111 lines
4.2 KiB
Systemverilog
111 lines
4.2 KiB
Systemverilog
// Copyright © 2019-2023
|
|
//
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
// you may not use this file except in compliance with the License.
|
|
// You may obtain a copy of the License at
|
|
// http://www.apache.org/licenses/LICENSE-2.0
|
|
//
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
// See the License for the specific language governing permissions and
|
|
// limitations under the License.
|
|
|
|
`include "VX_define.vh"
|
|
`include "VX_trace.vh"
|
|
|
|
module VX_ibuffer import VX_gpu_pkg::*; #(
|
|
parameter CORE_ID = 0,
|
|
parameter DOMAIN = WU_DOMAIN_SCALAR
|
|
) (
|
|
input wire clk,
|
|
input wire reset,
|
|
|
|
// inputs
|
|
VX_decode_if.slave decode_if,
|
|
|
|
// outputs
|
|
VX_ibuffer_if.master ibuffer_if [`ISSUE_WIDTH]
|
|
);
|
|
`UNUSED_PARAM (CORE_ID)
|
|
localparam ISW_WIDTH = `LOG2UP(`ISSUE_WIDTH);
|
|
localparam DATAW = `UUID_WIDTH + ISSUE_WIS_W + `NUM_THREADS + `XLEN + 1 + `EX_BITS + `INST_OP_BITS + `INST_MOD_BITS + 1 + 1 + `XLEN + (`NR_BITS * 4);
|
|
|
|
wire [`ISSUE_WIDTH-1:0] ibuf_ready_in;
|
|
|
|
wire [ISW_WIDTH-1:0] decode_isw = wid_to_isw(decode_if.data.wid);
|
|
wire [ISSUE_WIS_W-1:0] decode_wis = wid_to_wis(decode_if.data.wid);
|
|
wire decode_lane_in_domain = (DOMAIN == WU_DOMAIN_TENSOR) ? `IS_TENSOR_WARP(decode_if.data.wid) : `IS_SCALAR_WARP(decode_if.data.wid);
|
|
|
|
assign decode_if.ready = decode_lane_in_domain && ibuf_ready_in[decode_isw];
|
|
|
|
`ifdef SIMULATION
|
|
`ifdef DBG_TRACE_CORE_PIPELINE_VCS
|
|
wire decode_probe_pc = (decode_if.data.PC >= 32'h80000240) && (decode_if.data.PC <= 32'h80000260);
|
|
always @(posedge clk) begin
|
|
if (!reset && (CORE_ID == 0) && ($time > `TRACE_STARTTIME) && decode_if.valid && decode_probe_pc) begin
|
|
`TRACE(2, ("%d: core%0d-ibuffer-probe: wid=%0d, PC=0x%0h, valid=%b, ready=%b, isw=%0d, ibuf_ready=%b, tmask=%b, ex=0x%0h, op=0x%0h (#%0d)\n",
|
|
$time, CORE_ID, decode_if.data.wid, decode_if.data.PC, decode_if.valid, decode_if.ready,
|
|
decode_isw, ibuf_ready_in, decode_if.data.tmask, decode_if.data.ex_type, decode_if.data.op_type,
|
|
decode_if.data.uuid));
|
|
end
|
|
end
|
|
`endif
|
|
`endif
|
|
|
|
VX_ibuffer_if uop_sequencer_if [`ISSUE_WIDTH]();
|
|
|
|
for (genvar i = 0; i < `ISSUE_WIDTH; ++i) begin
|
|
localparam LANE_IN_DOMAIN = (DOMAIN == WU_DOMAIN_TENSOR) ? `IS_TENSOR_WARP(i) : `IS_SCALAR_WARP(i);
|
|
VX_elastic_buffer #(
|
|
.DATAW (DATAW),
|
|
.SIZE (`IBUF_SIZE),
|
|
.OUT_REG (1)
|
|
) instr_buf (
|
|
.clk (clk),
|
|
.reset (reset),
|
|
.valid_in (decode_if.valid && decode_isw == i && decode_lane_in_domain && LANE_IN_DOMAIN),
|
|
.ready_in (ibuf_ready_in[i]),
|
|
.data_in ({
|
|
decode_if.data.uuid,
|
|
decode_wis,
|
|
decode_if.data.tmask,
|
|
decode_if.data.ex_type,
|
|
decode_if.data.op_type,
|
|
decode_if.data.op_mod,
|
|
decode_if.data.wb,
|
|
decode_if.data.use_PC,
|
|
decode_if.data.use_imm,
|
|
decode_if.data.PC,
|
|
decode_if.data.imm,
|
|
decode_if.data.rd,
|
|
decode_if.data.rs1,
|
|
decode_if.data.rs2,
|
|
decode_if.data.rs3}),
|
|
|
|
.data_out (uop_sequencer_if[i].data),
|
|
.valid_out (uop_sequencer_if[i].valid),
|
|
.ready_out (uop_sequencer_if[i].ready)
|
|
);
|
|
|
|
`ifndef L1_ENABLE
|
|
assign decode_if.ibuf_pop[i] = uop_sequencer_if[i].valid && uop_sequencer_if[i].ready;
|
|
`endif
|
|
|
|
// tensor-core operation is controlled by a single macro-instruction at
|
|
// the ISA; internally, the uop_sequencer blitzs micro-ops (counterpart
|
|
// to Volta SASS set/step instructions) into the ibuffer upon encountering
|
|
// this macro-instruction. this becomes a pass-through for non-tensorcore
|
|
// instructions.
|
|
VX_uop_sequencer uop_sequencer (
|
|
.clk(clk),
|
|
.reset(reset),
|
|
|
|
.uop_sequencer_if(uop_sequencer_if[i]),
|
|
.ibuffer_if(ibuffer_if[i])
|
|
);
|
|
|
|
end
|
|
|
|
endmodule
|