当前位置:网站首页>AXI_ Round_ Robin_ Arbiter design - aw and W channels
AXI_ Round_ Robin_ Arbiter design - aw and W channels
2022-06-23 14:51:00 【Starry and】
Catalog
1. State machine design
Based on the state machine , The idea is to receive a package AW Channel data , Record its len, And then according to len and awid Receive corresponding AXI Master Of W Channel data , Guarantee AW Channels and W The channel sequence is consistent .

How does the weight change ? What I think here is Every time AW and W The channel completes one-time data collection , The weight changes once . Make sure AW Acceptance and W Reception must come from the same AXI Master
1.1. WAIT_AW
Waiting for a AW Channel and AXI Slave The handshake is complete , You can go to the next state
● according to req Output grant
Here req In fact, it's all about AXI Master Of awvalid Composed of many bit data , and grant Is the result of arbitration . Pay attention to WAIT_AW The state does not change
● State transition conditions :AW The channel handshake is complete
Note that the state transition condition of the original algorithm is req When something changes, it shifts , But here you need something Master Of awvalid Is high and awready Handshake only when the signal is high , Here's the picture

prio In fact, it is the weight . although req Changes have taken place in the middle , But the condition of state transition is the same as Master0 The condition is not transferred until the handshake is completed .
Be careful AW A channel is simply a one-time transmission of multiple control messages , No burst,awready&awvalid One effective clap can complete the handshake .
into WAIT_W Post state awready It should not be raised , Must wait from WAIT_W go back to WAIT_AW To pull it higher .
1.2. WAIT_W
Then enter WAIT_W state , Before WAIT_AW With which Master Finished shaking hands , This is what we have to wait for Master Of W The channel handshake is completed WAIT_W state .
● grant keep , Accept W Channel data
there grant Must be with WAIT_AW When the handshake is completed grant Agreement , At the same time, always accept W Data from the channel
● State transition conditions :W The channel handshake is complete
because Master Is constant , So just look at that Master Of wlast that will do , This signal represents the last data .

After the handshake is done , Return to WAIT_AW state . Note that this time prio The weight should also be based on grant Make corrections , meanwhile grant It can't be maintained any longer , into WAIT_AW After that, according to awvalid_mst Make combinational logic output
in summary , The state machine becomes

1.3. Code
module axi_round_robin_arbiter_aw_w #(
parameter AWID_WIDTH = 8,
parameter AWADDR_WIDTH = 16,
parameter WDATA_WIDTH = 16,
parameter MAX_PRIO_INITIAL_INDEX = 0
)(
input aclk,
input arstn,
//AW CHANNEL
input [(AWID_WIDTH+2)*4-1:0] awid_mst,
input [AWADDR_WIDTH*4-1:0] awaddr_mst,
input [8*4-1:0] awlen_mst,
input [3*4-1:0] awsize_mst,
input [2*4-1:0] awburst_mst,
input [1*4-1:0] awvalid_mst,
output [1*4-1:0] awready_mst,
output [(AWID_WIDTH+2)-1:0] awid,
output [AWADDR_WIDTH-1:0] awaddr,
output [8-1:0] awlen,
output [3-1:0] awsize,
output [2-1:0] awburst,
output awvalid,
input awready,
//W CHANNEL
input [WDATA_WIDTH*4-1:0] wdata_mst,
input [1*4-1:0] wlast_mst,
input [1*4-1:0] wvalid_mst,
output [1*4-1:0] wready_mst,
output [WDATA_WIDTH-1:0] wdata,
output [1-1:0] wlast,
output [1-1:0] wvalid,
input [1-1:0] wready
);
localparam WAIT_AW = 1'b0;
localparam WAIT_W = 1'b1;
reg [4-1:0] cur_state;
reg [4-1:0] nxt_state;
reg [4-1:0] grant;
reg [4-1:0] grant_tmp;
reg [4-1:0] prio;
reg wvalid_mst_cs;
reg wlast_mst_cs;
wire [4-1:0] inital_state;
wire [4:0] res1;
wire [4:0] res2;
assign inital_state = {
{
(4-MAX_PRIO_INITIAL_INDEX-1){
1'b0}},1'b1,{
MAX_PRIO_INITIAL_INDEX{
1'b0}}};
[email protected](posedge aclk or negedge arstn) begin
if(!arstn)
cur_state <= WAIT_AW;
else
cur_state <= nxt_state;
end
[email protected](*) begin
case(cur_state)
WAIT_AW:
if((|awvalid_mst) && awready)
nxt_state = WAIT_W;
else
nxt_state = WAIT_AW;
WAIT_W:
if(wvalid_mst_cs && wlast_mst_cs && wready)
nxt_state = WAIT_AW;
else
nxt_state = WAIT_W;
default: nxt_state = WAIT_AW;
endcase
end
[email protected](posedge aclk or negedge arstn) begin
if(!arstn)
prio <= inital_state;
else if(cur_state == WAIT_W && wvalid_mst_cs && wlast_mst_cs && wready)
prio <= (grant[3])? 4'b1:(grant << 1);
end
[email protected](*) begin
if(cur_state == WAIT_AW)
wlast_mst_cs = 1'b0;
else begin
case(grant)
4'b0001:
wlast_mst_cs = wlast_mst[0];
4'b0010:
wlast_mst_cs = wlast_mst[1];
4'b0100:
wlast_mst_cs = wlast_mst[2];
4'b1000:
wlast_mst_cs = wlast_mst[3];
default:
wlast_mst_cs = 1'b0;
endcase
end
end
[email protected](*) begin
if(cur_state == WAIT_W)
wvalid_mst_cs = 1'b0;
else begin
case(grant)
4'b0001:
wvalid_mst_cs = wvalid_mst[0];
4'b0010:
wvalid_mst_cs = wvalid_mst[1];
4'b0100:
wvalid_mst_cs = wvalid_mst[2];
4'b1000:
wvalid_mst_cs = wvalid_mst[3];
default:
wvalid_mst_cs = 1'b0;
endcase
end
end
[email protected](posedge aclk or negedge arstn) begin
if(!arstn)
grant_tmp <= 'd0;
else if(cur_state == WAIT_AW && (|awvalid_mst) && awready)
grant_tmp <= grant;
end
assign res1 = {
1'b0,~awvalid_mst} + {
1'b0,prio};
assign res2 = res1[4] + res1;
[email protected](*) begin
if(cur_state == WAIT_AW)
grant = awvalid_mst & res2;
else
grant = grant_tmp;
end
//-----------------AW channel chip select--------------------------
reg [(AWID_WIDTH+2)-1:0] awid_r;
reg [AWADDR_WIDTH-1:0] awaddr_r;
reg [8-1:0] awlen_r;
reg [3-1:0] awsize_r;
reg [2-1:0] awburst_r;
reg awvalid_r;
reg [1*4-1:0] awready_mst_r;
[email protected](*) begin
if(cur_state == WAIT_AW) begin
case(grant)
4'b0001:
awid_r = awid_mst[(AWID_WIDTH+2)-1:0];
4'b0010:
awid_r = awid_mst[(AWID_WIDTH+2)*2-1:(AWID_WIDTH+2)];
4'b0100:
awid_r = awid_mst[(AWID_WIDTH+2)*3-1:(AWID_WIDTH+2)*2];
4'b1000:
awid_r = awid_mst[(AWID_WIDTH+2)*4-1:(AWID_WIDTH+2)*3];
default:
awid_r = 'd0;
endcase
end
else
awid_r = 'd0;
end
assign awid = awid_r;
[email protected](*) begin
if(cur_state == WAIT_AW) begin
case(grant)
4'b0001:
awaddr_r = awaddr_mst[AWADDR_WIDTH-1:0];
4'b0010:
awaddr_r = awaddr_mst[AWADDR_WIDTH*2-1:AWADDR_WIDTH];
4'b0100:
awaddr_r = awaddr_mst[AWADDR_WIDTH*3-1:AWADDR_WIDTH*2];
4'b1000:
awaddr_r = awaddr_mst[AWADDR_WIDTH*4-1:AWADDR_WIDTH*3];
default:
awaddr_r = 'd0;
endcase
end
else
awaddr_r = 'd0;
end
assign awaddr = awaddr_r;
[email protected](*) begin
if(cur_state == WAIT_AW) begin
case(grant)
4'b0001:
awlen_r = awlen_mst[8-1:0];
4'b0010:
awlen_r = awlen_mst[8*2-1:8];
4'b0100:
awlen_r = awlen_mst[8*3-1:8*2];
4'b1000:
awlen_r = awlen_mst[8*4-1:8*3];
default:
awlen_r = 'd0;
endcase
end
else
awlen_r = 'd0;
end
assign awlen = awlen_r;
[email protected](*) begin
if(cur_state == WAIT_AW) begin
case(grant)
4'b0001:
awsize_r = awsize_mst[3-1:0];
4'b0010:
awsize_r = awsize_mst[3*2-1:3];
4'b0100:
awsize_r = awsize_mst[3*3-1:3*2];
4'b1000:
awsize_r = awsize_mst[3*4-1:3*3];
default:
awsize_r = 'd0;
endcase
end
else
awsize_r = 'd0;
end
assign awsize = awsize_r;
[email protected](*) begin
if(cur_state == WAIT_AW) begin
case(grant)
4'b0001:
awburst_r = awburst_mst[2-1:0];
4'b0010:
awburst_r = awburst_mst[2*2-1:2];
4'b0100:
awburst_r = awburst_mst[2*3-1:2*2];
4'b1000:
awburst_r = awburst_mst[2*4-1:2*3];
default:
awburst_r = 'd0;
endcase
end
else
awburst_r = 'd0;
end
assign awburst = awburst_r;
[email protected](*) begin
if(cur_state == WAIT_AW) begin
case(grant)
4'b0001:
awvalid_r = awvalid_mst[1-1:0];
4'b0010:
awvalid_r = awvalid_mst[1*2-1:1];
4'b0100:
awvalid_r = awvalid_mst[1*3-1:1*2];
4'b1000:
awvalid_r = awvalid_mst[1*4-1:1*3];
default:
awvalid_r = 'd0;
endcase
end
else
awvalid_r = 'd0;
end
assign awvalid = awvalid_r;
[email protected](*) begin
if(cur_state == WAIT_AW)
awready_mst_r = awready << grant;
else
awready_mst_r = 'd0;
end
assign awready_mst = awready_mst_r;
//-----------------W channel chip select--------------------------
reg [1*4-1:0] wready_mst_r;
reg [WDATA_WIDTH-1:0] wdata_r;
reg [1-1:0] wlast_r;
reg [1-1:0] wvalid_r;
[email protected](*) begin
if(cur_state == WAIT_AW)
wready_mst_r = 'd0;
else
wready_mst_r = wready << grant;
end
assign wready_mst = wready_mst_r;
[email protected](*) begin
if(cur_state == WAIT_AW)
wdata_r = 'd0;
else begin
case(grant)
4'b0001:
wdata_r = wdata_mst[WDATA_WIDTH-1:0];
4'b0010:
wdata_r = wdata_mst[WDATA_WIDTH*2-1:WDATA_WIDTH];
4'b0100:
wdata_r = wdata_mst[WDATA_WIDTH*3-1:WDATA_WIDTH*2];
4'b1000:
wdata_r = wdata_mst[WDATA_WIDTH*4-1:WDATA_WIDTH*3];
default:
wdata_r = 'd0;
endcase
end
end
assign wdata = wdata_r;
[email protected](*) begin
if(cur_state == WAIT_AW)
wlast_r = 'd0;
else begin
case(grant)
4'b0001:
wlast_r = wlast_mst[0];
4'b0010:
wlast_r = wlast_mst[1];
4'b0100:
wlast_r = wlast_mst[2];
4'b1000:
wlast_r = wlast_mst[3];
default:
wlast_r = 'd0;
endcase
end
end
assign wlast = wlast_r;
[email protected](*) begin
if(cur_state == WAIT_AW)
wvalid_r = 'd0;
else begin
case(grant)
4'b0001:
wvalid_r = wvalid_mst[0];
4'b0010:
wvalid_r = wvalid_mst[1];
4'b0100:
wvalid_r = wvalid_mst[2];
4'b1000:
wvalid_r = wvalid_mst[3];
default:
wvalid_r = 'd0;
endcase
end
end
assign wvalid = wvalid_r;
endmodule
2. Change the state machine to flow line
You can find , In the design of the above state machine ,WAIT_AW In the state of AXI Master Of AW Channels and AXI Slave Of AW Channel signals can communicate with each other , But this time W The channel is blocked . stay WAIT_W State, ,W The signal of the channel is opened , and AW The signal of the channel is blocked .
Now I think of it , stay WAIT_W State, ,AW The passage is directly blocked , Can you let go , Give Way AW Channel normal communication , namely WAIT_AW and WAIT_W The two states are parallel ?
Pipeline design can be transformed from state machine , For example, multiplier can be divided into first order bit Multiply 、 Then add them by dislocation , Then, can we make a list of the new data when we add them out of place bit Multiply ? Sure , According to this idea, the state machine is optimized into a pipeline .
RTL Multiplier IP Design
Of course, not all state machines can be fully optimized into pipelines , Some cannot be optimized at all ( for example USART), Some states can be optimized by pipeline .
that How to WAIT_AW and WAIT_W In parallel ?
AW Channels and W The root cause of channel collaborative design is ,W There's no channel wid, therefore AW The receiving order of the channel must be the same as W The receiving order of the channel is consistent .
therefore , Just every time AW Complete the handshake once , Record the handshake at this time Master Who is it? , namely grant. Each handshake will grant Put in FIFO, and W As for the channel, every time the handshake is completed, it starts from FIFO read out grant Know what to interact with next time Master, That's it AW Channels and W The running water of the passage .
Since the two channels use synchronization FIFO Isolation , Then it can be divided into two modules

2.1. axi_round_robin_arbiter_aw
AW Channel arbitration is easier to say , But still use the state machine , The status here should be changed to FIFO Is it full ,FIFO Full communication will be blocked .
Code
module axi_round_robin_arbiter_aw #(
parameter AWID_WIDTH = 8,
parameter AWADDR_WIDTH = 16,
parameter MAX_PRIO_INITIAL_INDEX = 0
)(
input aclk,
input arstn,
//AW CHANNEL
input [(AWID_WIDTH+2)*4-1:0] awid_mst,
input [AWADDR_WIDTH*4-1:0] awaddr_mst,
input [8*4-1:0] awlen_mst,
input [3*4-1:0] awsize_mst,
input [2*4-1:0] awburst_mst,
input [1*4-1:0] awvalid_mst,
output [1*4-1:0] awready_mst,
output [(AWID_WIDTH+2)-1:0] awid,
output [AWADDR_WIDTH-1:0] awaddr,
output [8-1:0] awlen,
output [3-1:0] awsize,
output [2-1:0] awburst,
output awvalid,
input awready,
input grant_fifo_full,
output [4-1:0] grant,
output grant_val
);
reg [4-1:0] cur_state;
reg [4-1:0] nxt_state;
wire [4-1:0] inital_state;
wire [4:0] res1;
wire [4:0] res2;
assign inital_state = {
{
(4-MAX_PRIO_INITIAL_INDEX-1){
1'b0}},1'b1,{
MAX_PRIO_INITIAL_INDEX{
1'b0}}};
[email protected](posedge aclk or negedge arstn) begin
if(!arstn)
cur_state <= inital_state;
else
cur_state <= nxt_state;
end
[email protected](*) begin
if(grant_val) begin
if(grant[3])
nxt_state = 4'b1;
else
nxt_state = grant << 1;
end
else
nxt_state = cur_state;
end
assign res1 = {
1'b0,~awvalid_mst} + {
1'b0,cur_state};
assign res2 = res1[4] + res1;
assign grant = awvalid_mst & res2;
assign grant_val = awvalid && awready;
//-----------------AW channel chip select--------------------------
reg [(AWID_WIDTH+2)-1:0] awid_r;
reg [AWADDR_WIDTH-1:0] awaddr_r;
reg [8-1:0] awlen_r;
reg [3-1:0] awsize_r;
reg [2-1:0] awburst_r;
reg awvalid_r;
reg [1*4-1:0] awready_mst_r;
[email protected](*) begin
if(!grant_fifo_full) begin
case(grant)
4'b0001:
awid_r = awid_mst[(AWID_WIDTH+2)-1:0];
4'b0010:
awid_r = awid_mst[(AWID_WIDTH+2)*2-1:(AWID_WIDTH+2)];
4'b0100:
awid_r = awid_mst[(AWID_WIDTH+2)*3-1:(AWID_WIDTH+2)*2];
4'b1000:
awid_r = awid_mst[(AWID_WIDTH+2)*4-1:(AWID_WIDTH+2)*3];
default:
awid_r = 'd0;
endcase
end
else
awid_r = 'd0;
end
assign awid = awid_r;
[email protected](*) begin
if(!grant_fifo_full) begin
case(grant)
4'b0001:
awaddr_r = awaddr_mst[AWADDR_WIDTH-1:0];
4'b0010:
awaddr_r = awaddr_mst[AWADDR_WIDTH*2-1:AWADDR_WIDTH];
4'b0100:
awaddr_r = awaddr_mst[AWADDR_WIDTH*3-1:AWADDR_WIDTH*2];
4'b1000:
awaddr_r = awaddr_mst[AWADDR_WIDTH*4-1:AWADDR_WIDTH*3];
default:
awaddr_r = 'd0;
endcase
end
else
awaddr_r = 'd0;
end
assign awaddr = awaddr_r;
[email protected](*) begin
if(!grant_fifo_full) begin
case(grant)
4'b0001:
awlen_r = awlen_mst[8-1:0];
4'b0010:
awlen_r = awlen_mst[8*2-1:8];
4'b0100:
awlen_r = awlen_mst[8*3-1:8*2];
4'b1000:
awlen_r = awlen_mst[8*4-1:8*3];
default:
awlen_r = 'd0;
endcase
end
else
awlen_r = 'd0;
end
assign awlen = awlen_r;
[email protected](*) begin
if(!grant_fifo_full) begin
case(grant)
4'b0001:
awsize_r = awsize_mst[3-1:0];
4'b0010:
awsize_r = awsize_mst[3*2-1:3];
4'b0100:
awsize_r = awsize_mst[3*3-1:3*2];
4'b1000:
awsize_r = awsize_mst[3*4-1:3*3];
default:
awsize_r = 'd0;
endcase
end
else
awsize_r = 'd0;
end
assign awsize = awsize_r;
[email protected](*) begin
if(!grant_fifo_full) begin
case(grant)
4'b0001:
awburst_r = awburst_mst[2-1:0];
4'b0010:
awburst_r = awburst_mst[2*2-1:2];
4'b0100:
awburst_r = awburst_mst[2*3-1:2*2];
4'b1000:
awburst_r = awburst_mst[2*4-1:2*3];
default:
awburst_r = 'd0;
endcase
end
else
awburst_r = 'd0;
end
assign awburst = awburst_r;
[email protected](*) begin
if(!grant_fifo_full) begin
case(grant)
4'b0001:
awvalid_r = awvalid_mst[1-1:0];
4'b0010:
awvalid_r = awvalid_mst[1*2-1:1];
4'b0100:
awvalid_r = awvalid_mst[1*3-1:1*2];
4'b1000:
awvalid_r = awvalid_mst[1*4-1:1*3];
default:
awvalid_r = 'd0;
endcase
end
else
awvalid_r = 'd0;
end
assign awvalid = awvalid_r;
[email protected](*) begin
if(!grant_fifo_full)
awready_mst_r = awready << grant;
else
awready_mst_r = 'd0;
end
assign awready_mst = awready_mst_r;
endmodule
2.2. axi_round_robin_arbiter_w
about W In terms of channels , Must be from FIFO read out grant Talent and grant Stipulated Master handshake , So we still follow the design process of state machine , There are two states

RD_GRANT
Read all the time in this state FIFO, meanwhile W The passage is blocked , until grant Read out

HANDSHAKE
here grant Will always keep , according to grant choice Master Wait for the handshake to complete .

Code
module axi_round_robin_arbiter_w #(
parameter WDATA_WIDTH = 16,
parameter MAX_PRIO_INITIAL_INDEX = 0
)(
input aclk,
input arstn,
//W CHANNEL
input [WDATA_WIDTH*4-1:0] wdata_mst,
input [1*4-1:0] wlast_mst,
input [1*4-1:0] wvalid_mst,
output [1*4-1:0] wready_mst,
output [WDATA_WIDTH-1:0] wdata,
output [1-1:0] wlast,
output [1-1:0] wvalid,
input [1-1:0] wready,
output fifo_rd_en,
input fifo_empty,
input [4-1:0] grant,
input grant_val
);
localparam IDLE = 2'b00;
localparam RD_GRANT = 2'b01;
localparam HANDSHAKE = 2'b11;
reg [1:0] cur_state;
reg [1:0] nxt_state;
reg fifo_rd_en_r;
[email protected](posedge aclk or negedge arstn) begin
if(!arstn)
cur_state <= IDLE;
else
cur_state <= nxt_state;
end
[email protected](*) begin
case(cur_state)
IDLE:
nxt_state = RD_GRANT;
RD_GRANT:
if(grant_val)
nxt_state = HANDSHAKE;
else
nxt_state = RD_GRANT;
HANDSHAKE:
if(wlast && wready)
nxt_state = HANDSHAKE;
else
nxt_state = RD_GRANT;
default: nxt_state = IDLE;
endcase
end
[email protected](posedge aclk or negedge arstn) begin
if(!arstn)
fifo_rd_en_r <= 1'b0;
else if(cur_state == IDLE)
fifo_rd_en_r <= 1'b1;
else if(cur_state == RD_GRANT) begin
if(!fifo_empty)
fifo_rd_en_r <= 1'b0;
end
else if(cur_state == HANDSHAKE) begin
if(wlast && wready)
fifo_rd_en_r <= 1'b1;
end
end
assign fifo_rd_en = fifo_rd_en_r;
//-----------------W channel chip select--------------------------
reg [1*4-1:0] wready_mst_r;
reg [WDATA_WIDTH-1:0] wdata_r;
reg [1-1:0] wlast_r;
reg [1-1:0] wvalid_r;
[email protected](*) begin
if(!cur_state[0]) //(cur_state == IDLE) && (cur_state == RD_GRANT)
wready_mst_r = 'd0;
else
wready_mst_r = wready << grant;
end
assign wready_mst = wready_mst_r;
[email protected](*) begin
if(!cur_state[0])
wdata_r = 'd0;
else begin
case(grant)
4'b0001:
wdata_r = wdata_mst[WDATA_WIDTH-1:0];
4'b0010:
wdata_r = wdata_mst[WDATA_WIDTH*2-1:WDATA_WIDTH];
4'b0100:
wdata_r = wdata_mst[WDATA_WIDTH*3-1:WDATA_WIDTH*2];
4'b1000:
wdata_r = wdata_mst[WDATA_WIDTH*4-1:WDATA_WIDTH*3];
default:
wdata_r = 'd0;
endcase
end
end
assign wdata = wdata_r;
[email protected](*) begin
if(!cur_state[0])
wlast_r = 'd0;
else begin
case(grant)
4'b0001:
wlast_r = wlast_mst[0];
4'b0010:
wlast_r = wlast_mst[1];
4'b0100:
wlast_r = wlast_mst[2];
4'b1000:
wlast_r = wlast_mst[3];
default:
wlast_r = 'd0;
endcase
end
end
assign wlast = wlast_r;
[email protected](*) begin
if(!cur_state[0])
wvalid_r = 'd0;
else begin
case(grant)
4'b0001:
wvalid_r = wvalid_mst[0];
4'b0010:
wvalid_r = wvalid_mst[1];
4'b0100:
wvalid_r = wvalid_mst[2];
4'b1000:
wvalid_r = wvalid_mst[3];
default:
wvalid_r = 'd0;
endcase
end
end
assign wvalid = wvalid_r;
endmodule
边栏推荐
- Illustration of ONEFLOW's learning rate adjustment strategy
- Hot Recruitment! The second Tencent light · public welfare innovation challenge is waiting for you to participate
- 2021-05-08
- The new version of Alibaba Seata finally solves the idempotence, suspension and empty rollback problems of the TCC mode
- Penetration test - right raising topic
- NFNet:NF-ResNet的延伸,不用BN的4096超大batch size训练 | 21年论文
- 一款自动生成单元测试的 IDEA 插件
- Uniswap 收购 NFT交易聚合器 Genie,NFT 交易市场将生变局?
- Instructions for laravel8 Beanstalk
- Auto - vérification recommandée! Les bogues MySQL ne font pas reculer les transactions, peut - être êtes - vous à risque!
猜你喜欢

Unity realizes the function of playing Ogg format video

今年英语高考,CMU用重构预训练交出134高分,大幅超越GPT3

建议自查!MySQL驱动Bug引发的事务不回滚问题,也许你正面临该风险!

2021-04-15

首个大众可用PyTorch版AlphaFold2复现,哥大开源OpenFold,star量破千

When I went to oppo for an interview, I got numb...

渗透测试-提权专题

Introduction to helm basics helm introduction and installation

信贷产品额度定价场景下的回归模型效果评估

小米为何深陷芯片泥潭?
随机推荐
[Level 2 warranty] which brand of Fortress machine is good for Level 2 warranty?
系统设计与分析-技术报告-定时清理验证码的一种解决方案
等保备案是什么意思?应该去哪里办理备案?
2021-06-07
2022 college entrance examination quarterly essay winners announced
期货怎么开户安全吗,期货手续费哪家期货公司比较低,适合散户开户?
Is flush a stock? Is it safe to open an account online now?
ASP. Net C pharmacy management information system (including thesis) graduation project [demonstration video]
Uniswap acquires genie, an NFT transaction aggregator. Will the NFT transaction market change?
LEGO announces price increase, speculators are more excited
如何解决 Iterative 半监督训练 在 ASR 训练中难以落地的问题丨RTC Dev Meetup
2021-05-08
Common methods and descriptions of beanstalk
When pandas met SQL, a powerful tool library was born
The data value reported by DTU cannot be filled into Tencent cloud database through Tencent cloud rule engine
The principle of redis cache consistency deep analysis
golang--文件的多个处理场景
小米为何深陷芯片泥潭?
Pyqt5 工具盒使用
The new version of Alibaba Seata finally solves the idempotence, suspension and empty rollback problems of the TCC mode