+"""core of the python-based POWER9 simulator
+
+this is part of a cycle-accurate POWER9 simulator. its primary purpose is
+not speed, it is for both learning and educational purposes, as well as
+a method of verifying the HDL.
+"""
+
from functools import wraps
from soc.decoder.orderedset import OrderedSet
from soc.decoder.selectable_int import (FieldSelectableInt, SelectableInt,
'VRSAVE': 256}
+def swap_order(x, nbytes):
+ x = x.to_bytes(nbytes, byteorder='little')
+ x = int.from_bytes(x, byteorder='big', signed=False)
+ return x
+
+
def create_args(reglist, extra=None):
args = OrderedSet()
for reg in reglist:
class Mem:
- def __init__(self, bytes_per_word=8, initial_mem=None):
+ def __init__(self, row_bytes=8, initial_mem=None):
self.mem = {}
- self.bytes_per_word = bytes_per_word
- self.word_log2 = math.ceil(math.log2(bytes_per_word))
+ self.bytes_per_word = row_bytes
+ self.word_log2 = math.ceil(math.log2(row_bytes))
+ print ("Sim-Mem", initial_mem, self.bytes_per_word, self.word_log2)
if not initial_mem:
return
- print ("Sim-Mem", initial_mem, self.bytes_per_word)
+
+ # different types of memory data structures recognised (for convenience)
+ if isinstance(initial_mem, list):
+ initial_mem = (0, initial_mem)
+ if isinstance(initial_mem, tuple):
+ startaddr, mem = initial_mem
+ initial_mem = {}
+ for i, val in enumerate(mem):
+ initial_mem[startaddr + row_bytes*i] = (val, row_bytes)
+
for addr, (val, width) in initial_mem.items():
- self.st(addr, val, width)
+ #val = swap_order(val, width)
+ self.st(addr, val, width, swap=False)
def _get_shifter_mask(self, wid, remainder):
shifter = ((self.bytes_per_word - wid) - remainder) * \
8 # bits per byte
+ # XXX https://bugs.libre-soc.org/show_bug.cgi?id=377
+ # BE/LE mode?
+ shifter = remainder * 8
mask = (1 << (wid * 8)) - 1
print ("width,rem,shift,mask", wid, remainder, hex(shifter), hex(mask))
return shifter, mask
# TODO: Implement ld/st of lesser width
- def ld(self, address, width=8):
+ def ld(self, address, width=8, swap=True):
print("ld from addr 0x{:x} width {:d}".format(address, width))
remainder = address & (self.bytes_per_word - 1)
address = address >> self.word_log2
print ("masking", hex(val), hex(mask<<shifter), shifter)
val = val & (mask << shifter)
val >>= shifter
+ if swap:
+ val = swap_order(val, width)
print("Read 0x{:x} from addr 0x{:x}".format(val, address))
return val
- def st(self, addr, v, width=8):
+ def st(self, addr, v, width=8, swap=True):
+ staddr = addr
remainder = addr & (self.bytes_per_word - 1)
addr = addr >> self.word_log2
- print("Writing 0x{:x} to addr 0x{:x}/{:x}".format(v, addr, remainder))
+ print("Writing 0x{:x} to ST 0x{:x} memaddr 0x{:x}/{:x}".format(v,
+ staddr, addr, remainder, swap))
assert remainder & (width - 1) == 0, "Unaligned access unsupported!"
+ if swap:
+ v = swap_order(v, width)
if width != self.bytes_per_word:
if addr in self.mem:
val = self.mem[addr]
def __call__(self, ridx):
return self[ridx]
-
-
+
class ISACaller:
# decoder2 - an instance of power_decoder2
# regfile - a list of initial values for the registers
+ # initial_{etc} - initial values for SPRs, Condition Register, Mem, MSR
def __init__(self, decoder2, regfile, initial_sprs=None, initial_cr=0,
- initial_mem=None, initial_msr=0):
+ initial_mem=None, initial_msr=0,
+ initial_insns=None):
if initial_sprs is None:
initial_sprs = {}
if initial_mem is None:
initial_mem = {}
+ if initial_insns is None:
+ initial_insns = {}
+ self.respect_pc = False
+ else:
+ # setup batch of instructions: we want to respect (follow) the PC
+ self.respect_pc = True
+
self.gpr = GPR(decoder2, regfile)
- self.mem = Mem(initial_mem=initial_mem)
+ self.mem = Mem(row_bytes=8, initial_mem=initial_mem)
+ self.insns = Mem(row_bytes=4, initial_mem=initial_insns)
self.pc = PC()
self.spr = SPR(decoder2, initial_sprs)
self.msr = SelectableInt(initial_msr, 64) # underlying reg
self.namespace['CA'] = self.spr['XER'][XER_bits['CA']].value
self.namespace['CA32'] = self.spr['XER'][XER_bits['CA32']].value
- def handle_carry_(self, inputs, outputs):
+ def handle_carry_(self, inputs, outputs, already_done):
inv_a = yield self.dec2.e.invert_a
if inv_a:
inputs[0] = ~inputs[0]
gts = [(x > output) for x in inputs]
print(gts)
cy = 1 if any(gts) else 0
- self.spr['XER'][XER_bits['CA']] = cy
+ if not (1 & already_done):
+ self.spr['XER'][XER_bits['CA']] = cy
print ("inputs", inputs)
# 32 bit carry
gts = [(x[32:64] > output[32:64]) == SelectableInt(1, 1)
for x in inputs]
cy32 = 1 if any(gts) else 0
- self.spr['XER'][XER_bits['CA32']] = cy32
+ if not (2 & already_done):
+ self.spr['XER'][XER_bits['CA32']] = cy32
def handle_overflow(self, inputs, outputs):
inv_a = yield self.dec2.e.invert_a
imm = yield self.dec2.e.imm_data.data
inputs.append(SelectableInt(imm, 64))
assert len(outputs) >= 1
+ print ("handle_overflow", inputs, outputs)
if len(inputs) >= 2:
output = outputs[0]
+
+ # OV (64-bit)
input_sgn = [exts(x.value, x.bits) < 0 for x in inputs]
output_sgn = exts(output.value, output.bits) < 0
ov = 1 if input_sgn[0] == input_sgn[1] and \
output_sgn != input_sgn[0] else 0
+ # OV (32-bit)
+ input32_sgn = [exts(x.value, 32) < 0 for x in inputs]
+ output32_sgn = exts(output.value, 32) < 0
+ ov32 = 1 if input32_sgn[0] == input32_sgn[1] and \
+ output32_sgn != input32_sgn[0] else 0
+
self.spr['XER'][XER_bits['OV']] = ov
+ self.spr['XER'][XER_bits['OV32']] = ov32
so = self.spr['XER'][XER_bits['SO']]
so = so | ov
self.spr['XER'][XER_bits['SO']] = so
def set_pc(self, pc_val):
self.namespace['NIA'] = SelectableInt(pc_val, 64)
self.pc.update(self.namespace)
-
def call(self, name):
# TODO, asmregs is from the spec, e.g. add RT,RA,RB
results = info.func(self, *inputs)
print(results)
+ # detect if CA/CA32 already in outputs (sra*, basically)
+ already_done = 0
+ if info.write_regs:
+ output_names = create_args(info.write_regs)
+ for name in output_names:
+ if name == 'CA':
+ already_done |= 1
+ if name == 'CA32':
+ already_done |= 2
+
+ print ("carry already done?", bin(already_done))
carry_en = yield self.dec2.e.output_carry
if carry_en:
- yield from self.handle_carry_(inputs, results)
- ov_en = yield self.dec2.e.oe
- if ov_en:
+ yield from self.handle_carry_(inputs, results, already_done)
+ ov_en = yield self.dec2.e.oe.oe
+ ov_ok = yield self.dec2.e.oe.ok
+ if ov_en & ov_ok:
yield from self.handle_overflow(inputs, results)
rc_en = yield self.dec2.e.rc.data
if rc_en:
# any modified return results?
if info.write_regs:
- output_names = create_args(info.write_regs)
for name, output in zip(output_names, results):
if isinstance(output, int):
output = SelectableInt(output, 256)