From: Luke Kenneth Casson Leighton Date: Sat, 16 May 2020 18:17:13 +0000 (+0100) Subject: code-munge on CR pipeline X-Git-Tag: div_pipeline~1128 X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=3e5975ed2814bca9bcd5c07ba6c4d800e38b3f4d;p=soc.git code-munge on CR pipeline --- diff --git a/src/soc/cr/main_stage.py b/src/soc/cr/main_stage.py index 1552c4b1..97120123 100644 --- a/src/soc/cr/main_stage.py +++ b/src/soc/cr/main_stage.py @@ -5,6 +5,14 @@ # This module however should not gate the carry or overflow, that's up # to the output stage +# NOTE: we really should be doing the field decoding which +# selectswhich bits of CR are to be read / written, back in the +# decoder / insn-isue, have both self.i.cr and self.o.cr +# be broken down into 4-bit-wide "registers", with their +# own "Register File" (indexed by bt, ba and bb), +# exactly how INT regs are done (by RA, RB, RS and RT) +# however we are pushed for time so do it as *one* register. + from nmigen import (Module, Signal, Cat, Repl, Mux, Const, Array) from nmutil.pipemodbase import PipeModBase from soc.cr.pipe_data import CRInputData, CROutputData @@ -40,8 +48,9 @@ class CRMainStage(PipeModBase): xl_fields = self.fields.instrs['XL'] xfx_fields = self.fields.instrs['XFX'] - cr_output = Signal.like(self.i.cr) - comb += cr_output.eq(self.i.cr) + # default: cr_o remains same as cr input unless modified, below + cr_o = Signal.like(self.i.cr) + comb += cr_o.eq(self.i.cr) # Generate array for cr input so bits can be selected cr_arr = Array([Signal(name=f"cr_arr_{i}") for i in range(32)]) @@ -52,20 +61,9 @@ class CRMainStage(PipeModBase): # selected by a signal cr_out_arr = Array([Signal(name=f"cr_out_{i}") for i in range(32)]) for i in range(32): - comb += cr_output[31-i].eq(cr_out_arr[i]) + comb += cr_o[31-i].eq(cr_out_arr[i]) comb += cr_out_arr[i].eq(cr_arr[i]) - # crand/cror and friends get decoded to the same opcode, but - # one of the fields inside the instruction is a 4 bit lookup - # table. This lookup table gets indexed by bits a and b from - # the CR to determine what the resulting bit should be. - - # Grab the lookup table for cr_op type instructions - lut = Signal(4, reset_less=True) - # There's no field, just have to grab it directly from the insn - comb += lut.eq(self.i.ctx.op.insn[6:10]) - - # Ugh. mtocrf and mtcrf have one random bit differentiating # them. This bit is not in any particular field, so this # extracts that bit from the instruction @@ -73,6 +71,7 @@ class CRMainStage(PipeModBase): comb += move_one.eq(self.i.ctx.op.insn[20]) with m.Switch(op.insn_type): + ##### mcrf ##### with m.Case(InternalOp.OP_MCRF): # MCRF copies the 4 bits of crA to crB (for instance # copying cr2 to cr1) @@ -86,13 +85,15 @@ class CRMainStage(PipeModBase): for i in range(4): comb += cr_out_arr[bf*4 + i].eq(cr_arr[bfa*4 + i]) + + ##### crand, cror, crnor etc. ##### with m.Case(InternalOp.OP_CROP): # Get the bit selector fields from the instruction bt = Signal(xl_fields['BT'][0:-1].shape()) - comb += bt.eq(xl_fields['BT'][0:-1]) ba = Signal(xl_fields['BA'][0:-1].shape()) - comb += ba.eq(xl_fields['BA'][0:-1]) bb = Signal(xl_fields['BB'][0:-1].shape()) + comb += bt.eq(xl_fields['BT'][0:-1]) + comb += ba.eq(xl_fields['BA'][0:-1]) comb += bb.eq(xl_fields['BB'][0:-1]) # Extract the two input bits from the CR @@ -101,31 +102,40 @@ class CRMainStage(PipeModBase): comb += bit_a.eq(cr_arr[ba]) comb += bit_b.eq(cr_arr[bb]) - bit_out = Signal(reset_less=True) + # crand/cror and friends get decoded to the same opcode, but + # one of the fields inside the instruction is a 4 bit lookup + # table. This lookup table gets indexed by bits a and b from + # the CR to determine what the resulting bit should be. + + # Grab the lookup table for cr_op type instructions + lut = Signal(4, reset_less=True) + # There's no field, just have to grab it directly from the insn + comb += lut.eq(self.i.ctx.op.insn[6:10]) # Use the two input bits to look up the result in the # lookup table + bit_out = Signal(reset_less=True) comb += bit_out.eq(Mux(bit_b, Mux(bit_a, lut[3], lut[1]), Mux(bit_a, lut[2], lut[0]))) # Set the output to the result above comb += cr_out_arr[bt].eq(bit_out) + ##### mtcrf ##### with m.Case(InternalOp.OP_MTCRF): fxm = Signal(xfx_fields['FXM'][0:-1].shape()) comb += fxm.eq(xfx_fields['FXM'][0:-1]) - # mtcrf + # replicate every fxm field in the insn to 4-bit, as a mask + fxl = [Repl(fxm[i], 4) for i in range(8)] mask = Signal(32, reset_less=True) + comb += mask.eq(Cat(*fxl)) - for i in range(8): - comb += mask[i*4:(i+1)*4].eq(Repl(fxm[i], 4)) - - comb += cr_output.eq((self.i.a[0:32] & mask) | - (self.i.cr & ~mask)) - + # put input (RA) - mask-selected - into output CR, leave + # rest of CR alone. + comb += cr_o.eq((self.i.a[0:32] & mask) | (self.i.cr & ~mask)) - comb += self.o.cr.eq(cr_output) + comb += self.o.cr.eq(cr_o) comb += self.o.ctx.eq(self.i.ctx) return m