DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH] ip_pipeline: add scripts file for pipeline to core mappings
@ 2016-05-06 18:32 Jasvinder Singh
  2016-05-06 19:04 ` [dpdk-dev] [PATCH v2] ip_pipeline: add script " Jasvinder Singh
  0 siblings, 1 reply; 6+ messages in thread
From: Jasvinder Singh @ 2016-05-06 18:32 UTC (permalink / raw)
  To: dev; +Cc: cristian.dumitrescu, Guruprasad Mukundarao

From: Guruprasad Mukundarao <guruprasadx.rao@intel.com>

This script parses the application configuration file and detects all the
pipelines specified therein, and then, it generates all the possible mappings
of those pipelines on the specified CPU core-list.

As a result, each of the possible pipeline-to-core mappings is saved as
separate output configuration file. For example- if input file is
edge_router_downstream.cfg with 3 pipeline (excluding pipeline 0) and
core-list is "1, 2", following combinations will be generated-

Pipeline 1      Pipeline 2       Pipeline 3
Core = 1        Core = 1         Core = 2
Core = 1        Core = 2         Core = 1
Core = 2        Core = 1         Core = 1
Core = 2        Core = 2         Core = 1
Core = 2        Core = 1         Core = 2
Core = 1        Core = 2         Core = 2
Core = C1       Core = C1H       Core = C2
Core = C1       Core = C2        Core = C1H
Core = C2       Core = C1        Core = C1H

This script will help users to analyse the performance of application by
evaluating all the generated configuration files with different
pipelines-to-core mappings and obtaining the application configuration
file with best performance.

Signed-off-by: Guruprasad Mukundarao <guruprasadx.rao@intel.com>
---
 .../ip_pipeline/config/pipeline-to-core-mapping.py | 990 +++++++++++++++++++++
 1 file changed, 990 insertions(+)
 create mode 100644 examples/ip_pipeline/config/pipeline-to-core-mapping.py

diff --git a/examples/ip_pipeline/config/pipeline-to-core-mapping.py b/examples/ip_pipeline/config/pipeline-to-core-mapping.py
new file mode 100644
index 0000000..6fdac91
--- /dev/null
+++ b/examples/ip_pipeline/config/pipeline-to-core-mapping.py
@@ -0,0 +1,990 @@
+#! /usr/bin/python3
+
+#   BSD LICENSE
+#
+#   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+#   All rights reserved.
+#
+#   Redistribution and use in source and binary forms, with or without
+#   modification, are permitted provided that the following conditions
+#   are met:
+#
+#     * Redistributions of source code must retain the above copyright
+#       notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in
+#       the documentation and/or other materials provided with the
+#       distribution.
+#     * Neither the name of Intel Corporation nor the names of its
+#       contributors may be used to endorse or promote products derived
+#       from this software without specific prior written permission.
+#
+#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#
+# This script maps the set of pipelines identified (MASTER pipelines are 
+# ignored) from the input configuration file to the set of cores 
+# provided as input argument and creates configuration files for each of 
+# the mapping combinations.
+#
+
+import sys
+import array
+import itertools
+import re
+import argparse
+import os
+from collections import namedtuple
+
+#default values
+enable_stage0_traceout = 1
+enable_stage1_traceout = 1
+enable_stage2_traceout = 1
+
+#enable_stage0_fileout = 0
+enable_stage1_fileout = 1
+enable_stage2_fileout = 1
+
+#pattern for physical core
+pattern_phycore = '^(s|S)\d(c|C)[1-9][0-9]*$'
+reg_phycore = re.compile(pattern_phycore)
+
+#----------------------------------------------------------------------------- 
+def popcount(mask):                                                                          
+    return bin(mask).count("1")
+
+#----------------------------------------------------------------------------- 
+def len2mask(length):
+    if (length == 0):
+        return 0
+
+    if (length > 64):
+        sys.exit('error: len2mask - lenght %i > 64. exiting' %length)
+    
+    return (0xFFFFFFFFFFFFFFFF >> (64 - length))
+
+#----------------------------------------------------------------------------- 
+def bitstring_write(n, n_bits):
+    tmpstr = ""
+    if (n_bits > 64):
+        return
+
+    i = n_bits - 1
+    while (i >= 0):
+        cond = (n & (1 << i))
+        if (cond):
+            print('1', end='')
+            tmpstr += '1'
+        else:
+            print('0', end='')
+            tmpstr += '0'
+        i -= 1
+    #end while
+    return tmpstr
+#end function
+
+#------------------------------------------------------------------------- 
+Constants = namedtuple('Constants', ['MAX_CORES', 'MAX_PIPELINES'])
+constants = Constants(16, 64)
+
+
+#------------------------------------------------------------------------- 
+class Cores0:
+    def __init__(self):
+        self.n_pipelines = 0
+
+class Cores1:
+    def __init__(self):
+        self.pipelines = 0
+        self.n_pipelines = 0
+
+class Cores2:
+    def __init__(self):
+        self.pipelines = 0
+        self.n_pipelines = 0
+        self.counter = 0
+        self.counter_max = 0
+        self.bitpos = array.array("L", itertools.repeat(0,constants.MAX_PIPELINES))
+
+
+#--------------------------------------------------------------------
+class Context0:
+    def __init__(self):
+        self.cores = [Cores0() for i in range(0, constants.MAX_CORES)]
+        self.n_cores = 0
+        self.n_pipelines = 0
+        self.n_pipelines0 = 0
+        self.pos = 0
+        self.file_comment = ""
+        self.c1 = None
+        self.c2 = None
+
+    #-------------------------------------------------------------
+    def stage0_print(self):
+        print('printing Context0 obj')
+        print('c0.cores(n_pipelines) = [ ', end='')
+        for cores_count in range(0, constants.MAX_CORES):
+            print(self.cores[cores_count].n_pipelines, end=' ')
+        print(']')
+        print('c0.n_cores = %d' %self.n_cores)
+        print('c0.n_pipelines = %d' %self.n_pipelines)
+        print('c0.n_pipelines0 = %d' %self.n_pipelines0)
+        print('c0.pos = %d' %self.pos)
+        print('c0.file_comment = %s' %self.file_comment)
+        if (self.c1 is not None):
+            print('c0.c1 = ', end='')
+            print(repr(self.c1))
+        else:
+            print('c0.c1 = None')
+
+        if (self.c2 is not None):
+            print('c0.c2 = ', end='')
+            print(repr(self.c2))
+        else:
+            print('c0.c2 = None')
+
+                    
+    #-------------------------------------------------------------
+    def stage0_init(self, num_cores, num_pipelines, c1, c2):
+        self.n_cores = num_cores
+        self.n_pipelines = num_pipelines
+        self.c1 = c1
+        self.c2 = c2
+
+    #-------------------------------------------------------------
+    def stage0_process(self):
+        #print('inside stage0_process')
+
+        # stage0 init
+        self.cores[0].n_pipelines = self.n_pipelines
+        self.n_pipelines0 = 0
+        self.pos = 1
+
+        while True:
+            #go forward
+            while True:
+                if ((self.pos < self.n_cores) and (self.n_pipelines0 > 0)):
+                    self.cores[self.pos].n_pipelines = min(self.cores[self.pos - 1].n_pipelines, self.n_pipelines0)
+                    self.n_pipelines0 -= self.cores[self.pos].n_pipelines
+                
+                    self.pos += 1
+                else:
+                    break
+
+            #end while
+            
+            # check solution
+            if (self.n_pipelines0 == 0):
+                self.stage0_log()
+                self.c1.stage1_init(self, self.c2) # self is object c0
+                self.c1.stage1_process()
+        
+            # go backward
+            while True:
+                if (self.pos == 0):
+                    return
+
+                self.pos -= 1
+                if ((self.cores[self.pos].n_pipelines >1) and
+                        (self.pos != (self.n_cores -1))):
+                    break
+                
+                self.n_pipelines0 += self.cores[self.pos].n_pipelines
+                self.cores[self.pos].n_pipelines = 0
+            #end while
+
+            # rearm
+            self.cores[self.pos].n_pipelines -= 1
+            self.n_pipelines0 += 1
+            self.pos += 1
+        #end while         
+    #end function    
+
+
+    #-------------------------------------------------------------
+    def stage0_log(self):
+        tmp_file_comment = ""
+        if(enable_stage0_traceout != 1):
+            return
+
+        print('STAGE0: ', end='')
+        tmp_file_comment += 'STAGE0: '
+        for cores_count in range(0, self.n_cores):
+            print('C%d = %d\t' \
+            %(cores_count, self.cores[cores_count].n_pipelines), end='')
+            tmp_file_comment += "C{} = {}\t".format(cores_count, \
+                    self.cores[cores_count].n_pipelines)
+        #end for
+        print('')
+        self.c1.stage0_file_comment = tmp_file_comment
+        self.c2.stage0_file_comment = tmp_file_comment
+
+    # end function
+
+#end class Context0
+
+
+#-------------------------------------------------------------
+class Context1:
+    #class attribute
+    _fileTrace = None
+
+    def __init__(self):
+        self.cores = [Cores1() for i in range(constants.MAX_CORES)]
+        self.n_cores = 0
+        self.n_pipelines = 0
+        self.pos = 0
+        self.stage0_file_comment = ""
+        self.stage1_file_comment = ""
+
+        self.c2 = None
+        self.arr_pipelines2cores = []
+    #end init
+
+    #-------------------------------------------------------------
+    def stage1_reset(self):
+        for i in range(constants.MAX_CORES):
+            self.cores[i].pipelines = 0
+            self.cores[i].n_pipelines = 0
+        
+        self.n_cores = 0
+        self.n_pipelines = 0
+        self.pos = 0
+        self.c2 = None
+        self.arr_pipelines2cores.clear()
+    #end def stage1_reset
+
+    #-------------------------------------------------------------
+    def stage1_print(self):
+        print('printing Context1 obj')
+        print('c1.cores(pipelines,n_pipelines) = [ ', end='')
+        for cores_count in range(0, constants.MAX_CORES):
+            print('(%d,%d)' %(self.cores[cores_count].pipelines, 
+                              self.cores[cores_count].n_pipelines), end=' ')
+        print(']')
+        print('c1.n_cores = %d' %self.n_cores)
+        print('c1.n_pipelines = %d' %self.n_pipelines)
+        print('c1.pos = %d' %self.pos)
+        print('c1.stage0_file_comment = %s' %self.stage0_file_comment)
+        print('c1.stage1_file_comment = %s' %self.stage1_file_comment)
+        if (self.c2 is not None):
+            print('c1.c2 = ', end='')
+            print(self.c2)
+        else:
+            print('c1.c2 = None')
+    #end stage1_print
+
+    #-------------------------------------------------------------
+    def stage1_init(self, c0, c2):
+        self.stage1_reset()  
+        self.n_cores = 0
+        while (c0.cores[self.n_cores].n_pipelines > 0):
+            self.n_cores += 1
+        
+        self.n_pipelines = c0.n_pipelines
+        self.c2 = c2
+
+        self.arr_pipelines2cores = [0] * self.n_pipelines
+
+        i = 0
+        while (i < self.n_cores):
+            self.cores[i].n_pipelines = c0.cores[i].n_pipelines
+            i += 1
+        #end while
+    #end stage1_init
+    #-------------------------------------------------------------
+    def stage1_process(self):
+        pipelines_max = len2mask(self.n_pipelines)
+
+        while True:
+            pos = 0
+            overlap = 0
+
+            if (self.cores[self.pos].pipelines == pipelines_max):
+                if (self.pos == 0):
+                    return
+
+                self.cores[self.pos].pipelines = 0
+                self.pos -= 1
+                continue
+            #end if
+
+            self.cores[self.pos].pipelines += 1
+            
+            if (popcount(self.cores[self.pos].pipelines) \
+                            != self.cores[self.pos].n_pipelines):
+                continue
+
+            overlap = 0
+            pos = 0
+            while (pos < self.pos):
+                if ((self.cores[self.pos].pipelines) & \
+                        (self.cores[pos].pipelines)):
+                    overlap = 1
+                    break
+                pos += 1
+            #end while
+
+            if (overlap):
+                continue
+
+            
+            if ((self.pos > 0) and \
+               ((self.cores[self.pos].n_pipelines) == (self.cores[self.pos - 1].n_pipelines)) and \
+               ((self.cores[self.pos].pipelines) < (self.cores[self.pos - 1].pipelines))):
+                continue
+
+            if (self.pos == self.n_cores - 1):
+                self.stage1_log()
+                self.c2.stage2_init(self)
+                self.c2.stage2_process()
+
+                if (self.pos == 0):
+                    return
+
+                self.cores[self.pos].pipelines = 0
+                self.pos -= 1
+                continue
+            #endif
+
+            self.pos += 1
+        #end for
+    #end stage1_process
+            
+            
+    #-------------------------------------------------------------
+    def stage1_log(self):
+        tmp_file_comment = ""
+        if(enable_stage1_traceout == 1):
+            print('STAGE1: ', end = '')
+            tmp_file_comment += 'STAGE1: '
+            i = 0
+            while (i < self.n_cores):
+                print('C%d = [' %i, end='')
+                tmp_file_comment += "C{} = [".format(i)
+
+                j = self.n_pipelines - 1
+                while (j >= 0):
+                    cond = ((self.cores[i].pipelines) & (1 << j))
+                    if (cond):
+                        print('1', end='')
+                        tmp_file_comment += '1'
+                    else:
+                        print('0', end='')
+                        tmp_file_comment += '0'
+                    j -= 1
+            
+                print(']\t', end='')
+                tmp_file_comment += ']\t'
+                i += 1
+            #end while
+            print('\n', end ='')
+            #tmp_file_comment += '\n'
+            self.stage1_file_comment = tmp_file_comment
+            self.c2.stage1_file_comment = tmp_file_comment
+        #endif
+
+        #check if file traceing is enabled         
+        if(enable_stage1_fileout != 1):
+            return
+
+        #spit out the combination to file
+        self.stage1_process_file()
+    #end function stage1_log
+
+
+    #------------------------------------------------------------------------ 
+    def stage1_updateCoresInBuf(self, nPipeline, sCore):
+
+        rePipeline = self._fileTrace.arr_pipelines[nPipeline]
+        rePipeline = rePipeline.replace("[","\[").replace("]","\]")
+        reCore = 'core\s*=\s*((\d*)|(((s|S)\d)?(c|C)[1-9][0-9]*)).*\n'
+        sSubs = 'core = ' + sCore + '\n'
+
+        reg_pipeline = re.compile(rePipeline)
+        search_match = reg_pipeline.search(self._fileTrace.in_buf)
+        #debug
+        #print(search_match)
+
+        if(search_match):
+            pos = search_match.start()
+            substr1 = self._fileTrace.in_buf[:pos]
+            substr2 = self._fileTrace.in_buf[pos:]
+            substr2 = re.sub(reCore, sSubs, substr2,1)
+            self._fileTrace.in_buf = substr1 + substr2
+        #endif
+    #end function stage1_updateCoresInBuf
+
+    #------------------------------------------------------------------------ 
+    def stage1_process_file(self):
+        outFileName = os.path.join(self._fileTrace.out_path, \
+                self._fileTrace.prefix_outfile)
+        
+        i = 0 #represents core number
+        while (i < self.n_cores):
+            outFileName += '_'
+            outFileName += str(self.cores[i].pipelines)
+
+            j = self.n_pipelines - 1
+            pipeline_idx = 0
+            while(j >= 0):
+                cond = ((self.cores[i].pipelines) & (1 << j))
+                if (cond):
+                    #update the pipelines array to match the core
+                    # only in case of cond match
+                    self.arr_pipelines2cores[pipeline_idx] = fileTrace.in_physical_cores[i]
+                #endif
+                j -= 1
+                pipeline_idx += 1
+            #end while
+            i += 1
+        #end while
+
+        # update the in_buf as per the arr_pipelines2cores      
+        for pipeline_idx in range(len(self.arr_pipelines2cores)):
+            self.stage1_updateCoresInBuf(pipeline_idx,self.arr_pipelines2cores[pipeline_idx])
+
+        
+        #by now the in_buf is all set to be written to file 
+        outFileName += self._fileTrace.suffix_outfile
+        outputFile = open(outFileName, "w")
+        
+        #write out the comments 
+        outputFile.write("; =============== Pipeline-to-Core Mapping ================\n")
+        outputFile.write("; Generated from file {}\n".format(self._fileTrace.in_file_namepath))
+        outputFile.write("; Input pipelines = {}\n; Input cores = {}\n" \
+                .format(fileTrace.arr_pipelines, fileTrace.in_physical_cores))
+
+        strTruncated = ("", "(Truncated)") [self._fileTrace.ncores_truncated]
+        outputFile.write("; N_PIPELINES = {} N_CORES = {} {}\n"\
+                .format(self._fileTrace.n_pipelines, self._fileTrace.n_cores, strTruncated))
+
+        outputFile.write("; {}\n".format(self.stage0_file_comment)) #stage0 comment
+        outputFile.write("; {}\n".format(self.stage1_file_comment)) #stage1 comment
+        #debugging
+        #outputFile.write("; <<<<printing stage1 arr_pipelines2cores>>>>")
+        #outputFile.write("; stage1_arr_pipelines2cores = {}".format(self.arr_pipelines2cores))
+        outputFile.write("; ========================================================\n")
+        outputFile.write(";\n")
+
+        #
+        outputFile.write(self._fileTrace.in_buf)
+        outputFile.flush()
+        outputFile.close()
+    #end function stage1_process_file
+# end class Context1
+
+
+#----------------------------------------------------------------------------- 
+class Context2:
+    #class attribute
+    _fileTrace = None
+
+    def __init__(self):
+        self.cores = [Cores2() for i in range(constants.MAX_CORES)]
+        self.n_cores = 0
+        self.n_pipelines = 0
+        self.pos = 0
+        self.stage0_file_comment = ""
+        self.stage1_file_comment = ""
+        self.stage2_file_comment = ""
+
+        #each array entry is a pipeline mapped to core stored as string
+        # pipeline ranging from 1 to n, however stored in zero based array
+        self.arr2_pipelines2cores = []
+
+    #-------------------------------------------------------------
+    def stage2_print(self):
+        print('printing Context2 obj')
+        print('c2.cores(pipelines, n_pipelines, counter, counter_max) =')
+        for cores_count in range(0, constants.MAX_CORES):
+            print('core[%d] = (%d,%d,%d,%d)' %(cores_count,
+                                    self.cores[cores_count].pipelines, \
+                                    self.cores[cores_count].n_pipelines, \
+                                    self.cores[cores_count].counter, \
+                                    self.cores[cores_count].counter_max))
+
+            print('c2.n_cores = %d' %self.n_cores, end='')
+            print('c2.n_pipelines = %d' %self.n_pipelines, end='')
+            print('c2.pos = %d' %self.pos)
+            print('c2.stage0_file_comment = %s' %self.self.stage0_file_comment)
+            print('c2.stage1_file_comment = %s' %self.self.stage1_file_comment)
+            print('c2.stage2_file_comment = %s' %self.self.stage2_file_comment)
+        #end for
+    #end function stage2_print
+            
+
+    #-------------------------------------------------------------
+    def stage2_reset(self):
+        for i in range(0, constants.MAX_CORES):
+            self.cores[i].pipelines = 0
+            self.cores[i].n_pipelines = 0;
+            self.cores[i].counter = 0;
+            self.cores[i].counter_max = 0
+            
+            for idx in range(0, constants.MAX_PIPELINES):
+                self.cores[i].bitpos[idx] = 0
+
+        self.n_cores = 0
+        self.n_pipelines = 0
+        self.pos = 0
+
+        self.arr2_pipelines2cores.clear() 
+    #end stage2_reset
+
+    #-------------------------------------------------------------
+    def bitpos_load(self, coreidx):
+        i = j = 0
+        while (i < self.n_pipelines):
+            if ((self.cores[coreidx].pipelines) & \
+                 (1 << i)):
+                self.cores[coreidx].bitpos[j] = i
+                j += 1
+            i += 1
+        self.cores[coreidx].n_pipelines = j
+
+
+    #-------------------------------------------------------------
+    def bitpos_apply(self, in_buf, pos, n_pos):
+        out = 0
+        for i in range(0, n_pos):
+            out |= (in_buf & (1 << i)) << (pos[i] - i)
+
+        return out
+
+
+    #-------------------------------------------------------------
+    def stage2_init(self, c1):
+        self.stage2_reset()
+        self.n_cores = c1.n_cores
+        self.n_pipelines = c1.n_pipelines
+
+        self.arr2_pipelines2cores = [''] * self.n_pipelines
+
+        core_idx = 0
+        while (core_idx < self.n_cores):
+            self.cores[core_idx].pipelines = c1.cores[core_idx].pipelines
+
+            self.bitpos_load(core_idx)
+            core_idx += 1
+        #end while 
+    #end function stage2_init
+
+
+    #-------------------------------------------------------------
+    def stage2_log(self):
+        tmp_file_comment = ""
+        if(enable_stage2_traceout == 1):
+            print('STAGE2: ', end='')
+            tmp_file_comment += 'STAGE2: '
+
+            for i in range(0, self.n_cores):
+                mask = len2mask(self.cores[i].n_pipelines)
+                pipelines_ht0 = self.bitpos_apply((~self.cores[i].counter) & mask, \
+                                self.cores[i].bitpos, \
+                                self.cores[i].n_pipelines)
+
+                pipelines_ht1 = self.bitpos_apply(self.cores[i].counter, \
+                                self.cores[i].bitpos, \
+                                self.cores[i].n_pipelines)
+
+                print('C%dHT0 = [' %i, end='')
+                tmp_file_comment += "C{}HT0 = [".format(i)
+                tmp_file_comment += bitstring_write(pipelines_ht0, self.n_pipelines)
+
+                print(']\tC%dHT1 = [' %i, end='')
+                tmp_file_comment += "]\tC{}HT1 = [".format(i)
+                tmp_file_comment += bitstring_write(pipelines_ht1, self.n_pipelines)
+                print(']\t', end='')
+                tmp_file_comment += ']\t'
+            #end for
+            print('')
+            self.stage2_file_comment = tmp_file_comment
+        #endif
+
+        #check if file traceing is enabled         
+        if(enable_stage2_fileout != 1):
+            return 
+        #spit out the combination to file
+        self.stage2_process_file()
+
+    #end function stage2_log
+
+    #-------------------------------------------------------------
+    def stage2_updateCoresInBuf(self, nPipeline, sCore):
+        rePipeline = self._fileTrace.arr_pipelines[nPipeline]
+        rePipeline = rePipeline.replace("[","\[").replace("]","\]")
+        reCore = 'core\s*=\s*((\d*)|(((s|S)\d)?(c|C)[1-9][0-9]*)).*\n'
+        sSubs = 'core = ' + sCore + '\n'
+        #sSubs = 'core = ' + self._fileTrace.in_physical_cores[sCore] + '\n'
+
+        reg_pipeline = re.compile(rePipeline)
+        search_match = reg_pipeline.search(self._fileTrace.in_buf)
+
+        if(search_match):
+            pos = search_match.start()
+            substr1 = self._fileTrace.in_buf[:pos]
+            substr2 = self._fileTrace.in_buf[pos:]
+            substr2 = re.sub(reCore, sSubs, substr2,1)
+            self._fileTrace.in_buf = substr1 + substr2
+        #endif
+    #end function stage2_updateCoresInBuf
+
+
+    #-------------------------------------------------------------
+    def pipelines2cores(self, n, n_bits, nCore, bHT):
+        if (n_bits > 64):
+            return
+
+        i = n_bits - 1
+        pipeline_idx = 0
+        while (i >= 0):
+            cond = (n & (1 << i))
+            if (cond):
+                #update the pipelines array to match the core
+                # only in case of cond match
+                # PIPELINE0 and core 0 are reserved
+                if(bHT):
+                    tmpCore = fileTrace.in_physical_cores[nCore] + 'h'
+                    self.arr2_pipelines2cores[pipeline_idx] = tmpCore
+                else:
+                    self.arr2_pipelines2cores[pipeline_idx] = \
+                            fileTrace.in_physical_cores[nCore]
+                #endif
+            #endif
+            i -= 1
+            pipeline_idx += 1
+        #end while
+
+    #end function pipelines2cores
+
+    #-------------------------------------------------------------
+    def stage2_process_file(self):
+        outFileName = os.path.join(self._fileTrace.out_path, \
+                self._fileTrace.prefix_outfile)
+
+        for i in range(0, self.n_cores):
+            mask = len2mask(self.cores[i].n_pipelines)
+            pipelines_ht0 = self.bitpos_apply((~self.cores[i].counter) & mask, \
+                                self.cores[i].bitpos, \
+                                self.cores[i].n_pipelines)
+
+            pipelines_ht1 = self.bitpos_apply(self.cores[i].counter, \
+                                self.cores[i].bitpos, \
+                                self.cores[i].n_pipelines)
+   
+            outFileName += '_'
+            outFileName += str(pipelines_ht0)
+            outFileName += '_'
+            outFileName += str(pipelines_ht1)
+            outFileName += 'HT'            
+
+            #update pipelines to core mapping
+            self.pipelines2cores(pipelines_ht0, self.n_pipelines, i , False)
+            self.pipelines2cores(pipelines_ht1, self.n_pipelines, i, True)
+        #end for
+
+        # update the in_buf as per the arr_pipelines2cores 
+        for pipeline_idx in range(len(self.arr2_pipelines2cores)):
+            self.stage2_updateCoresInBuf(pipeline_idx, self.arr2_pipelines2cores[pipeline_idx])
+       
+        #by now the in_buf is all set to be written to file 
+        outFileName += self._fileTrace.suffix_outfile
+        outputFile = open(outFileName, "w")
+
+        #write out the comments
+        outputFile.write("; ========= Pipeline-to-Core Mapping =====================\n")
+        outputFile.write("; Generated from file {}\n".format(self._fileTrace.in_file_namepath))
+        outputFile.write("; Input pipelines = {}\n; Input cores = {}\n" \
+                .format(fileTrace.arr_pipelines, fileTrace.in_physical_cores))
+
+        strTruncated = ("", "(Truncated)") [self._fileTrace.ncores_truncated]
+        outputFile.write("; N_PIPELINES = {} N_CORES = {} {}\n"\
+                .format(self._fileTrace.n_pipelines, self._fileTrace.n_cores, strTruncated))
+
+        outputFile.write("; {}\n".format(self.stage0_file_comment)) #stage0 comment
+        outputFile.write("; {}\n".format(self.stage1_file_comment)) #stage1 comment
+        outputFile.write("; {}\n".format(self.stage2_file_comment)) #stage2 comment
+        outputFile.write("; ========================================================\n")
+        outputFile.write(";\n")
+
+        outputFile.write(self._fileTrace.in_buf)
+        outputFile.flush()
+        outputFile.close()
+
+    #end function stage2_process_file 
+
+    #-------------------------------------------------------------
+    def stage2_process(self):
+        i = 0
+        while(i < self.n_cores):
+            self.cores[i].counter_max = len2mask(self.cores[i].n_pipelines - 1)
+            i += 1
+        #end while
+        
+        self.pos = self.n_cores - 1
+        while True:
+            if (self.pos == self.n_cores - 1):
+                self.stage2_log()
+
+            if (self.cores[self.pos].counter == self.cores[self.pos].counter_max):
+                if (self.pos == 0):
+                    return
+                #endif
+
+                self.cores[self.pos].counter = 0
+                self.pos -= 1
+                continue
+            #endif
+
+            self.cores[self.pos].counter += 1
+
+            if(self.pos < self.n_cores - 1):
+                self.pos += 1
+            #endif
+        #end while
+#end class Context2
+
+#----------------------------------------------------------------------------- 
+class FileTrace:
+    #initialize default parameters
+    def __init__(self,filenamepath):
+        self.in_file_namepath = os.path.abspath(filenamepath)
+        self.in_filename = os.path.basename(self.in_file_namepath) 
+        self.in_path = os.path.dirname(self.in_file_namepath)
+        
+        #set output folder same as input
+        self.out_path = self.in_path
+
+        filenamesplit = self.in_filename.split('.')
+        self.prefix_outfile = filenamesplit[0]
+        self.suffix_outfile = ".cfg"
+        self.in_buf = None
+        self.arr_pipelines = []  # holds the positions of search
+        
+        self.max_cores = 15 
+        self.max_pipelines = 15
+
+        self.in_physical_cores = None
+        self.hyper_thread = None
+
+        # save the num of pipelines determined from input file
+        self.n_pipelines = 0
+        # save the num of cores input (or the truncated value) 
+        self.n_cores = 0
+        self.ncores_truncated = False
+       
+    #end init
+
+    def print_TraceFile(self):
+        print("self.in_file_namepath = ", self.in_file_namepath) 
+        print("self.in_filename = ", self.in_filename)
+        print("self.in_path = ", self.in_path)
+        print("self.out_path = ", self.out_path)
+        print("self.prefix_outfile = ", self.prefix_outfile)
+        print("self.suffix_outfile = ", self.suffix_outfile)
+        print("self.in_buf = ", self.in_buf)
+        print("self.arr_pipelines =", self.arr_pipelines)
+        print("self.in_physical_cores", self.in_physical_cores)
+        print("self.hyper_thread", self.hyper_thread)
+    #end function print_TraceFile
+
+#end class FileTrace
+
+
+#-----------------------------------------------------------------------------
+# main process method
+#
+def process(n_cores, n_pipelines, fileTrace):
+    if (n_cores == 0):
+        sys.exit('N_CORES is 0, exiting')
+    #endif
+
+    if (n_pipelines == 0):
+        sys.exit('N_PIPELINES is 0, exiting')
+    #endif
+
+    if (n_cores > n_pipelines):
+        print('\nToo many cores, truncating N_CORES to N_PIPELINES')
+        n_cores = n_pipelines
+        fileTrace.ncores_truncated = True
+    #endif
+    fileTrace.n_pipelines = n_pipelines
+    fileTrace.n_cores = n_cores
+
+    strTruncated = ("", "(Truncated)") [fileTrace.ncores_truncated]
+    print("N_PIPELINES = {}, N_CORES = {} {}" \
+            .format(n_pipelines,n_cores, strTruncated))
+    print("---------------------------------------------------------------")
+
+    c0 = Context0()
+    c1 = Context1()
+    c2 = Context2()
+
+    #initialize the class variables
+    c1._fileTrace = fileTrace
+    c2._fileTrace = fileTrace
+
+    c0.stage0_init(n_cores, n_pipelines, c1, c2)
+    c0.stage0_process()
+
+#end function process
+
+
+
+#--------------------------------------------------------------------------
+def validate_core(core):
+	match = reg_phycore.match(core)
+	if(match):
+		return True
+	else:
+		return False
+        #endif
+#end function validate_core
+
+
+#--------------------------------------------------------------------------
+def validate_phycores(phy_cores):
+        #eat up whitespaces
+        phy_cores = phy_cores.strip().split(',')
+
+        #check if the core list is unique
+        if(len(phy_cores) != len(set(phy_cores))):
+                print('list of physical cores has duplicates')
+                return False
+        #endif
+
+        for core in phy_cores:
+                if(validate_core(core) != True):
+                    print('invalid physical core specified.')
+                    return None
+                else:
+                    return phy_cores
+                #endif
+        #endfor
+#end function validate_phycores
+
+#--------------------------------------------------------------------------
+def scanconfigfile(fileTrace):
+    #debug
+    #fileTrace.print_TraceFile()
+
+    # open file
+    filetoscan = open(fileTrace.in_file_namepath, 'r')
+    fileTrace.in_buf = filetoscan.read()
+
+    #reset iterator on open file
+    filetoscan.seek(0)
+
+    # scan input file for pipelines
+    # master pipelines to be ignored
+    pattern_pipeline = r'\[PIPELINE\d*\]'
+    pattern_mastertype = r'type\s*=\s*MASTER'
+
+    pending_pipeline = False
+    for line in filetoscan:
+        match_pipeline = re.search(pattern_pipeline, line)
+        match_type = re.search('type\s*=', line)
+        match_mastertype = re.search(pattern_mastertype, line)
+
+        if(match_pipeline):
+            sPipeline = line[match_pipeline.start():match_pipeline.end()]
+            #sPipeline = sPipeline.strip('[]')
+            pending_pipeline = True
+        elif(match_type):
+            # found a type definition...
+            if(match_mastertype == None):
+                # and this is not a master pipeline...
+                if(pending_pipeline == True):
+                    # add it to the list of pipelines to be mapped
+                    fileTrace.arr_pipelines.append(sPipeline)
+                    pending_pipeline = False
+            else:
+                # and this is a master pipeline...
+                # ignore the current and move on to next
+                sPipeline = ""
+                pending_pipeline = False
+            #endif
+        #endif
+    #endfor
+    filetoscan.close()
+
+    # validate if pipelines are unique
+    if(len(fileTrace.arr_pipelines) != len(set(fileTrace.arr_pipelines))):
+        sys.exit('Error: duplicate pipelines in input file')
+    #endif
+
+    num_pipelines = len(fileTrace.arr_pipelines)
+    num_cores = len(fileTrace.in_physical_cores)
+
+    #debug
+    #print(fileTrace.matches_pipeline)
+    print("-------------------Pipeline-to-core mapping--------------------")
+    print("Input pipelines = {}\nInput cores = {}" \
+            .format(fileTrace.arr_pipelines, fileTrace.in_physical_cores))
+
+    #input configuration file validations goes here
+    if (num_cores > fileTrace.max_cores):
+        sys.exit('Error: number of cores specified > max_cores (%d)' %fileTrace.max_cores)
+
+    if (num_pipelines > fileTrace.max_pipelines):
+        sys.exit('Error: number of pipelines in input cfg file > max_pipelines (%d)' %fileTrace.max_pipelines)
+
+    #call process to generate pipeline-to-core mapping, trace and log
+    process(num_cores, num_pipelines, fileTrace)
+
+#end function scanconfigfile
+
+
+#-----------------------------------------------------------------------------
+# python trick - so that our Python files can act as either reusable modules, 
+# or as standalone programs
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser(description='mappipelines')
+
+    reqNamedGrp = parser.add_argument_group('required named args')
+    reqNamedGrp.add_argument('-i', '--input-file', type=argparse.FileType('r'), help='Input config file', required=True)
+
+    #--physical-cores “<core>, <core>, …”, with <core> = s<SOCKETID>c<COREID>
+    reqNamedGrp.add_argument('-pc', '--physical-cores', type=validate_phycores, help='''Enter available CPU cores in format:\"<core>,<core>,...\"
+        where each core format: \"s<SOCKETID>c<COREID>\"
+        where SOCKETID={0..9}, COREID={1-99}''', required=True)
+
+    #add optional arguments
+    parser.add_argument('-ht', '--hyper-thread', help='enable/disable hyper threading. default is ON', default='ON', choices=['ON','OFF'])
+
+    parser.add_argument('-nO', '--no-output-file', help='disable output config file generation. Output file generation is enabled by default', action="store_true")
+
+    args = parser.parse_args()
+
+    # create object of FileTrace and initialise
+    fileTrace = FileTrace(args.input_file.name)
+    fileTrace.in_physical_cores = args.physical_cores
+    fileTrace.hyper_thread = args.hyper_thread
+
+    if(fileTrace.hyper_thread == 'OFF'):
+        print("!!!!disabling stage2 HT!!!!")
+        enable_stage2_traceout = 0
+        enable_stage2_fileout = 0
+    #endif
+
+    if(args.no_output_file == True):
+        print("!!!!disabling stage1 and stage2 fileout!!!!")
+        enable_stage1_fileout = 0
+        enable_stage2_fileout = 0
+    #endif
+
+    scanconfigfile(fileTrace)
+
+#end main
-- 
2.5.5

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2016-06-14 19:19 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-05-06 18:32 [dpdk-dev] [PATCH] ip_pipeline: add scripts file for pipeline to core mappings Jasvinder Singh
2016-05-06 19:04 ` [dpdk-dev] [PATCH v2] ip_pipeline: add script " Jasvinder Singh
2016-05-31  8:58   ` [dpdk-dev] [PATCH v3] " Jasvinder Singh
2016-05-31 12:23     ` Mcnamara, John
2016-06-10 15:55     ` [dpdk-dev] [PATCH v4] " Jasvinder Singh
2016-06-14 19:19       ` Thomas Monjalon

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).