-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathnn.tlv
503 lines (474 loc) · 22.8 KB
/
nn.tlv
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
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
\m4_TLV_version 1d: tl-x.org
\SV
m4+definitions(['
m4_define(['M4_PRETRAINED'],1)
m4_define(['M4_INPUTDATAWIDTH'],8)
m4_define(['M4_NUMWEIGHT_LAYER1'],30)
m4_define(['M4_NUMWEIGHT_LAYER2'],30)
m4_define(['M4_NUMWEIGHT_LAYER3'],30)
m4_define(['M4_WEIGHTINTWIDTH'],1)
m4_define(['M4_ADDRESSWIDTH'],5)
m4_define(['M4_NUM_LAYER1_NEURONS'],30)
m4_define(['M4_NUM_LAYER2_NEURONS'],30)
m4_define(['M4_NUM_LAYER3_NEURONS'],30)
'])
//Bias Memory - it contains only a single BIAS value (hence only need for 'rd_en' and 'rd_address' ports)
//parameter : /_top -> top scope for bias memory
// /_biasmem_scope -> scope of bias memory
// #_inputdatawidth -> inputDataWidth
// @_stage_wr -> Write stage (prensently not used)
// @_stage_rd -> Read stage (FETCH stage)
// $_biasmem_rd_data -> Output from bias memory
\TLV biasmem(/_top, /_biasmem_scope, #_inputdatawidth, @_stage_wr, @_stage_rd, $_biasmem_rd_data)
m4_default(['M4_PRETRAINED'], 1)
m4_ifelse_block(M4_PRETRAINED, 1, ['
/_top
@_stage_wr
\SV_plus
logic [#_inputdatawidth-1:0] biasmem [0:0];
assign biasmem = '{
{#_inputdatawidth'b1000}
};
/_biasmem_scope
$value[#_inputdatawidth-1:0] = *biasmem\[0\];
'], ['
/_top
@_stage_wr
/_biasmem_scope
$wr = $biasmem_wr_en;
$value[#_inputdatawidth-1:0] = $reset ? '0 : $biasmem_wr_data;
'])
/_top
@_stage_rd
$_biasmem_rd_data[#_inputdatawidth-1:0] = /_biasmem_scope>>m4_stage_eval(1)$value;
//Weight Memory - it contains Weights value stored in array (presently doesnt support write operation)
//parameter : /_top -> top scope for weight memory
// /_mem_scope -> scope of weight memory
// #_inputdatawidth -> inputDataWidth
// #_numdepth -> Number of weights per neuron
// @_stage_wr -> Write stage (prensently not used)
// @_stage_rd -> Read stage (FETCH stage)
// $_mem_rd_en -> weigth memory read enable
// $_mem_rd_address -> weight memory rrread address
// $_mem_rd_data -> Output data from weight memory
\TLV weightmem(/_top, /_mem_scope, #_inputdatawidth, #_numdepth, @_stage_wr, @_stage_rd, $_mem_rd_en, $_mem_rd_address, $_mem_rd_data)
m4_pushdef(['m4_mem_scope'], m4_strip_prefix(/_mem_scope))
m4_default(['M4_PRETRAINED'], 1)
m4_ifelse_block(M4_PRETRAINED, 1, ['
/_top
@_stage_wr
\SV_plus
logic [#_inputdatawidth-1:0] mem [#_numdepth-1:0];
//\$readmemb(weightfile, mem);
assign mem = '{
{#_inputdatawidth'b000},
{#_inputdatawidth'b10},
{#_inputdatawidth'b1},
{#_inputdatawidth'b01},
{#_inputdatawidth'b0},
{#_inputdatawidth'b0},
{#_inputdatawidth'b00},
{#_inputdatawidth'b000},
{#_inputdatawidth'b0},
{#_inputdatawidth'b0},
{#_inputdatawidth'b101},
{#_inputdatawidth'b11},
{#_inputdatawidth'b00},
{#_inputdatawidth'b11},
{#_inputdatawidth'b1},
{#_inputdatawidth'b10},
{#_inputdatawidth'b00},
{#_inputdatawidth'b0},
{#_inputdatawidth'b0},
{#_inputdatawidth'b0},
{#_inputdatawidth'b0},
{#_inputdatawidth'b1},
{#_inputdatawidth'b011},
{#_inputdatawidth'b10},
{#_inputdatawidth'b00},
{#_inputdatawidth'b0},
{#_inputdatawidth'b001},
{#_inputdatawidth'b0},
{#_inputdatawidth'b0},
{#_inputdatawidth'b0011}
};
/_mem_scope[#_numdepth-1:0]
$value[#_inputdatawidth-1:0] = *mem\[#m4_mem_scope\];
'], ['
/_top
@_stage_wr
/_mem_scope[#_numdepth-1:0]
$wr = $mem_wr_en && ($mem_wr_address == #m4_mem_scope);
$value[#_inputdatawidth-1:0] = $reset ? '0 :
$wr ? $mem_wr_data :
$RETAIN;
'])
/_top
@_stage_rd
$_mem_rd_data[#_inputdatawidth-1:0] = ($_mem_rd_en) ? /_mem_scope[$_mem_rd_address]>>m4_stage_eval(1)$value : '0;
m4_popdef(['m4_mem_scope'])
//Neuron - Architecture of single neuron
//parameter : /_top -> top scope for neuron
// /_biasmem -> scope for bias memory
// /_weightmem -> scope of weight memory
// #_layernum -> Indicates this neuron is part of which layer (i.e. column)
// #_neuronnum -> Indicates the number of this neuron (i.e. row)
// #_pipedepth -> No. of Pipeline depth in neuron (presently supports only 5 cofiguration with max 4-cycle of depth)
// #_numinputweights -> No. of input weights(i.e. previous layer's neuronnum) to this neuron.
// #_inputdatawidth -> inputDataWidth
// #_weightintwidth -> weight int value(because of fixed point format)
// $_reset -> external reset
// $_myinput -> input data to neuron(1-by-1)
// $_myinputvalid -> input data valid
// $_out -> output from neuron (after all the (#_numinputweights + #_pipedepth) cycles from 1st valid input)
// $_outvalid -> 1-cycle output valid
//Working :- The valid "input value' gets multipled with "valid weight value" and is accumulated and at last cycle
// the finalsum is added with bias, which then passes though "ReLU" activation function
// (due to fixed #_inputdatawidth, we saturate the max relu value to 1 when it overflows).
\TLV neuron(/_top, /_neuron, /_biasmem, /_weightmem, #_layernum, #_neuronnum, #_pipedepth, #_numinputweights, #_inputdatawidth, #_weightintwidth, $_reset, $_myinput, $_myinputvalid, $_out, $_outvalid)
m4_pushdef(['m4_pipenum'], #_pipedepth) // configuring stages based on #_pipedepth
m4_default(['m4_pipenum'], 1)
m4_case(m4_pipenum, 1, ['
m4_pushdef(['M4_FETCH_STAGE'],0)
m4_pushdef(['M4_MUL_STAGE'],0)
m4_pushdef(['M4_SUM_STAGE'],0)
m4_pushdef(['M4_ACT_STAGE'],0)
m4_pushdef(['M4_OUT_STAGE'],0)
'], 2, ['
m4_pushdef(['M4_FETCH_STAGE'],0)
m4_pushdef(['M4_MUL_STAGE'],0)
m4_pushdef(['M4_SUM_STAGE'],0)
m4_pushdef(['M4_ACT_STAGE'],1)
m4_pushdef(['M4_OUT_STAGE'],1)
'], 3, ['
m4_pushdef(['M4_FETCH_STAGE'],0)
m4_pushdef(['M4_MUL_STAGE'],1)
m4_pushdef(['M4_SUM_STAGE'],1)
m4_pushdef(['M4_ACT_STAGE'],2)
m4_pushdef(['M4_OUT_STAGE'],2)
'], 4, ['
m4_pushdef(['M4_FETCH_STAGE'],0)
m4_pushdef(['M4_MUL_STAGE'],1)
m4_pushdef(['M4_SUM_STAGE'],2)
m4_pushdef(['M4_ACT_STAGE'],3)
m4_pushdef(['M4_OUT_STAGE'],3)
'], 5, ['
m4_pushdef(['M4_FETCH_STAGE'],0)
m4_pushdef(['M4_MUL_STAGE'],1)
m4_pushdef(['M4_SUM_STAGE'],2)
m4_pushdef(['M4_ACT_STAGE'],3)
m4_pushdef(['M4_OUT_STAGE'],4)
']
)
m4_define(['M4_ADDRESSWIDTH'], \$clog2(#_numinputweights))
/_neuron
@M4_FETCH_STAGE
$irst = /_top<>0$_reset;
$input[#_inputdatawidth-1:0] = $rand[#_inputdatawidth-1:0]; // /_top$_myinput[#_inputdatawidth-1:0];
$input_valid = /_top$_myinputvalid;
m4+biasmem(/_neuron,/_biasmem, #_inputdatawidth, @M4_FETCH_STAGE, @M4_FETCH_STAGE, $biasReg)
/_neuron
@M4_FETCH_STAGE
$bias[(2 * #_inputdatawidth) - 1 : 0] = {$biasReg[#_inputdatawidth-1:0] ,{#_inputdatawidth{1'b0}}};
$mem_rd_address[M4_ADDRESSWIDTH:0] = ($irst | >>m4_eval(M4_OUT_STAGE - M4_FETCH_STAGE + 1)$_outvalid ) ? '0 :
($input_valid) ? >>1$mem_rd_address + 1'b1 :
$RETAIN;
$mem_rd_en = $input_valid;
m4+weightmem(/_neuron, /_weightmem, #_inputdatawidth, #_numinputweights, @M4_FETCH_STAGE, @M4_MUL_STAGE, $mem_rd_en, $mem_rd_address, $mem_out)
/_neuron
@M4_MUL_STAGE
$mul[(2 * #_inputdatawidth)-1:0] = \$signed($input) * \$signed($mem_out); // $signed multiplication
/_neuron
@M4_SUM_STAGE
$sumAdd[(2 * #_inputdatawidth)-1:0] = $mul + >>1$sum;
$biasAdd[(2 * #_inputdatawidth)-1:0] = $bias + >>1$sum;
$sum[(2 * #_inputdatawidth)-1:0] = ($irst | >>m4_eval(M4_OUT_STAGE - M4_SUM_STAGE + 1)$_outvalid) ? '0 :
(($mem_rd_address == #_numinputweights) && (!$input_valid & >>1$input_valid)) ? (
((! $bias[(2 * #_inputdatawidth)-1]) & (! >>1$sum[(2 * #_inputdatawidth)-1]) & ( $biasAdd[(2 * #_inputdatawidth)-1])) ? {1'b0, {((2 * #_inputdatawidth)-1){1'b1}}} : // Overflow between bias and sum
(( $bias[(2 * #_inputdatawidth)-1]) & ( >>1$sum[(2 * #_inputdatawidth)-1]) & (! $biasAdd[(2 * #_inputdatawidth)-1])) ? {1'b1, {((2 * #_inputdatawidth)-1){1'b0}}} : // Underflow between bias and sum
{$biasAdd} ) :
($input_valid) ? (
((! $mul[(2 * #_inputdatawidth)-1]) & (! >>1$sum[(2 * #_inputdatawidth)-1]) & ( $sumAdd[(2 * #_inputdatawidth)-1])) ? {1'b0, {((2 * #_inputdatawidth)-1){1'b1}}} : // Overflow between weight and sum
(( $mul[(2 * #_inputdatawidth)-1]) & ( >>1$sum[(2 * #_inputdatawidth)-1]) & (! $sumAdd[(2 * #_inputdatawidth)-1])) ? {1'b1, {((2 * #_inputdatawidth)-1){1'b0}}} : // Underflow between weight and sum
{$sumAdd} ) : '0;
@M4_ACT_STAGE
$actvalid = (($mem_rd_address == #_numinputweights) && (!$input_valid & >>1$input_valid));
// "ReLU" activation function
?$actvalid
$out_act[#_inputdatawidth-1:0] = ($sum[(2 * #_inputdatawidth)-1] == 0) ? (
(| $sum[(2 * #_inputdatawidth)-1 -: #_weightintwidth+1]) ? {1'b0, {(#_inputdatawidth-1){1'b1}}} : {$sum[(2 * #_inputdatawidth)-1-#_weightintwidth -: #_inputdatawidth]}
) : '0;
@M4_OUT_STAGE
$_outvalid = $actvalid;
$_out[#_inputdatawidth-1:0] = $out_act;
m4_popdef(['m4_pipenum'])
m4_popdef(['M4_FETCH_STAGE'])
m4_popdef(['M4_MUL_STAGE'])
m4_popdef(['M4_SUM_STAGE'])
m4_popdef(['M4_ACT_STAGE'])
m4_popdef(['M4_OUT_STAGE'])
//Laeyer - Architecture of Layer
//parameter : /_top -> top scope for Layer
// /_layer -> scope for layer
// /_layerhier -> scope for layer replication (i.e. hierarchy)
// /_neuron -> scope for _neuron
// /_biasmem -> scope for bias memory
// /_weightmem -> scope of weight memory
// #_layernum -> Indicates layer no.
// #_numneuron -> Indicates the number of neuron present in this layer
// #_pipedepth -> No. of Pipeline depth in neuron (presently supports only 5 cofiguration with max 4-cycle of depth)
// #_numinputweights -> No. of input weights(i.e. previous layer's neuronnum) to this neuron.
// #_inputdatawidth -> inputDataWidth
// #_weightintwidth -> weight int value(because of fixed point format)
// $_reset -> external reset
// $_myinput -> input data to all neurons in layer at once (1-by-1)
// $_myinputvalid -> input data valid
// $_out -> output from layer
// $_outvalid -> 1-cycle output valid
\TLV layer(/_top, /_layer, /_layerhier, /_neuron, /_biasmem, /_weightmem, #_layernum, #_numneuron, #_pipedepth , #_numinputweights, #_inputdatawidth, #_weightintwidth, $_reset, $_myinput, $_myinputvalid, $_out, $_outvalid)
m4_pushdef(['m4_layerhier'], m4_strip_prefix(/_layerhier))
m4_pushdef(['m4_pipenum'], #_pipedepth)
m4_default(['m4_pipenum'], 1)
m4_case(m4_pipenum, 1, ['
m4_pushdef(['m4_fetch_stage'],0)
m4_pushdef(['m4_out_stage'],0)
'], 2, ['
m4_pushdef(['m4_fetch_stage'],0)
m4_pushdef(['m4_out_stage'],1)
'], 3, ['
m4_pushdef(['m4_fetch_stage'],0)
m4_pushdef(['m4_out_stage'],2)
'], 4, ['
m4_pushdef(['m4_fetch_stage'],0)
m4_pushdef(['m4_out_stage'],3)
'], 5, ['
m4_pushdef(['m4_fetch_stage'],0)
m4_pushdef(['m4_out_stage'],4)
']
)
/_layer
@m4_fetch_stage
$reset = /_top<>0$_reset;
$myinputvalid = /_top$_myinputvalid;
//$myinput[#_inputdatawidth - 1:0] = /_top$_myinput[#_inputdatawidth - 1:0];
/_layerhier[m4_eval(#_numneuron - 1):0]
m4_pushdef(['m4_neuronnum'], #m4_layerhier)
m4+neuron(/_top/_layer, /_neuron, /_biasmem, /_weightmem, #_layernum, m4_neuronnum, #_pipedepth, #_numinputweights, #_inputdatawidth, #_weightintwidth, $reset, $myinput, $myinputvalid, $out, $outvalid)
@0
$test[31:0] = m4_neuronnum;
`BOGUS_USE($test)
m4_popdef(['m4_neuronnum'])
@m4_out_stage
$_out[(#_numneuron * #_inputdatawidth) - 1:0] = /_layerhier[*]/_neuron$out;
$_outvalid[(#_numneuron) - 1:0] = /_layerhier[*]/_neuron$outvalid;
m4_popdef(['m4_layernum'])
m4_popdef(['m4_pipenum'])
m4_popdef(['m4_fetch_stage'])
m4_popdef(['m4_out_stage'])
\SV
m4_makerchip_module // (Expanded in Nav-TLV pane.)
\TLV
\viz_js
box: {strokeWidth: 0, left: -100, top: -75, width: 550, height: 1400, fill: "#BBBBBB"},
init() {
let widgets = {}
widgets.title = new fabric.Text("Neural Network Architecture", {
left: 180, top: -50,
originX: "center",
fontSize: 22, fontFamily: "Courier New", fontWeight: "bold"
})
widgets.input = new fabric.Text("Input", {
left: -80, top: 520,
fontSize: 20, fontFamily: "Courier New",
})
return widgets
}
|pipe1
@0
$reset = *reset;
$cnt[\$clog2(M4_NUMWEIGHT_LAYER1) : 0] = $reset ? '0 :
>>1$cnt + 1;
$myinputvalid1 = (($cnt >= 1) && ($cnt <= M4_NUMWEIGHT_LAYER1));
\viz_js
//template: {dot: ["Circle", {radius: 30, fill: "red"}]},
box: {strokeWidth: 0, left: 20, top: 35, width: 50, height: 400},
init() {
let widgets = {}
widgets.title = new fabric.Text("Layer 1", {
left: 50, top: 0,
originX: "center",
fontSize: 20, fontFamily: "Courier New",
})
return widgets
}
//m4+neuron(|pipe, /neuron, /biasmem, /weightmem, 0, 0, 5, M4_NUMWEIGHT, M4_INPUTDATAWIDTH, M4_WEIGHTINTWIDTH, $reset, $myinput, $myinputvalid, $out, $outvalid)
m4+layer(|pipe1, /layer1, /layernum1, /neuron1, /biasmem1, /weightmem1, 0, 30, 5, M4_NUMWEIGHT_LAYER1, M4_INPUTDATAWIDTH, M4_WEIGHTINTWIDTH, $reset, $myinput1, $myinputvalid1, $out1, $outvalid1)
@4
/layer1
/layernum1[29:0]
\viz_js
layout: "vertical",
box: {top: 30, left: 40, strokeWidth: 0, width: 40, height: 40},
render() {
let num = '/neuron1$out'.asInt()*15
return [
new fabric.Circle({
radius: 10,
fill: `rgb(${num}, 0, 0)`,
style: {
margin: 2
}
}),
]
},
where: {top: 80, left: 80},
|pipe2
@0
$ANY = /top|pipe1/layer1>>4$ANY;
$count_layer1[\$clog2(30) : 0] = (/top|pipe1<>0$reset | (>>1$count_layer1 == M4_NUMWEIGHT_LAYER1)) ? '0 :
($outvalid1[0] & ! >>1$outvalid1[0]) ? 1 :
(>>1$count_layer1 >= 1) ? >>1$count_layer1 + 1'b1 : $RETAIN;
$holddata1[m4_eval(M4_NUMWEIGHT_LAYER1 * M4_INPUTDATAWIDTH) - 1:0] = (/top|pipe1<>0$reset | (>>1$count_layer1 == M4_NUMWEIGHT_LAYER1)) ? '0 :
($outvalid1[0] & ! >>1$outvalid1[0]) ? $out1 :
(>>1$count_layer1 >= 1) ? (>>1$holddata1 >> M4_INPUTDATAWIDTH) : $RETAIN;
$myinput2[M4_INPUTDATAWIDTH - 1:0] = $holddata1[M4_INPUTDATAWIDTH - 1:0];
$myinputvalid2 = (/top|pipe1<>0$reset | (>>1$count_layer1 == M4_NUMWEIGHT_LAYER1)) ? '0 :
($outvalid1[0] & ! >>1$outvalid1[0]) ? 1'b1 :
(>>1$count_layer1 >= 1) ? 1'b1 : $RETAIN;
\viz_js
box: {strokeWidth: 0, left: 150, top: 35, width: 50, height: 400},
init() {
let widgets = {}
widgets.title = new fabric.Text("Layer 2", {
left: 170, top: 0,
originX: "center",
fontSize: 20, fontFamily: "Courier New",
})
return widgets
}
m4+layer(|pipe2, /layer2, /layernum2, /neuron2, /biasmem2, /weightmem2, 1, 30, 5, M4_NUMWEIGHT_LAYER2, M4_INPUTDATAWIDTH, M4_WEIGHTINTWIDTH, $reset, $myinput2, $myinputvalid2, $out2, $outvalid2)
@4
/layer2
/layernum2[29:0]
\viz_js
layout: "vertical",
box: {top: 30, left: 0, strokeWidth: 0, width: 40, height: 40},
render() {
let num = '/neuron2$out'.asInt()*15
return [
new fabric.Circle({
radius: 10,
fill: `rgb(${num}, 0,0)`,
})
]
},
where: {top: 80, left: 160},
@4
/layer2
/layernum2[29:0]
\viz_js
layout: {top: 0, left: 0}
/layernum1[29:0]
\viz_js
box: {strokeWidth: 0},
layout: {top: 40, left: 0},
render(){
let valr = '/layernum2/neuron2$out'.asInt()*10
let valg = '/top|pipe1/layer1/layernum1/neuron1$out'.asInt()*3
console.log(valg)
return [new fabric.Line([
0, 0, 100,
(this.getIndex("layernum2") - this.getIndex("layernum1")) * 40],
{stroke: `rgb(${valr}, ${valg}, 0)`, strokeWidth: 3})
]
},
where: {top: 60, left: 60},
|pipe3
@0
$ANY = /top|pipe2/layer2>>4$ANY;
$count_layer2[\$clog2(M4_NUMWEIGHT_LAYER2) : 0] = (/top|pipe1<>0$reset | (>>1$count_layer2 == M4_NUMWEIGHT_LAYER2)) ? '0 :
($outvalid2[0] & ! >>1$outvalid2[0]) ? 1 :
(>>1$count_layer2 >= 1) ? >>1$count_layer2 + 1'b1 : $RETAIN;
$holddata2[m4_eval(M4_NUMWEIGHT_LAYER2 * M4_INPUTDATAWIDTH) - 1:0] = (/top|pipe1<>0$reset | (>>1$count_layer2 == M4_NUMWEIGHT_LAYER2)) ? '0 :
($outvalid2[0] & ! >>1$outvalid2[0]) ? $out2 :
(>>1$count_layer2 >= 1) ? (>>1$holddata2 >> M4_INPUTDATAWIDTH) : $RETAIN;
$myinput3[M4_INPUTDATAWIDTH - 1:0] = $holddata2[M4_INPUTDATAWIDTH - 1:0];
$myinputvalid3 = (/top|pipe1<>0$reset | (>>1$count_layer2 == M4_NUMWEIGHT_LAYER2)) ? '0 :
($outvalid2[0] & ! >>1$outvalid2[0]) ? 1'b1 :
(>>1$count_layer2 >= 1) ? 1'b1 : $RETAIN;
\viz_js
box: {strokeWidth: 0, left: 290, top: 35, width: 50, height: 400},
init() {
let widgets = {}
widgets.title = new fabric.Text("Layer 3", {
left: 320, top: 0,
originX: "center",
fontSize: 20, fontFamily: "Courier New",
})
return widgets
}
m4+layer(|pipe3, /layer3, /layernum3, /neuron3, /biasmem3, /weightmem3, 2, 30, 5, M4_NUMWEIGHT_LAYER3, M4_INPUTDATAWIDTH, M4_WEIGHTINTWIDTH, $reset, $myinput3, $myinputvalid3, $out3, $outvalid3)
@4
/layer3
/layernum3[29:0]
\viz_js
layout: "vertical",
box: {top: 30, left: 100, strokeWidth: 0, width: 40, height: 40},
render() {
let $valid = '/neuron3$outvalid'.asInt()
let num = '/neuron3$out'.asInt()*10
return [
new fabric.Circle({
radius: 10,
fill: `rgb(${num}, 0, 0)`,
style: {
margin: 2
}
}),
]
},
where: {top: 80, left: 400},
@5
/layer3
/layernum3[29:0]
\viz_js
layout: {top: 0, left: 0}
/layernum2[29:0]
\viz_js
box: {strokeWidth: 0},
layout: {top: 40, left: 0},
render(){
let valr = '/layernum3/neuron3$out'.asInt()*10
let valg = '/top|pipe2/layer2/layernum2/neuron2$out'.asInt()*3
return [new fabric.Line([
0, 0, 120,
(this.getIndex("layernum3") - this.getIndex("layernum2")) * 40],
{stroke: `rgb(${valr}, ${valg}, 0)`, strokeWidth: 3})
]
},
where: {top: 60, left: 180},
|pipe
@0
$result[m4_eval(30 * M4_INPUTDATAWIDTH)-1:0] = /top|pipe3/layer3>>4$out3;
$resultvalid = /top|pipe3/layer3>>4$outvalid3;
\viz_js
box: {strokeWidth: 0},
init() {
let widgets = {}
widgets.title = new fabric.Text("Output", {
left: 400, top: 520,
originX: "center",
fontSize: 20, fontFamily: "Courier New",
})
widgets.layer_name = new fabric.Text("Hidden Layers", {
left: 170, top: 1250,
originX: "center",
fontSize: 20, fontFamily: "Courier New",
})
return widgets
}
// Assert these to end simulation (before Makerchip cycle limit).
*passed = *cyc_cnt > 200;
*failed = 1'b0;
\SV
endmodule