1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
|
`include "gfx/gfx_defs.sv"
module mat_mat_mul
(
input logic clk,
rst_n,
input mat4 a,
b,
input logic in_valid,
out_ready,
output mat4 q,
output logic in_ready,
out_valid
);
mat4 a_hold, b_hold, b_transpose, q_hold, mul_b;
vec4 mul_q;
logic mul_in_ready, mul_in_valid, mul_out_ready, mul_out_valid;
index4 in_index, out_index;
assign in_ready = mul_in_ready && in_index == `INDEX4_MIN;
assign out_valid = mul_out_valid && out_index == `INDEX4_MAX;
assign mul_in_valid = in_valid || in_index != `INDEX4_MIN;
assign mul_out_ready = out_ready || out_index != `INDEX4_MAX;
transpose transpose
(
.in(b),
.out(b_transpose)
);
mat_vec_mul mul
(
.a(in_index == `INDEX4_MIN ? a : a_hold),
.x(mul_b[in_index]),
.q(mul_q),
.in_ready(mul_in_ready),
.in_valid(mul_in_valid),
.out_ready(mul_out_ready),
.out_valid(mul_out_valid),
.*
);
always_comb begin
mul_b = b_hold;
mul_b[0] = b_transpose[0];
q = q_hold;
q[`VECS_PER_MAT - 1] = mul_q;
end
always_ff @(posedge clk or negedge rst_n)
if (!rst_n) begin
in_index <= `INDEX4_MIN;
out_index <= `INDEX4_MIN;
end else begin
if (mul_in_ready && mul_in_valid)
in_index <= in_index + 1;
if (mul_out_ready && mul_out_valid)
out_index <= out_index + 1;
end
always_ff @(posedge clk) begin
if (in_ready) begin
a_hold <= a;
b_hold <= b_transpose;
end
if (mul_out_ready && mul_out_valid)
q_hold[out_index] <= mul_q;
end
endmodule
|
