107 lines
3.5 KiB
Verilog
107 lines
3.5 KiB
Verilog
`include "VX_define.vh"
|
|
|
|
module VX_gpu_unit #(
|
|
parameter CORE_ID = 0
|
|
) (
|
|
`SCOPE_IO_VX_gpu_unit
|
|
|
|
input wire clk,
|
|
input wire reset,
|
|
|
|
// Inputs
|
|
VX_gpu_req_if gpu_req_if,
|
|
|
|
// Outputs
|
|
VX_warp_ctl_if warp_ctl_if,
|
|
VX_commit_if gpu_commit_if
|
|
);
|
|
|
|
`UNUSED_PARAM (CORE_ID)
|
|
`UNUSED_VAR (clk)
|
|
`UNUSED_VAR (reset)
|
|
|
|
gpu_tmc_t tmc;
|
|
gpu_wspawn_t wspawn;
|
|
gpu_barrier_t barrier;
|
|
gpu_split_t split;
|
|
|
|
wire is_wspawn = (gpu_req_if.op_type == `INST_GPU_WSPAWN);
|
|
wire is_tmc = (gpu_req_if.op_type == `INST_GPU_TMC);
|
|
wire is_split = (gpu_req_if.op_type == `INST_GPU_SPLIT);
|
|
wire is_bar = (gpu_req_if.op_type == `INST_GPU_BAR);
|
|
wire is_pred = (gpu_req_if.op_type == `INST_GPU_PRED);
|
|
|
|
wire [31:0] rs1_data = gpu_req_if.rs1_data[gpu_req_if.tid];
|
|
|
|
wire [`NUM_THREADS-1:0] taken_tmask;
|
|
wire [`NUM_THREADS-1:0] not_taken_tmask;
|
|
|
|
for (genvar i = 0; i < `NUM_THREADS; i++) begin
|
|
wire taken = gpu_req_if.rs1_data[i][0];
|
|
assign taken_tmask[i] = gpu_req_if.tmask[i] & taken;
|
|
assign not_taken_tmask[i] = gpu_req_if.tmask[i] & ~taken;
|
|
end
|
|
|
|
// tmc
|
|
|
|
wire [`NUM_THREADS-1:0] pred_mask = (taken_tmask != 0) ? taken_tmask : gpu_req_if.tmask;
|
|
|
|
assign tmc.valid = is_tmc || is_pred;
|
|
assign tmc.tmask = is_pred ? pred_mask : rs1_data[`NUM_THREADS-1:0];
|
|
|
|
// wspawn
|
|
|
|
wire [31:0] wspawn_pc = gpu_req_if.rs2_data;
|
|
wire [`NUM_WARPS-1:0] wspawn_wmask;
|
|
for (genvar i = 0; i < `NUM_WARPS; i++) begin
|
|
assign wspawn_wmask[i] = (i < rs1_data);
|
|
end
|
|
assign wspawn.valid = is_wspawn;
|
|
assign wspawn.wmask = wspawn_wmask;
|
|
assign wspawn.pc = wspawn_pc;
|
|
|
|
// split
|
|
|
|
assign split.valid = is_split;
|
|
assign split.diverged = (| taken_tmask) && (| not_taken_tmask);
|
|
assign split.then_tmask = taken_tmask;
|
|
assign split.else_tmask = not_taken_tmask;
|
|
assign split.pc = gpu_req_if.next_PC;
|
|
|
|
// barrier
|
|
|
|
assign barrier.valid = is_bar;
|
|
assign barrier.id = rs1_data[`NB_BITS-1:0];
|
|
assign barrier.size_m1 = (`NW_BITS)'(gpu_req_if.rs2_data - 1);
|
|
|
|
// output
|
|
|
|
wire stall = ~gpu_commit_if.ready && gpu_commit_if.valid;
|
|
|
|
VX_pipe_register #(
|
|
.DATAW (1 + `NW_BITS + `NUM_THREADS + 32 + `NR_BITS + 1 + `GPU_TMC_BITS + `GPU_WSPAWN_BITS + `GPU_SPLIT_BITS + `GPU_BARRIER_BITS),
|
|
.RESETW (1)
|
|
) pipe_reg (
|
|
.clk (clk),
|
|
.reset (reset),
|
|
.enable (!stall),
|
|
.data_in ({gpu_req_if.valid, gpu_req_if.wid, gpu_req_if.tmask, gpu_req_if.PC, gpu_req_if.rd, gpu_req_if.wb, tmc, wspawn, split, barrier}),
|
|
.data_out ({gpu_commit_if.valid, gpu_commit_if.wid, gpu_commit_if.tmask, gpu_commit_if.PC, gpu_commit_if.rd, gpu_commit_if.wb, warp_ctl_if.tmc, warp_ctl_if.wspawn, warp_ctl_if.split, warp_ctl_if.barrier})
|
|
);
|
|
|
|
assign gpu_commit_if.eop = 1'b1;
|
|
|
|
assign warp_ctl_if.valid = gpu_commit_if.valid && gpu_commit_if.ready;
|
|
assign warp_ctl_if.wid = gpu_commit_if.wid;
|
|
|
|
// can accept new request?
|
|
assign gpu_req_if.ready = ~stall;
|
|
|
|
`SCOPE_ASSIGN (gpu_rsp_valid, warp_ctl_if.valid);
|
|
`SCOPE_ASSIGN (gpu_rsp_wid, warp_ctl_if.wid);
|
|
`SCOPE_ASSIGN (gpu_rsp_tmc, warp_ctl_if.tmc);
|
|
`SCOPE_ASSIGN (gpu_rsp_wspawn, warp_ctl_if.wspawn);
|
|
`SCOPE_ASSIGN (gpu_rsp_split, warp_ctl_if.split);
|
|
`SCOPE_ASSIGN (gpu_rsp_barrier, warp_ctl_if.barrier);
|
|
|
|
endmodule |