trafodion-codereview mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From DaveBirdsall <...@git.apache.org>
Subject [GitHub] incubator-trafodion pull request #750: TRAFODION-2265
Date Mon, 10 Oct 2016 17:02:29 GMT
Github user DaveBirdsall commented on a diff in the pull request:

    https://github.com/apache/incubator-trafodion/pull/750#discussion_r82643315
  
    --- Diff: core/sql/exp/exp_fixup.cpp ---
    @@ -1740,6 +1741,103 @@ Lng32 ex_conv_clause::findIndexIntoInstrArray(ConvInstruction
ci)
       return -1; // not found
     }
     
    +bool   ex_conv_clause::sv_instrOffsetIndexPopulated = false;
    +short  ex_conv_clause::sv_MaxOpTypeValue = -1;
    +int   *ex_conv_clause::sv_convIndexSparse = 0;
    +
    +/* 
    + * Scans convInstrInfo[] and populates a 
    + * sparsely populated index: sv_convIndexSparse. 
    + * 
    + * For a given op type: op_type, 
    + * if (sv_convIndexSparse[op_type>] != -1) then 
    + * it contains the offset into the first row of 
    + * convInstrInfo[]
    + */
    +void ex_conv_clause::populateInstrOffsetIndex()
    +{
    +  if (sv_instrOffsetIndexPopulated) {
    +    return;
    +  }
    +
    +  typedef struct convInstrIndexStruct {
    +    short type_op1;
    +    short offset;
    +  } convInstrIndexDef;
    +  
    +  short lv_MaxIndex = -1;
    +
    +  // Currently, the number of distinct 'type_op1' values in the convInstrInf[] is 32
    +  const int lv_MaxIndexCapacity = 100;
    +  convInstrIndexDef lv_convIndex[lv_MaxIndexCapacity];
    +
    +  Int32 i = 0;
    +  Int32 lv_source_type = -1;
    +  Int32 lv_max_array_size = sizeof(ex_conv_clause::convInstrInfo) / sizeof(ex_conv_clause::ConvInstrStruct);
    +  
    +  // collect distinct type_op1 values and their 1st offset in convInstrInfo[]
    +  while (i < lv_max_array_size) {
    +    if (ex_conv_clause::convInstrInfo[i].type_op1 != lv_source_type) {
    +      lv_source_type = ex_conv_clause::convInstrInfo[i].type_op1;
    +      if ((lv_source_type > -1) && 
    +	  (lv_MaxIndex < (lv_MaxIndexCapacity - 1))) {
    +	lv_convIndex[++lv_MaxIndex].type_op1 = lv_source_type;
    +	lv_convIndex[lv_MaxIndex].offset = i;
    +      }
    +    }
    +    i++;
    +  }
    +
    +  // Find the max op type value in lv_convIndex
    +  int j = 0;
    +  while (j <= lv_MaxIndex) {
    +    if (sv_MaxOpTypeValue < lv_convIndex[j].type_op1) {
    +      sv_MaxOpTypeValue = lv_convIndex[j].type_op1;
    +    }
    +    j++;
    +  }
    +
    +  // just in case some op type has been assigned a pretty large value
    +  // we dont want to allocate a very large sparse value
    +  if (sv_MaxOpTypeValue > 8000) {
    +    sv_MaxOpTypeValue = 8000;
    +  }
    +  
    +  // Allocate a sparsely populated array
    +  sv_convIndexSparse = (int *) calloc(sizeof(int), 
    --- End diff --
    
    I think this logic is idempotent to it probably doesn't matter but it looks like two threads
could do the calloc, one of them will win in setting sv_convIndexSparse, then both of them
could interleave in the initialization logic below. Could be the fast one sets the sv_instrOffsetIndexPopulated
flag to true while the slow one is in the Initialize to -1 loop so theoretically we could
get wrong answers. Might be safer to use a local variable here and only set sv_convIndexSparse
at the very end just before setting the boolean. Then the worst thing that happens is we allocated
the array twice but it will never be inconsistent.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

Mime
View raw message