18, x0, four # if ( midkey > key ) goto four:
# if block for midkey > key
addi x12, x11, -1 # idx_max = idx_mid - 1
# end of if block
four:
add x20, x10, x12 # idx_min + idx_max
srai x11, x20, 1 # idx_mid = ( idx_min + idx_max ) / 2
slt x21, x12, x10 # idx_max < idx_min
or x22, x21, x13 # done || (idx_max < idx_min)
# while
# ( !(done || (idx_max < idx_min)) )
# goto one:
beq x22, x0, one
addi x7, x7, 1 # i++
bne x7, x3, zero # if (i < srch_sz) goto 0:
# end of the program
csrw proc2mngr, x0 > 0
nop
nop
nop
nop
nop
nop
"""
mem_image = assemble(text)
# load data by manually create data sections using binutils
d_keys_section = mk_section(".data", c_bin_search_d_keys_ptr, d_keys)
mem_image.add_section(d_keys_section)
d_values_section = mk_section(".data", c_bin_search_d_values_ptr,
d_values)
mem_image.add_section(d_values_section)
s_keys_section = mk_section(".data", c_bin_search_s_keys_ptr, s_keys)
mem_image.add_section(s_keys_section)
return mem_image