[API] Multiple report file parsing support

python/heterocl/report.py

@@ -3,8 +3,6 @@
 import time
 import xmltodict
 import pandas as pd
-# Support for graphical display of the report
-#import matplotlib.pyplot as plt
 from .report_config import RptSetup
 from tabulate import tabulate
 from .schedule import Stage
@@ -57,6 +55,10 @@ class Displayer(object):
 
     get_max(col)
         Sort the latency in a decreasing order for specific latency category.
+
+    add_fields(val)
+        Dictionary representation of the Displayer object's member of 
+        report files.
 
     display(loops=None, level=None, cols=None)
         Display the report table with appropriate query arguments.
@@ -373,6 +375,31 @@ def get_max(self, col):
         tup_lst = list(map(lambda x: (x[0], x[1], x[2].count('+')), tup_lst))
         return list(reversed(sorted(tup_lst, key=lambda x: int(x[1]))))
 
+    def add_fields(self, val): 
+        """Append additional data present in separate report files.
+
+        Parameters
+        ----------
+        val: dict
+            Dictionary representation of the Displayer object's member
+            of report files.
+
+        Returns
+        ----------
+        None
+        """
+        self._category = val['_category']
+        self._category_aux = val['_category_aux']
+        self._loop_name += val['_loop_name']
+        self._loop_name_aux += val['_loop_name_aux']
+        self._max_level = val['_max_level'] if val['_max_level'] > self._max_level else self._max_level
+
+        for key in self._category_aux:
+            try:
+                self._data[key] += val['_data'][key]
+            except:
+                self._data[key] = val['_data'][key]
+
     def display(self, loops=None, level=None, cols=None):
         """Display the report file.
 
@@ -431,7 +458,7 @@ def display(self, loops=None, level=None, cols=None):
         print('* Units in {}'.format(self.unit))
         splt = df.loc[rows, cols].to_string().split("\n")
         pd.set_option('max_colwidth', len(splt[0]) * 100)
-        return df.loc[rows, cols].to_string()
+        return df.loc[rows, cols].to_string() 
 
 def parse_js(path, print_flag=False):
     js_file = os.path.join(path, "kernel/reports/lib/report_data.js")
@@ -456,17 +483,25 @@ def parse_js(path, print_flag=False):
 def parse_xml(path, xml_path, prod_name, print_flag=False):
     xml_file = os.path.join(path, xml_path)
 
+    # Collect files other than the main one.
+    p = xml_file.rsplit('/', 1)[0]
+    other_xml_file = []
+    for file in os.listdir(p):
+        if file.endswith("_csynth.xml") and file != "test_csynth.xml":
+            fpath = os.path.join(p, file)
+            other_xml_file.append(fpath)
+
+    # Display the general information in `test_csynth.xml`
     if not os.path.isfile(xml_file):
         raise RuntimeError("Cannot find {}, run csyn first".format(xml_file))
     json_file = os.path.join(path,"report.json")
     outfile = open(json_file, "w")
     with open(xml_file, "r") as xml:
         profile = xmltodict.parse(xml.read())["profile"]
         json.dump(profile, outfile, indent=2)
-
+    
     config = RptSetup(profile, prod_name)
     config.eval_members()
-
     res = {}
     res["HLS Version"] = config.prod_name + " " + config.version
     res["Product family"] = config.prod_family
@@ -504,17 +539,40 @@ def parse_xml(path, xml_path, prod_name, print_flag=False):
     tablestr.insert(5, splitline)
     table = '\n'.join(tablestr)
 
-    # Latency information extraction
     clock_unit = config.performance_unit
+
+    # Parse latency information in the main report file (if it exists)
     summary = config.loop_latency
+    out_info_table = Displayer(clock_unit)
+    try:
+        out_info_table.init_table(summary)
+        out_info_table.collect_data(summary)
+    except:
+        pass
+
+    # Latency information extraction
+    for lat_xml_file in other_xml_file:
+        with open(lat_xml_file, "r") as xml:
+            profile = xmltodict.parse(xml.read())["profile"]
+
+        config = RptSetup(profile, prod_name)
+        config.eval_members()
+
+        summary = config.loop_latency
 
-    info_table = Displayer(clock_unit)
-    info_table.init_table(summary)
-    info_table.collect_data(summary)
+        try:
+            info_table = Displayer(clock_unit)
+            info_table.init_table(summary)
+            info_table.collect_data(summary)
+            res = vars(info_table)
+            out_info_table.add_fields(res)
+        except:
+            print("Report for issue")
+            pass
 
     if print_flag:
-        print(table)
-    return info_table
+        print(table) 
+    return out_info_table
 
 def report_stats(target, folder):
     path = folder

python/heterocl/report_config.py

@@ -82,8 +82,7 @@ def _lookup(self, keys):
         try:
             return reduce(operator.getitem, keys, self.profile)
         except KeyError:
-            print("Invalid key")
-            raise
+            pass
 
     def eval_members(self):
         """Initialize each attribute to appropriate values.

tests/test_hls_report.py

@@ -1,5 +1,6 @@
 import heterocl as hcl
 import numpy as np
+import os
 import re
 import json
 import xmltodict
@@ -92,6 +93,33 @@ def spam_filter():
     pass
 # END TODO
 
+def stages():
+
+    A = hcl.placeholder((32, 32), "A")
+    C = hcl.placeholder((32, 32), "C")
+    def kernel(A, C):
+        B = hcl.compute(A.shape, lambda i, j : A[i, j] + 1, "B")
+        D = hcl.compute(A.shape, lambda i, j : B[i, j] + 1, "D")
+        E = hcl.compute(A.shape, lambda i, j : C[i, j] + 1, "E")
+        F = hcl.compute(A.shape, lambda i, j : D[i, j] + E[i, j], "F")
+        return F
+
+    target = hcl.Platform.xilinx_zc706
+    target.config(compiler="vivado_hls", mode="csyn", project="stages-tvm.prj")
+    s = hcl.create_schedule([A, C], kernel)
+    s.to(kernel.B, s[kernel.D])
+    s.to(kernel.D, s[kernel.F])
+    s.to(kernel.E, s[kernel.F])
+    mod = hcl.build(s, target=target)
+    np_A = np.zeros((32, 32))
+    np_C = np.zeros((32, 32))
+    np_F = np.zeros((32, 32))
+    hcl_A = hcl.asarray(np_A)
+    hcl_C = hcl.asarray(np_C)
+    hcl_F = hcl.asarray(np_F)
+    mod(hcl_A, hcl_C, hcl_F)
+    return mod.report()
+
 def refine(res_tbl):
     lst = res_tbl.split("\n")
     pattern = re.compile(r'\s\s+') 
@@ -119,11 +147,33 @@ def get_rpt(config):
         with open(xml_file, "r") as xml:
             profile = xmltodict.parse(xml.read())["profile"]
         clock_unit = profile["PerformanceEstimates"]["SummaryOfOverallLatency"]["unit"]
-        summary = profile["PerformanceEstimates"]["SummaryOfLoopLatency"]
-
+
         rpt = hcl.report.Displayer(clock_unit)
-        rpt.init_table(summary)
-        rpt.collect_data(summary)
+        try:
+            summary = profile["PerformanceEstimates"]["SummaryOfLoopLatency"]
+            rpt.init_table(summary)
+            rpt.collect_data(summary)
+        except:
+            pass
+
+        path = config['algorithm']['data_path']
+        if path:
+            other_rpt = []
+            path = str(pathlib.Path(__file__).parent.absolute()) + path
+            for file in os.listdir(path):
+                if file.endswith("_csynth.xml") and file != "test_csynth.xml":
+                    fpath = os.path.join(path, file)
+                    other_rpt.append(fpath)
+
+            for files in other_rpt:
+                with open(files, "r") as xml:
+                    profile = xmltodict.parse(xml.read())["profile"]
+                summary = profile["PerformanceEstimates"]["SummaryOfLoopLatency"]
+                rpt_inner = hcl.report.Displayer(clock_unit)
+                rpt_inner.init_table(summary)
+                rpt_inner.collect_data(summary)
+                res = vars(rpt_inner)
+                rpt.add_fields(res) 
     return rpt
 
 def _test_rpt(config):
@@ -212,6 +262,7 @@ def test_knn_digitrec(vhls):
         'has_algorithm' : 1,
         'algorithm' : {
             'report_path' : '/test_report_data/digitrec_report.xml',
+            'data_path' : '',
             'name' : 'knn_digitrec'
         },
         'get_max' : 'Latency',
@@ -253,6 +304,7 @@ def test_kmeans(vhls):
         'has_algorithm' : 1,
         'algorithm' : {
             'report_path' : '/test_report_data/kmeans_report.xml',
+            'data_path' : '',
             'name' : 'kmeans'
         },
         'get_max' : 'Absolute Time Latency',
@@ -294,6 +346,7 @@ def test_sobel(vhls):
         'has_algorithm' : 0,
         'algorithm' : {
             'report_path' : '/test_report_data/sobel_report.xml',
+            'data_path' : '',
             'name' : 'sobel'
         },
         'get_max' : 'Latency',
@@ -336,6 +389,7 @@ def test_sobel_partial(vhls):
         'has_algorithm' : 0,
         'algorithm' : {
             'report_path' : '/test_report_data/sobel_report_partial.xml',
+            'data_path' : '',
             'name' : 'sobel_partial'
         },
         'get_max' : 'Latency',
@@ -378,6 +432,7 @@ def test_canny(vhls):
         'has_algorithm' : 0,
         'algorithm' : {
             'report_path' : '/test_report_data/canny_report.xml',
+            'data_path' : '',
             'name' : 'canny'
         },
         'get_max' : 'Max Latency',
@@ -421,6 +476,7 @@ def test_spam_filter(vhls):
         'has_algorithm' : 0,
         'algorithm' : {
             'report_path' : '/test_report_data/spam_filter_report.xml',
+            'data_path' : '',
             'name' : 'spam_filter'
         },
         'get_max' : 'Latency',
@@ -457,10 +513,56 @@ def test_spam_filter(vhls):
     }
     _test_rpt(config)
 
+def test_multi_rpt(vhls):
+    # `report_path` can also be switched into 
+    # `/test_report_data/multi_report.xml`
+    config = {
+        'vhls' : vhls,
+        'has_algorithm' : 0,
+        'algorithm' : {
+            'report_path' : '/test_report_data/stages_report/test_csynth.xml',
+            'data_path' : '/test_report_data/stages_report/',
+            'name' : 'stages'
+        },
+        'get_max' : 'Latency',
+        'col' : 'Category',
+        'info' : 'NoQuery',
+        'loop_query' : {
+            'query' : ['B', 'F'],
+            'name' : 'LoopQuery'
+        },
+        'column_query' : {
+            'query' : ['Trip Count', 'Latency', 'Iteration Latency', 
+                        'Pipeline II', 'Pipeline Depth'],
+            'name' : 'ColumnQuery'
+        },
+        'level_query' : {
+            'val' : 0,
+            'name' : 'LevelQuery'
+        },
+        'level_out_of_bound' : {
+            'val' : [5, -2],
+            'name' : 'LevelQueryOOB'
+        },
+        'multi_query' : {
+            'row_query' : ['D'],
+            'level_query' : 1,
+            'name' : 'MultiQuery'
+        },
+        'all_query' : {
+            'row_query' : ['B', 'E'],
+            'col_query' : ['Latency'],
+            'level_query' : 0,
+            'name' : 'AllQuery'
+        }
+    }
+    _test_rpt(config)
+
 if __name__ == '__main__':
     test_knn_digitrec(False)
     test_kmeans(False)
     test_sobel(False)
     test_sobel_partial(False)
     test_canny(False)
     test_spam_filter(False) 
+    test_multi_rpt(False)

tests/test_report_data/expected.json

@@ -64,5 +64,16 @@
     "LevelQueryOOB" : ["Trip Count, Latency Iteration Latency Pipeline II Pipeline Depth", "outer_loop_x_outer_loop_y, 22500, 92610000, 4116, N/A, N/A", "+ data_local_x_outer_data_local_x_inner, 1024, 1026, N/A, 1, 4", "+ dot_product_loop_x_outer1_dot_product_loop_x_inner1, 1024, 1027, N/A, 1, 5", "+ grad_x_outer2_grad_x_inner2, 1024, 1026, N/A, 1, 4", "+ update_param_loop_x_outer3_update_param_loop_x_inner3, 1024, 1026, N/A, 1, 4"],
     "MultiQuery" : ["Trip Count Latency Iteration Latency Pipeline II Pipeline Depth", "+ update_param_loop_x_outer3_update_param_loop_x_inner3, 1024, 1026, N/A, 1, 4"],
     "AllQuery" : ["Latency", "+ dot_product_loop_x_outer1_dot_product_loop_x_inner1, 1027"]
+  },
+  "stages" : {
+    "GetMax" : {"E_i2": {"3136": 0}, "B_i": {"3136": 0}, "F_i3": {"2112": 0}, "D_i1": {"2112": 0}, "E_j2": {"96": 1}, "B_j": {"96": 1}, "F_j3": {"64": 1}, "D_j1": {"64": 1}},
+    "Category" : "Trip Count Latency Iteration Latency Pipeline II Pipeline Depth",
+    "NoQuery" : ["Trip Count Latency Iteration Latency Pipeline II Pipeline Depth", "B_i, 32, 3136, 98, N/A, N/A", "+ B_j, 32, 96, 3, N/A, N/A", "D_i1, 32, 2112, 66, N/A, N/A", "+ D_j1, 32, 64, 2, N/A, N/A", "E_i2, 32, 3136, 98, N/A, N/A", "+ E_j2, 32, 96, 3, N/A, N/A", "F_i3, 32, 2112, 66, N/A, N/A", "+ F_j3, 32, 64, 2, N/A, N/A"],
+    "LoopQuery" : ["Trip Count Latency Iteration Latency Pipeline II Pipeline Depth", "B_i, 32, 3136, 98, N/A, N/A", "+ B_j, 32, 96, 3, N/A, N/A", "F_i3, 32, 2112, 66, N/A, N/A", "+ F_j3, 32, 64, 2, N/A, N/A"],
+    "ColumnQuery" : ["Trip Count Latency Iteration Latency Pipeline II Pipeline Depth", "B_i, 32, 3136, 98, N/A, N/A", "+ B_j, 32, 96, 3, N/A, N/A", "D_i1, 32, 2112, 66, N/A, N/A", "+ D_j1, 32, 64, 2, N/A, N/A", "E_i2, 32, 3136, 98, N/A, N/A", "+ E_j2, 32, 96, 3, N/A, N/A", "F_i3, 32, 2112, 66, N/A, N/A", "+ F_j3, 32, 64, 2, N/A, N/A"],
+    "LevelQuery" : ["Trip Count Latency Iteration Latency Pipeline II Pipeline Depth", "B_i, 32, 3136, 98, N/A, N/A", "D_i1, 32, 2112, 66, N/A, N/A", "E_i2, 32, 3136, 98, N/A, N/A", "F_i3, 32, 2112, 66, N/A, N/A"],
+    "LevelQueryOOB" : ["Trip Count Latency Iteration Latency Pipeline II Pipeline Depth", "B_i, 32, 3136, 98, N/A, N/A", "+ B_j, 32, 96, 3, N/A, N/A", "D_i1, 32, 2112, 66, N/A, N/A", "+ D_j1, 32, 64, 2, N/A, N/A", "E_i2, 32, 3136, 98, N/A, N/A", "+ E_j2, 32, 96, 3, N/A, N/A", "F_i3, 32, 2112, 66, N/A, N/A", "+ F_j3, 32, 64, 2, N/A, N/A"],
+    "MultiQuery" : ["Trip Count Latency Iteration Latency Pipeline II Pipeline Depth", "D_i1, 32, 2112, 66, N/A, N/A", "+ D_j1, 32, 64, 2, N/A, N/A"],
+    "AllQuery" : ["Latency", "B_i, 3136", "E_i2, 3136"]
   }
 }