Stable AXI-IP, working in xsim, verilator and FPGA

.github/workflows/verify.yml

@@ -96,4 +96,4 @@ jobs:
 
   #       mkdir -p run/work_resnet
   #       cd run/work_resnet
-  #       python ../resnet_50.py
+  #       python ../resnet_50.py

deepsocflow/c/example.c

@@ -0,0 +1,5 @@
+#include "runtime.h"
+
+int example_main (){
+  return 0;
+}

deepsocflow/c/xilinx_example.c

@@ -16,4 +16,4 @@ int main()
     hardware_cleanup();
 
     return 0;
-}
+}

deepsocflow/py/model.py

@@ -335,7 +335,6 @@ def verify_inference(self, SIM, SIM_PATH):
         '''
         hw.simulate(SIM=SIM, SIM_PATH=SIM_PATH)
 
-
         '''
         CHECK ERROR
         '''
@@ -400,4 +399,4 @@ def predict_performance(self):
 
         time = clocks_total / (self.hw.FREQ * 1e6)
         mem_bytes = mem_bits / 8
-        return time, mem_bytes
+        return time, mem_bytes

deepsocflow/rtl/counter.sv

@@ -1,28 +1,28 @@
-`timescale 1ns/1ps
-
-module counter #(parameter W = 8)(
-  input  logic clk, rstn_g, rst_l, en,
-  input  logic [W-1:0] max_in,
-  output logic [W-1:0] count,
-  output logic last, last_clk, first
-);
-  logic [W-1:0] max;
-  wire  [W-1:0] count_next = last ? max : count - 1;
-
-  always_ff @(posedge clk)
-    if (!rstn_g)
-      {count, max, last} <= '0;
-    else if (rst_l) begin
-      count  <= max_in;
-      max    <= max_in;
-      last   <= max_in==0;
-    end
-    else if (en) begin
-      last   <= count_next == 0;
-      count  <= count_next;
-    end
-
-  assign last_clk = en && last && rstn_g && !rst_l;
-  assign first = count == max;
-
+`timescale 1ns/1ps
+
+module counter #(parameter W = 8)(
+  input  logic clk, rstn_g, rst_l, en,
+  input  logic [W-1:0] max_in,
+  output logic [W-1:0] count,
+  output logic last, last_clk, first
+);
+  logic [W-1:0] max;
+  wire  [W-1:0] count_next = last ? max : count - 1;
+
+  always_ff @(posedge clk)
+    if (!rstn_g)
+      {count, max, last} <= '0;
+    else if (rst_l) begin
+      count  <= max_in;
+      max    <= max_in;
+      last   <= max_in==0;
+    end
+    else if (en) begin
+      last   <= count_next == 0;
+      count  <= count_next;
+    end
+
+  assign last_clk = en && last && rstn_g && !rst_l;
+  assign first = count == max;
+
 endmodule

deepsocflow/tcl/asic/outputGen.tcl

@@ -11,4 +11,4 @@ write_sdf -view WC_VIEW ../asic/outputs/${design}_WC.sdf
 setAnalysisMode -hold
 set_analysis_view -setup BC_VIEW -hold BC_VIEW
 do_extract_model -view BC_VIEW -format dotlib ${design}_BC.lib
-write_sdf -view BC_VIEW ../asic/outputs/${design}_BC.sdf
+write_sdf -view BC_VIEW ../asic/outputs/${design}_BC.sdf

deepsocflow/test/py/.gitattributes

@@ -0,0 +1 @@
+**/*.ipynb linguist-vendored

deepsocflow/test/py/performance.ipynb

@@ -0,0 +1,123 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "c:\\ProgramData\\Miniconda3\\envs\\qkeras\\lib\\site-packages\\numpy\\_distributor_init.py:30: UserWarning: loaded more than 1 DLL from .libs:\n",
+      "c:\\ProgramData\\Miniconda3\\envs\\qkeras\\lib\\site-packages\\numpy\\.libs\\libopenblas.FB5AE2TYXYH2IJRDKGDGQ3XBKLKTF43H.gfortran-win_amd64.dll\n",
+      "c:\\ProgramData\\Miniconda3\\envs\\qkeras\\lib\\site-packages\\numpy\\.libs\\libopenblas64__v0.3.21-gcc_10_3_0.dll\n",
+      "  warnings.warn(\"loaded more than 1 DLL from .libs:\"\n"
+     ]
+    }
+   ],
+   "source": [
+    "from tensorflow.keras.applications import resnet50\n",
+    "import keras\n",
+    "import numpy as np\n",
+    "\n",
+    "model = resnet50.ResNet50(include_top=True, weights='imagenet', input_shape=(224, 224, 3))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "clocks=51317730.0, mb=96.698956, time_per_image=0.026, fps=38.97, mb_per_image=96.70, latency=0.205\n"
+     ]
+    }
+   ],
+   "source": [
+    "ROWS = 8\n",
+    "COLS = 96\n",
+    "KH_MAX = 7\n",
+    "XB = 1\n",
+    "YB = 3\n",
+    "KB = 1\n",
+    "XN = ROWS\n",
+    "MHZ = 250\n",
+    "\n",
+    "total_clocks = 0\n",
+    "total_mem = 0\n",
+    "\n",
+    "for i, layer in enumerate(model.layers):\n",
+    "    if isinstance(layer, keras.layers.convolutional.conv2d.Conv2D):\n",
+    "\n",
+    "        YH, YW, CO = layer.output.shape[1:]\n",
+    "        KH, KW, CI, _ = layer.kernel.shape\n",
+    "        SH, SW = layer.strides\n",
+    "        XH = YH*SH\n",
+    "        XW = YW*SW\n",
+    "\n",
+    "    elif isinstance(layer, keras.layers.core.dense.Dense):\n",
+    "        XH = XN\n",
+    "        CI, CO = layer.kernel.shape\n",
+    "        XW = KH = KW = 1\n",
+    "    \n",
+    "    else:\n",
+    "        continue\n",
+    "        \n",
+    "    L  = np.ceil(XH/ROWS)\n",
+    "    IT = np.ceil(CO/np.floor(COLS/KW))\n",
+    "\n",
+    "    clocks = IT * (1+ XN*L*XW*(1+CI*KH))\n",
+    "    mem_access = \\\n",
+    "        XB * (IT * XN * L * XW * CI * (ROWS + KH_MAX-1)) +\\\n",
+    "        KB * (IT * CI * KH * COLS) +\\\n",
+    "        YB * (IT * XN * L * XW * np.floor(COLS/KW) * ROWS)\n",
+    "    \n",
+    "    if layer.name in ['conv2_block1_0_conv', 'conv3_block1_0_conv', 'conv4_block1_0_conv', 'conv5_block1_0_conv']:\n",
+    "        continue\n",
+    "        total_mem += YB * (IT * XN * L * XW * np.floor(COLS/KW) * ROWS)\n",
+    "\n",
+    "    total_clocks += clocks\n",
+    "    total_mem += mem_access\n",
+    "\n",
+    "\n",
+    "time = total_clocks/MHZ/1e6/XN\n",
+    "mem_access = total_mem/1024/XN/1024\n",
+    "\n",
+    "print(f'clocks={total_clocks}, mb={mem_access:0f}, time_per_image={time:.3f}, fps={1/time:.2f}, mb_per_image={mem_access:.2f}, latency={XN*time:.3f}')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "qkeras",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.9"
+  },
+  "orig_nbformat": 4
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}