From: Luke Kenneth Casson Leighton Date: Wed, 5 Jun 2019 05:43:14 +0000 (+0100) Subject: add mirror copy of FU_Regs Dep Matrix, names changed, adapted to LD/ST X-Git-Tag: div_pipeline~1890 X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=912ff8cecfdc1ed7db0d72449c97606a3a2c3597;p=soc.git add mirror copy of FU_Regs Dep Matrix, names changed, adapted to LD/ST wr -> ld dest -> ld rd -> st src1 -> st global search and replace. --- diff --git a/src/scoreboard/mem_dependence_cell.py b/src/scoreboard/mem_dependence_cell.py new file mode 100644 index 00000000..e84b9d17 --- /dev/null +++ b/src/scoreboard/mem_dependence_cell.py @@ -0,0 +1,118 @@ +from nmigen.compat.sim import run_simulation +from nmigen.cli import verilog, rtlil +from nmigen import Module, Signal, Elaboratable, Array, Cat, Repl +from nmutil.latch import SRLatch + + +class MemDepRow(Elaboratable): + """ implements 1st phase Memory Depencency cell + """ + def __init__(self, n_reg): + self.n_reg = n_reg + # inputs + self.ld_i = Signal(n_reg, reset_less=True) # Dest in (top) + self.st_i = Signal(n_reg, reset_less=True) # oper1 in (top) + self.issue_i = Signal(reset_less=True) # Issue in (top) + + self.st_pend_i = Signal(n_reg, reset_less=True) # Read pend in (top) + self.ld_pend_i = Signal(n_reg, reset_less=True) # Write pend in (top) + self.st_rsel_o = Signal(n_reg, reset_less=True) # Read pend out (bot) + self.ld_rsel_o = Signal(n_reg, reset_less=True) # Write pend out (bot) + + self.go_ld_i = Signal(reset_less=True) # Go Write in (left) + self.go_st_i = Signal(reset_less=True) # Go Read in (left) + self.go_die_i = Signal(reset_less=True) # Go Die in (left) + + # for Register File Select Lines (vertical) + self.ld_rsel_o = Signal(n_reg, reset_less=True) # dest reg sel (bot) + self.st_rsel_o = Signal(n_reg, reset_less=True) # src1 reg sel (bot) + + # for Function Unit "forward progress" (horizontal) + self.ld_fwd_o = Signal(n_reg, reset_less=True) # dest FU fw (right) + self.st_fwd_o = Signal(n_reg, reset_less=True) # src1 FU fw (right) + + def elaborate(self, platform): + m = Module() + m.submodules.ld_c = ld_c = SRLatch(sync=False, llen=self.n_reg) + m.submodules.st_c = st_c = SRLatch(sync=False, llen=self.n_reg) + + # connect go_rd / go_wr (dest->wr, src->rd) + ld_die = Signal(reset_less=True) + st_die = Signal(reset_less=True) + m.d.comb += ld_die.eq(self.go_ld_i | self.go_die_i) + m.d.comb += st_die.eq(self.go_st_i | self.go_die_i) + m.d.comb += ld_c.r.eq(Repl(ld_die, self.n_reg)) + m.d.comb += st_c.r.eq(Repl(st_die, self.n_reg)) + + # connect input reg bit (unary) + i_ext = Repl(self.issue_i, self.n_reg) + m.d.comb += ld_c.s.eq(i_ext & self.ld_i) + m.d.comb += st_c.s.eq(i_ext & self.st_i) + + # connect up hazard checks: read-after-write and write-after-read + m.d.comb += self.ld_fwd_o.eq(ld_c.q & self.st_pend_i) + m.d.comb += self.st_fwd_o.eq(st_c.q & self.ld_pend_i) + + # connect reg-sel outputs + st_ext = Repl(self.go_st_i, self.n_reg) + ld_ext = Repl(self.go_ld_i, self.n_reg) + m.d.comb += self.ld_rsel_o.eq(ld_c.qlq & ld_ext) + m.d.comb += self.st_rsel_o.eq(st_c.qlq & st_ext) + + # to be accumulated to indicate if register is in use (globally) + # after ORing, is fed back in to st_pend_i / ld_pend_i + m.d.comb += self.st_rsel_o.eq(st_c.qlq) + m.d.comb += self.ld_rsel_o.eq(ld_c.qlq) + + return m + + def __iter__(self): + yield self.ld_i + yield self.st_i + yield self.st_pend_i + yield self.ld_pend_i + yield self.issue_i + yield self.go_ld_i + yield self.go_st_i + yield self.go_die_i + yield self.ld_rsel_o + yield self.st_rsel_o + yield self.ld_fwd_o + yield self.st_fwd_o + + def ports(self): + return list(self) + + +def dcell_sim(dut): + yield dut.ld_i.eq(1) + yield dut.issue_i.eq(1) + yield + yield dut.issue_i.eq(0) + yield + yield dut.st_i.eq(1) + yield dut.issue_i.eq(1) + yield + yield + yield + yield dut.issue_i.eq(0) + yield + yield dut.go_st_i.eq(1) + yield + yield dut.go_st_i.eq(0) + yield + yield dut.go_ld_i.eq(1) + yield + yield dut.go_ld_i.eq(0) + yield + +def test_dcell(): + dut = MemDepRow(4) + vl = rtlil.convert(dut, ports=dut.ports()) + with open("test_mem_drow.il", "w") as f: + f.write(vl) + + run_simulation(dut, dcell_sim(dut), vcd_name='test_mem_dcell.vcd') + +if __name__ == '__main__': + test_dcell() diff --git a/src/scoreboard/mem_fu_matrix.py b/src/scoreboard/mem_fu_matrix.py new file mode 100644 index 00000000..83a6e219 --- /dev/null +++ b/src/scoreboard/mem_fu_matrix.py @@ -0,0 +1,218 @@ +from nmigen.compat.sim import run_simulation +from nmigen.cli import verilog, rtlil +from nmigen import Module, Signal, Elaboratable, Array, Cat + +from scoreboard.mem_dependence_cell import MemDepRow +from scoreboard.mem_fu_pending import MemFU_Pend +from scoreboard.mem_select import Mem_Rsv +from scoreboard.global_pending import GlobalPending + +""" + +""" + +class MemFUDepMatrix(Elaboratable): + """ implements 1st phase Memory-to-FU Dependency Matrix + """ + def __init__(self, n_fu_row, n_reg_col): + self.n_fu_row = n_fu_row # Y (FUs) ^v + self.n_reg_col = n_reg_col # X (Regs) <> + self.ld_i = Signal(n_reg_col, reset_less=True) # LD in (top) + self.st_i = Signal(n_reg_col, reset_less=True) # ST in (top) + + # Register "Global" vectors for determining RaW and WaR hazards + self.ld_pend_i = Signal(n_reg_col, reset_less=True) # ld pending (top) + self.st_pend_i = Signal(n_reg_col, reset_less=True) # st pending (top) + self.ld_rsel_o = Signal(n_reg_col, reset_less=True) # ld pending (bot) + self.st_rsel_o = Signal(n_reg_col, reset_less=True) # st pending (bot) + + self.issue_i = Signal(n_fu_row, reset_less=True) # Issue in (top) + self.go_ld_i = Signal(n_fu_row, reset_less=True) # Go LOAD in (left) + self.go_st_i = Signal(n_fu_row, reset_less=True) # Go STOR in (left) + self.go_die_i = Signal(n_fu_row, reset_less=True) # Go Die in (left) + + # for Register File Select Lines (horizontal), per-reg + self.ld_rsel_o = Signal(n_reg_col, reset_less=True) # ld reg (bot) + self.st_rsel_o = Signal(n_reg_col, reset_less=True) # st reg (bot) + + # for Function Unit "forward progress" (vertical), per-FU + self.ld_pend_o = Signal(n_fu_row, reset_less=True) # ld pending (right) + self.st_pend_o = Signal(n_fu_row, reset_less=True) # st pending (right) + + def elaborate(self, platform): + m = Module() + + # --- + # matrix of dependency cells + # --- + dm = Array(MemDepRow(self.n_reg_col) for r in range(self.n_fu_row)) + for fu in range(self.n_fu_row): + setattr(m.submodules, "dr_fu%d" % fu, dm[fu]) + + # --- + # array of Function Unit Pending vectors + # --- + fupend = Array(MemFU_Pend(self.n_reg_col) for f in range(self.n_fu_row)) + for fu in range(self.n_fu_row): + setattr(m.submodules, "fu_fu%d" % (fu), fupend[fu]) + + # --- + # array of Register Reservation vectors + # --- + regrsv = Array(Mem_Rsv(self.n_fu_row) for r in range(self.n_reg_col)) + for rn in range(self.n_reg_col): + setattr(m.submodules, "rr_r%d" % (rn), regrsv[rn]) + + # --- + # connect Function Unit vector + # --- + ld_pend = [] + st_pend = [] + for fu in range(self.n_fu_row): + dc = dm[fu] + fup = fupend[fu] + ld_fwd_o = [] + st_fwd_o = [] + for rn in range(self.n_reg_col): + # accumulate cell fwd outputs for dest/src1 + ld_fwd_o.append(dc.ld_fwd_o[rn]) + st_fwd_o.append(dc.st_fwd_o[rn]) + # connect cell fwd outputs to FU Vector in [Cat is gooood] + m.d.comb += [fup.ld_fwd_i.eq(Cat(*ld_fwd_o)), + fup.st_fwd_i.eq(Cat(*st_fwd_o)), + ] + # accumulate FU Vector outputs + ld_pend.append(fup.reg_ld_pend_o) + st_pend.append(fup.reg_st_pend_o) + + # ... and output them from this module (vertical, width=FUs) + m.d.comb += self.ld_pend_o.eq(Cat(*ld_pend)) + m.d.comb += self.st_pend_o.eq(Cat(*st_pend)) + + # --- + # connect Reg Selection vector + # --- + ld_rsel = [] + st_rsel = [] + for rn in range(self.n_reg_col): + rsv = regrsv[rn] + ld_rsel_o = [] + st_rsel_o = [] + for fu in range(self.n_fu_row): + dc = dm[fu] + # accumulate cell reg-select outputs dest/src1 + ld_rsel_o.append(dc.ld_rsel_o[rn]) + st_rsel_o.append(dc.st_rsel_o[rn]) + # connect cell reg-select outputs to Reg Vector In + m.d.comb += [rsv.ld_rsel_i.eq(Cat(*ld_rsel_o)), + rsv.st_rsel_i.eq(Cat(*st_rsel_o)), + ] + # accumulate Reg-Sel Vector outputs + ld_rsel.append(rsv.ld_rsel_o) + st_rsel.append(rsv.st_rsel_o) + + # ... and output them from this module (horizontal, width=REGs) + m.d.comb += self.ld_rsel_o.eq(Cat(*ld_rsel)) + m.d.comb += self.st_rsel_o.eq(Cat(*st_rsel)) + + # --- + # connect Dependency Matrix dest/src1/issue to module d/s/s/i + # --- + for fu in range(self.n_fu_row): + dc = dm[fu] + # wire up inputs from module to row cell inputs (Cat is gooood) + m.d.comb += [dc.ld_i.eq(self.ld_i), + dc.st_i.eq(self.st_i), + dc.st_pend_i.eq(self.st_pend_i), + dc.ld_pend_i.eq(self.ld_pend_i), + ] + + # accumulate rsel bits into read/write pending vectors. + st_pend_v = [] + ld_pend_v = [] + for fu in range(self.n_fu_row): + dc = dm[fu] + st_pend_v.append(dc.st_rsel_o) + ld_pend_v.append(dc.ld_rsel_o) + st_v = GlobalPending(self.n_reg_col, st_pend_v) + ld_v = GlobalPending(self.n_reg_col, ld_pend_v) + m.submodules.st_v = st_v + m.submodules.ld_v = ld_v + + m.d.comb += self.st_rsel_o.eq(st_v.g_pend_o) + m.d.comb += self.ld_rsel_o.eq(ld_v.g_pend_o) + + # --- + # connect Dep issue_i/go_st_i/go_ld_i to module issue_i/go_rd/go_wr + # --- + go_st_i = [] + go_ld_i = [] + go_die_i = [] + issue_i = [] + for fu in range(self.n_fu_row): + dc = dm[fu] + # accumulate cell fwd outputs for dest/src1 + go_st_i.append(dc.go_st_i) + go_ld_i.append(dc.go_ld_i) + go_die_i.append(dc.go_die_i) + issue_i.append(dc.issue_i) + # wire up inputs from module to row cell inputs (Cat is gooood) + m.d.comb += [Cat(*go_st_i).eq(self.go_st_i), + Cat(*go_ld_i).eq(self.go_ld_i), + Cat(*go_die_i).eq(self.go_die_i), + Cat(*issue_i).eq(self.issue_i), + ] + + return m + + def __iter__(self): + yield self.ld_i + yield self.st_i + yield self.issue_i + yield self.go_ld_i + yield self.go_st_i + yield self.go_die_i + yield self.ld_rsel_o + yield self.st_rsel_o + yield self.ld_pend_o + yield self.st_pend_o + yield self.ld_pend_i + yield self.st_pend_i + yield self.ld_rsel_o + yield self.st_rsel_o + + def ports(self): + return list(self) + +def d_matrix_sim(dut): + """ XXX TODO + """ + yield dut.ld_i.eq(1) + yield dut.issue_i.eq(1) + yield + yield dut.issue_i.eq(0) + yield + yield dut.st_i.eq(1) + yield dut.issue_i.eq(1) + yield + yield dut.issue_i.eq(0) + yield + yield dut.go_st_i.eq(1) + yield + yield dut.go_st_i.eq(0) + yield + yield dut.go_ld_i.eq(1) + yield + yield dut.go_ld_i.eq(0) + yield + +def test_d_matrix(): + dut = MemFUDepMatrix(n_fu_row=3, n_reg_col=3) + vl = rtlil.convert(dut, ports=dut.ports()) + with open("test_fu_mem_matrix.il", "w") as f: + f.write(vl) + + run_simulation(dut, d_matrix_sim(dut), vcd_name='test_fu_mem_matrix.vcd') + +if __name__ == '__main__': + test_d_matrix() diff --git a/src/scoreboard/mem_fu_pending.py b/src/scoreboard/mem_fu_pending.py new file mode 100644 index 00000000..951f7ac1 --- /dev/null +++ b/src/scoreboard/mem_fu_pending.py @@ -0,0 +1,22 @@ +from nmigen import Elaboratable, Module, Signal, Cat + + +class MemFU_Pend(Elaboratable): + """ these are allocated per-FU (horizontally), + and are of length reg_count + """ + def __init__(self, reg_count): + self.reg_count = reg_count + self.ld_fwd_i = Signal(reg_count, reset_less=True) + self.st_fwd_i = Signal(reg_count, reset_less=True) + + self.reg_ld_pend_o = Signal(reset_less=True) + self.reg_st_pend_o = Signal(reset_less=True) + + def elaborate(self, platform): + m = Module() + m.d.comb += self.reg_ld_pend_o.eq(self.ld_fwd_i.bool()) + m.d.comb += self.reg_st_pend_o.eq(self.st_fwd_i.bool()) + + return m + diff --git a/src/scoreboard/mem_select.py b/src/scoreboard/mem_select.py new file mode 100644 index 00000000..627d7d10 --- /dev/null +++ b/src/scoreboard/mem_select.py @@ -0,0 +1,20 @@ +from nmigen import Elaboratable, Module, Signal + + +class Mem_Rsv(Elaboratable): + """ these are allocated per-Register (vertically), + and are each of length fu_count + """ + def __init__(self, fu_count): + self.fu_count = fu_count + self.ld_rsel_i = Signal(fu_count, reset_less=True) + self.st_rsel_i = Signal(fu_count, reset_less=True) + self.ld_rsel_o = Signal(reset_less=True) + self.st_rsel_o = Signal(reset_less=True) + + def elaborate(self, platform): + m = Module() + m.d.comb += self.ld_rsel_o.eq(self.ld_rsel_i.bool()) + m.d.comb += self.st_rsel_o.eq(self.st_rsel_i.bool()) + return m +