github-mirrors/cve2
1
0
Fork 0
mirror of https://github.com/openhwgroup/cve2.git synced 2025-06-28 01:29:10 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

adding back obi2ahbm example

Browse source
This commit is contained in:
Davide Schiavone 2024-06-06 11:38:53 +02:00
parent 42f76f433c
commit cd43ee5dbf
2 changed files with 283 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

BIN
examples/obi2ahb/OBI2AHBM_Adapter Design Spec.pdf Executable file
View file

Binary file not shown.

283
examples/obi2ahb/obi2ahbm_adapter.sv Executable file
View file

@ -0,0 +1,283 @@
// Copyright 2023 Intrinsix Corp.
// SPDX-License-Identifier: Apache-2.0 WITH SHL-2.0
// ============================================================================
//
// Description : OBI to AHB-Lite Master Adaptor
// This adapter accepts OBI v1.5.0 transfers and translates them into
// AHB-lite (ARM IHI 0033C) transfers.
module obi2ahbm_adapter (
// Clock and reset
input hclk_i, // (I) AHB clock
input hresetn_i, // (I) AHB reset, active LOW
// AHB master interface
output logic [31:0] haddr_o, // (O) 32-bit AHB system address bus
output logic [2:0] hburst_o, // (O) Burst type
output logic hmastlock_o, // (O) Sequence lock
output logic [3:0] hprot_o, // (O) Protection control
output logic [2:0] hsize_o, // (O) Transfer size
output logic [1:0] htrans_o, // (O) Transfer type
output logic [31:0] hwdata_o, // (O) 32-bit AHB write data bus
output logic hwrite_o, // (O) Transfer direction
input logic [31:0] hrdata_i, // (I) 32-bit AHB read data bus
input logic hready_i, // (I) Status of transfer
input logic hresp_i, // (I) Transfer response
// Data interface from core
input logic data_req_i, // (I) Request ready
output logic data_gnt_o, // (O) The other side accepted the request
output logic data_rvalid_o, // (O) Read data valid when high
input logic data_we_i, // (I) Write enable (active HIGH)
input logic [3:0] data_be_i, // (I) Byte enable
input logic [31:0] data_addr_i, // (I) Address
input logic [31:0] data_wdata_i, // (I) Write data
output logic [31:0] data_rdata_o, // (O) Read data
output logic data_err_o, // (O) Error
input logic pending_dbus_xfer_i, // (I) Asserted if data bus is busy from other transactions
// Miscellaneous
input logic priv_mode_i // (I) Privilege mode (from core. 1=machine mode, 0=user mode)
);
// ********** //
// Parameters //
// ********** //
parameter HPROT_NONCACHEABLE = 1'b0;
parameter HPROT_NONBUFFERABLE = 1'b0;
parameter HPROT_DATAACCESS = 1'b1;
parameter MACHINE_MODE = 1'b1;
parameter HBURST_SINGLE = 3'b000;
parameter TIE_LO = 1'b0;
parameter AHB_FSM_WAIT = 2'b00;
parameter AHB_FSM_DATA = 2'b10;
// ********** //
// Wires/Regs //
// ********** //
logic [1:0] ahb_fsm_reg;
logic [1:0] ahb_fsm_reg_nxt;
logic [31:0] haddr_m_reg;
logic [3:0] hprot_m_reg;
logic [2:0] hsize_m_reg;
logic hwrite_m_reg;
logic data_err_o_nxt;
logic [31:0] data_rdata_o_reg;
logic [31:0] hwdata_m_nxt;
logic prev_data_gnt_o;
// ********************** //
// Continuous assignments //
// ********************** //
// These signals are unused, so OK that they are tied to a constant
// lint_checking TIELOG off
// Only single bursts are supported
assign hburst_o = HBURST_SINGLE;
// hmastlock_o is not used, so it is tied low
assign hmastlock_o = TIE_LO;
// lint_checking TIELOG on
// A grant should only happen on a request, and if the AHB side is ready
// to respond. Don't initiate an AHB xfer if the data bus is busy from
// other pending xfers
assign data_gnt_o = hready_i && data_req_i && !pending_dbus_xfer_i;
// The valid signal should only assert when transitioning to the DATA state
assign data_rvalid_o = ahb_fsm_reg == AHB_FSM_DATA ? hready_i : 1'b0;
// ************* //
// Clocked Logic //
// ************* //
always @ (posedge hclk_i or negedge hresetn_i) begin
if (!hresetn_i) begin
data_err_o <= 1'b0;
data_rdata_o_reg <= 32'h00000000;
haddr_m_reg <= 32'h00000000;
hprot_m_reg <= {HPROT_NONCACHEABLE, HPROT_NONBUFFERABLE, MACHINE_MODE, HPROT_DATAACCESS};
hsize_m_reg <= 3'b000;
// Write data needs to come one cycle after per AHB protocol
hwdata_o <= 32'h00000000;
hwrite_m_reg <= 1'b0;
ahb_fsm_reg <= AHB_FSM_WAIT;
prev_data_gnt_o <= 1'b0;
end
else begin
data_err_o <= data_err_o_nxt;
data_rdata_o_reg <= data_rdata_o;
haddr_m_reg <= haddr_o;
hprot_m_reg <= hprot_o;
hsize_m_reg <= hsize_o;
hwdata_o <= hwdata_m_nxt;
hwrite_m_reg <= hwrite_o;
ahb_fsm_reg <= ahb_fsm_reg_nxt;
if (!data_rvalid_o) begin
prev_data_gnt_o <= data_gnt_o | prev_data_gnt_o;
end else begin
prev_data_gnt_o <= data_gnt_o;
end
end
end
// ******************* //
// Combinational Logic //
// ******************* //
// Signal that the current request errored out and the read is invalid if
// an AHB error response is received
assign data_err_o_nxt = hresp_i;
//AHB Logic
always_comb begin
hprot_o = hprot_m_reg;
haddr_o = haddr_m_reg;
hwrite_o = hwrite_m_reg;
// lint_checking TIELOG off
// If htrans_m is not driven to non-seq while data_gnt_o is set
// then it should be idle and set to (2'b00)
htrans_o = 2'b00;
// lint_checking TIELOG on
hwdata_m_nxt = hwdata_o;
hsize_o = hsize_m_reg;
unique case (ahb_fsm_reg)
AHB_FSM_WAIT: begin
// Idles until data_gnt_o is received
// Once data_gnt_o is received, this is the address phase of AHB
// and all ahb signals are fed through from the core
if (data_gnt_o) begin
// lint_checking TIELOG off
// Non-bufferable, non-cacheable data accesses are supported. The
// privilege bit, hprot_o[1], is driven by the privilege mode
// that the core is currently in (1 for machine, 0 for user)
hprot_o = {HPROT_NONCACHEABLE, HPROT_NONBUFFERABLE, priv_mode_i, HPROT_DATAACCESS};
// lint_checking TIELOG on
// The address, transfer request, and write enable can be fed
// through from the core to the corresponding AHB signals. For
// htrans_o, if data_gnt_o is given, the ahb is in address phase,
// and the transaction is non-sequential (1'b10).
haddr_o = data_addr_i;
hwrite_o = data_we_i;
// lint_checking TIELOG off
htrans_o = 2'b10;
// lint_checking TIELOG on
// Write data needs to be in data phase (After data_gnt_o is de-asserted)
// So the data fromthe core is registered to save it for completing the transaction
hwdata_m_nxt = data_wdata_i;
// lint_checking TIELOG off
// Only word (32-bit), half-word (16-bit), or byte (8-bit)
// aligned transfers are supported. Invalid responses default to
hsize_o[2] = TIE_LO;
// lint_checking TIELOG on
hsize_o[1] = &data_be_i;
hsize_o[0] = (data_be_i[3] & data_be_i[2]) ^ (data_be_i[1] & data_be_i[0]);
end
end
AHB_FSM_DATA: begin
// DATA phase
// read data is fed through from the core
data_rdata_o = hrdata_i;
// If data_gnt_o is also given during the data phase, then it is also
// the subsequent address phase, so the necessary signals get
// passed through from the core
if (data_gnt_o) begin
// lint_checking TIELOG off
// Non-bufferable, non-cacheable data accesses are supported. The
// privilege bit, hprot_o[1], is driven by the privilege mode
// that the core is currently in (1 for machine, 0 for user)
hprot_o = {HPROT_NONCACHEABLE, HPROT_NONBUFFERABLE, priv_mode_i, HPROT_DATAACCESS};
// lint_checking TIELOG on
// The address, transfer request, and write enable can be fed
// through from the core to the corresponding AHB signals. For
// htrans_o, if data_gnt_o is given, the ahb is in address phase,
// and the transaction is non-sequential (1'b10).
haddr_o = data_addr_i;
hwrite_o = data_we_i;
// lint_checking TIELOG off
htrans_o = 2'b10;
// lint_checking TIELOG on
// Write data needs to be in data phase (After data_gnt_o is de-asserted)
// So the data fromthe core is registered to save it for completing the transaction
hwdata_m_nxt = data_wdata_i;
// lint_checking TIELOG off
// Only word (32-bit), half-word (16-bit), or byte (8-bit)
// aligned transfers are supported. Invalid responses default to
hsize_o[2] = TIE_LO;
// lint_checking TIELOG on
hsize_o[1] = &data_be_i;
hsize_o[0] = (data_be_i[3] & data_be_i[2]) ^ (data_be_i[1] & data_be_i[0]);
end
end
default: begin
hprot_o = {HPROT_NONCACHEABLE, HPROT_NONBUFFERABLE, priv_mode_i, HPROT_DATAACCESS};
haddr_o = data_addr_i;
hwrite_o = data_we_i;
htrans_o = 2'b00;
hsize_o = {TIE_LO,TIE_LO,TIE_LO};
end
endcase
end
//FSM
always_comb begin
ahb_fsm_reg_nxt = ahb_fsm_reg;
unique case (ahb_fsm_reg)
// IDLE/ADDRESS phase
// Waits until "ADDRESS Phase" occurs during data_gnt_o
AHB_FSM_WAIT: begin
if (data_gnt_o) begin
ahb_fsm_reg_nxt = AHB_FSM_DATA;
end
end
// DATA/ADDRESS phase
// DATA phase for alread started transaction.
// If data_gnt_o it is also the ADDRESS phase for the next set of data so it stays
// in the data phase on the next cycle
AHB_FSM_DATA: begin
if (data_rvalid_o) begin
if (data_gnt_o) begin
ahb_fsm_reg_nxt = AHB_FSM_DATA;
end else begin
ahb_fsm_reg_nxt = AHB_FSM_WAIT;
end
end
end
default: begin
ahb_fsm_reg_nxt = ahb_fsm_reg;
end
endcase
end
// lint_checking TRNMBT on
// lint_checking HASUPC on
endmodule : obi2ahbm_adapter