nested loops for speed optimization
Hello. I need help to optimize for speed (in LabVIEW) code (matlab) following:
-----------------------------------
a = rand(1024, 200);
b = rand(1024, 512);
c = Zeros (200,512);
ICT
for i = 1: 200
j = 1:512
c (i, j) = (Sum ((:,i).*b(:,j))) a ABS;
end
end
Table of contents
end
-------------------------------------------------------------
Basically I have two 2d tables, a and b, one with 200 columns and rows of 1024, other columns a 512 and 1024 lines. What I have to do is multiply (dot product) each column a each column of b, the sum and take the absolute value of the result. The output, c, will be an array of 2d of size 200 x 512. Is it possible to do without the help of 2 loops of nesting? The two paintings 2d, a and b\, are produced by my material.
What is the best strategy to get the maximum speed? I have a powerful enough processor.
Thank you very much?
So I looked at this problem, said «seems familiar, but different...» ». Did a little math, encoded at top and improved the speed by a factor of 100 (with absolutely no parallelism). How about apples?
If we let A and B tables 1024 x 512 and 1024 x 200 and turn them into matrices, the problem is (mathematically) the absolute value of the transposed from A times B, where "the times" are the multiplication of matrices. I "guessed" that NEITHER would have included the Multiplication of matrices in his bag of tricks and maybe optimized, so I coded it upward, adding the calendar.
Using the first double Lynn - to multiply, sum, formulation of abs and a 'do it as a matrix' on my machine, the Array 1004 msec, the matrix method took 10 msec, as well as the results differ by less than one part in 1 million (I generated at random (0, 1) floats, and made the comparison by subtracting the two tables and checking that the absolute all the differences were less than 1E-6).
Bob Schor
Tags: NI Software
Similar Questions
-
Hello guys,.
Am a beginner, so can someone help me with this problem.
This is the sequence
Table 2D-> indexed by a table 2 x 2-> find the Max and Min difference in table 2 x 2-> result item must be compared with the participation of entry each array element 2 x 2, if its rained 1 0 otherwise -> finally replace the output array should be identical to the size of the input array.
Attachment Vi for reference. How to do this? Thanks in advance.
The average function does not work on the 2D tables, but you can use 'add elements to the table' and divide by four.
The rest of my code works for you?
(EDIT: new version of upladed.) You must be careful not to overflow U8 adding).
-
Nested loops in the XSLT handling
Hello
I want to manage nested loops for each within the transformation roadmap.
I have two collections of source and a target collection.
I have to search the value of key in source1 source2 and if the key is found, I map the value of the target, and if the key is not found, maybe I want to the key itself to target.
This is the input samples and the results expected from the transformation roadmap. Please help turn in the required format.
Entry
----------------------
------------------------
< Source1Collection >
< Group1 >
< key > < / keys >
i < value > < / value >
< / Group1 >
< Group2 >
b < Key > < / key >
II < value > < / value >
< / Group2 >
Group < 3 >
< Key > c < / keys >
III < value > < / value >
< / Group 3 >
< / Source1Collection >
---------------------------------------
< Source2Collection >
< Feild1 > z < / Feild1 >
< Feild1 > x < / Feild1 >
c < Feild1 > < / Feild1 >
< Feild1 > v < / Feild1 >
b < Feild1 > < / Feild1 >
< / Source2Collection >
----------------------------------------
Output
--------------------------------------
--------------------------------------
< TargetCollection >
< Field1 > one < / Feild1 >
b < Field1 > < / Feild1 >
III < Field1 > < / Feild1 >
< / TargetCollection >is it possible that I can keep the position of the corresponding source2Key "?" This will help me to map the corresponding value of the sourceKey2 to the target.
For something like that...
1 one 2 two 3 three You can use the following to get the value of a respective key...
Thus, when $key = '2' the expression above will return "two"...
I hope this helps...
See you soon,.
Vlad -
Simple as possible"nested loop"XML (XPATH)?
What is the best way to view the entries to an XML file with the following structure as a nested list (actually it must become a form where it is possible to control all entries).
I can out of each level (see the code), but it seems not to be possible to nest loops for.
Advice if it is directly possible with XPath or should I check the result of a loop to output in the other or it would even be better to make a query on the XML result?
Thank you, excellent!
I tried somwhere along this before, but it didn't work then. I have propably just forgot to bring the book of signs to the correct place or something ;-)
But yes, your solution is exactly what I was looking for!
Thank you
-
use OPT_ESTIMATE or CARDINALITY tip for correct estimation of nested loops
I'm using Oracle 11.2.0.3. Below the execution below plan, how can I use tip OPT_ESTIMATE or CARDINALITY to teach optimization E-lines for ID 9 (Nested Loop) 30553 instead of 6.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | Id | Operation | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers | OMem | 1Mem | Used-Mem | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | 0 | SELECT STATEMENT | | 1 | | | 4604 (100)| | 30553 |00:00:02.56 | 208K| | | | | 1 | SORT ORDER BY | | 1 | 6 | 7044 | 4604 (1)| 00:01:05 | 30553 |00:00:02.56 | 208K| 41M| 2086K| 36M (0)| |* 2 | HASH JOIN OUTER | | 1 | 6 | 7044 | 4603 (1)| 00:01:05 | 30553 |00:00:02.10 | 208K| 38M| 3120K| 39M (0)| |* 3 | HASH JOIN OUTER | | 1 | 6 | 6870 | 4599 (1)| 00:01:05 | 30553 |00:00:01.97 | 208K| 38M| 3120K| 39M (0)| |* 4 | HASH JOIN OUTER | | 1 | 6 | 6744 | 4591 (1)| 00:01:05 | 30553 |00:00:01.85 | 208K| 37M| 3121K| 39M (0)| |* 5 | HASH JOIN OUTER | | 1 | 6 | 6570 | 4584 (1)| 00:01:05 | 30553 |00:00:01.74 | 208K| 37M| 3121K| 38M (0)| |* 6 | HASH JOIN OUTER | | 1 | 6 | 6414 | 4576 (1)| 00:01:05 | 30553 |00:00:01.60 | 208K| 37M| 3121K| 38M (0)| | 7 | NESTED LOOPS | | 1 | | | | | 30553 |00:00:01.44 | 208K| | | | | 8 | NESTED LOOPS | | 1 | 6 | 6318 | 4572 (1)| 00:01:05 | 30553 |00:00:01.31 | 182K| | | | | 9 | NESTED LOOPS | | 1 | 6 | 1830 | 4568 (1)| 00:01:04 | 30553 |00:00:01.11 | 124K| | | | |* 10 | HASH JOIN | | 1 | 2069 | 270K| 2499 (1)| 00:00:35 | 30646 |00:00:00.46 | 23738 | 6539K| 2033K| 7965K (0)| |* 11 | TABLE ACCESS BY INDEX ROWID | DSCLR | 1 | 5079 | 158K| 533 (1)| 00:00:08 | 91395 |00:00:00.28 | 6460 | | | | | 12 | BITMAP CONVERSION TO ROWIDS | | 1 | | | | | 197K|00:00:00.16 | 615 | | | | | 13 | BITMAP AND | | 1 | | | | | 4 |00:00:00.14 | 615 | | | | | 14 | BITMAP OR | | 1 | | | | | 4 |00:00:00.07 | 276 | | | | | 15 | BITMAP CONVERSION FROM ROWIDS| | 1 | | | | | 2 |00:00:00.02 | 69 | | | | |* 16 | INDEX RANGE SCAN | XIF913DSCLR | 1 | 274K| | 14 (0)| 00:00:01 | 68407 |00:00:00.01 | 69 | | | | | 17 | BITMAP CONVERSION FROM ROWIDS| | 1 | | | | | 4 |00:00:00.05 | 207 | | | | |* 18 | INDEX RANGE SCAN | XIF913DSCLR | 1 | 274K| | 42 (0)| 00:00:01 | 209K|00:00:00.03 | 207 | | | | | 19 | BITMAP CONVERSION FROM ROWIDS | | 1 | | | | | 5 |00:00:00.06 | 339 | | | | |* 20 | INDEX RANGE SCAN | XIF910DSCLR | 1 | 274K| | 67 (0)| 00:00:01 | 239K|00:00:00.04 | 339 | | | | | 21 | NESTED LOOPS | | 1 | | | | | 21749 |00:00:00.10 | 17278 | | | | | 22 | NESTED LOOPS | | 1 | 3404 | 339K| 1965 (1)| 00:00:28 | 22772 |00:00:00.03 | 1246 | | | | | 23 | TABLE ACCESS FULL | TMP_RPT_BD_STATE_DATA_DWNLD | 1 | 3255 | 257K| 11 (0)| 00:00:01 | 3255 |00:00:00.01 | 27 | | | | |* 24 | INDEX RANGE SCAN | XIE2OCRN | 3255 | 7 | | 1 (0)| 00:00:01 | 22772 |00:00:00.03 | 1219 | | | | |* 25 | TABLE ACCESS BY INDEX ROWID | OCRN | 22772 | 1 | 21 | 1 (0)| 00:00:01 | 21749 |00:00:00.06 | 16032 | | | | |* 26 | VIEW PUSHED PREDICATE | | 30646 | 1 | 171 | 1 (0)| 00:00:01 | 30553 |00:00:00.63 | 101K| | | | | 27 | WINDOW BUFFER | | 30646 | 1 | 21 | 1 (0)| 00:00:01 | 46946 |00:00:00.59 | 101K| 2048 | 2048 | 2048 (0)| |* 28 | TABLE ACCESS BY INDEX ROWID | DSCLR_FLNG | 30646 | 1 | 21 | 1 (0)| 00:00:01 | 46946 |00:00:00.32 | 101K| | | | |* 29 | INDEX RANGE SCAN | XIE1DSCLR_FLNG | 30646 | 2 | | 1 (0)| 00:00:01 | 46946 |00:00:00.21 | 59862 | | | | |* 30 | INDEX RANGE SCAN | XPKH760_RGLTY_ACTN_DSCLR | 30553 | 1 | | 1 (0)| 00:00:01 | 30553 |00:00:00.18 | 57541 | | | | | 31 | TABLE ACCESS BY INDEX ROWID | H760_RGLTY_ACTN_DSCLR | 30553 | 1 | 748 | 1 (0)| 00:00:01 | 30553 |00:00:00.11 | 26218 | | | | | 32 | TABLE ACCESS FULL | DSCLR_ST_TYPE_TBL | 1 | 4 | 64 | 4 (0)| 00:00:01 | 4 |00:00:00.01 | 5 | | | | | 33 | TABLE ACCESS FULL | RGLTY_RSLTN_TYPE_TBL | 1 | 12 | 312 | 7 (0)| 00:00:01 | 12 |00:00:00.01 | 10 | | | | | 34 | TABLE ACCESS FULL | INTTR_TYPE_TBL | 1 | 14 | 406 | 7 (0)| 00:00:01 | 14 |00:00:00.01 | 10 | | | | | 35 | TABLE ACCESS FULL | SNCTN_TYPE_TBL | 1 | 15 | 315 | 7 (0)| 00:00:01 | 15 |00:00:00.01 | 10 | | | | | 36 | TABLE ACCESS FULL | PRDCT_TYPE_TBL | 1 | 25 | 725 | 4 (0)| 00:00:01 | 25 |00:00:00.01 | 5 | | | | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Query Block Name / Object Alias (identified by operation id): ------------------------------------------------------------- 1 - SEL$A8B7A3F4 11 - SEL$A8B7A3F4 / D@SEL$13 23 - SEL$A8B7A3F4 / OH@SEL$1 24 - SEL$A8B7A3F4 / O@SEL$13 25 - SEL$A8B7A3F4 / O@SEL$13 26 - SEL$13512960 / DF@SEL$13 27 - SEL$13512960 28 - SEL$13512960 / DF1@SEL$14 29 - SEL$13512960 / DF1@SEL$14 30 - SEL$A8B7A3F4 / H7@SEL$2 31 - SEL$A8B7A3F4 / H7@SEL$2 32 - SEL$A8B7A3F4 / DS@SEL$9 33 - SEL$A8B7A3F4 / RST@SEL$11 34 - SEL$A8B7A3F4 / ITT@SEL$3 35 - SEL$A8B7A3F4 / ST@SEL$5 36 - SEL$A8B7A3F4 / PT@SEL$7 Outline Data ------------- /*+ BEGIN_OUTLINE_DATA IGNORE_OPTIM_EMBEDDED_HINTS OPTIMIZER_FEATURES_ENABLE('11.2.0.3') DB_VERSION('11.2.0.3') OPT_PARAM('optimizer_index_cost_adj' 20) ALL_ROWS OUTLINE_LEAF(@"SEL$13512960") PUSH_PRED(@"SEL$A8B7A3F4" "DF"@"SEL$13" 11) OUTLINE_LEAF(@"SEL$A8B7A3F4") ELIMINATE_JOIN(@"SEL$2D47CDEA" "DTT"@"SEL$13") OUTLINE(@"SEL$14") OUTLINE(@"SEL$A8B7A3F4") ELIMINATE_JOIN(@"SEL$2D47CDEA" "DTT"@"SEL$13") OUTLINE(@"SEL$2D47CDEA") MERGE(@"SEL$3FAACB6D") OUTLINE(@"SEL$15") OUTLINE(@"SEL$3FAACB6D") MERGE(@"SEL$11") MERGE(@"SEL$A1A6E401") OUTLINE(@"SEL$12") OUTLINE(@"SEL$11") OUTLINE(@"SEL$A1A6E401") MERGE(@"SEL$9") MERGE(@"SEL$BFB1842A") OUTLINE(@"SEL$10") OUTLINE(@"SEL$9") OUTLINE(@"SEL$BFB1842A") MERGE(@"SEL$7") MERGE(@"SEL$EEDE2B8C") OUTLINE(@"SEL$8") OUTLINE(@"SEL$7") OUTLINE(@"SEL$EEDE2B8C") MERGE(@"SEL$5") MERGE(@"SEL$A6B38458") OUTLINE(@"SEL$6") OUTLINE(@"SEL$5") OUTLINE(@"SEL$A6B38458") MERGE(@"SEL$171BE69E") MERGE(@"SEL$3") OUTLINE(@"SEL$4") OUTLINE(@"SEL$171BE69E") MERGE(@"SEL$AF02BF1E") OUTLINE(@"SEL$3") OUTLINE(@"SEL$2") OUTLINE(@"SEL$AF02BF1E") MERGE(@"SEL$13") OUTLINE(@"SEL$1") OUTLINE(@"SEL$13") FULL(@"SEL$A8B7A3F4" "OH"@"SEL$1") INDEX(@"SEL$A8B7A3F4" "O"@"SEL$13" ("OCRN"."ORG_PK" "OCRN"."DSCLR_DSCLB_FL" "OCRN"."DSCLR_RPTBL_FL")) BITMAP_TREE(@"SEL$A8B7A3F4" "D"@"SEL$13" AND(OR(1 1 ("DSCLR"."FORM_TYPE_CD") 2 ("DSCLR"."FORM_TYPE_CD")) ("DSCLR"."DSCLR_TYPE_CD"))) NO_ACCESS(@"SEL$A8B7A3F4" "DF"@"SEL$13") INDEX(@"SEL$A8B7A3F4" "H7"@"SEL$2" ("H760_RGLTY_ACTN_DSCLR"."FLNG_PK" "H760_RGLTY_ACTN_DSCLR"."DSCLR_PK" "H760_RGLTY_ACTN_DSCLR"."REC_SEQ_NB")) FULL(@"SEL$A8B7A3F4" "DS"@"SEL$9") FULL(@"SEL$A8B7A3F4" "RST"@"SEL$11") FULL(@"SEL$A8B7A3F4" "ITT"@"SEL$3") FULL(@"SEL$A8B7A3F4" "ST"@"SEL$5") FULL(@"SEL$A8B7A3F4" "PT"@"SEL$7") LEADING(@"SEL$A8B7A3F4" "OH"@"SEL$1" "O"@"SEL$13" "D"@"SEL$13" "DF"@"SEL$13" "H7"@"SEL$2" "DS"@"SEL$9" "RST"@"SEL$11" "ITT"@"SEL$3" "ST"@"SEL$5" "PT"@"SEL$7") USE_NL(@"SEL$A8B7A3F4" "O"@"SEL$13") NLJ_BATCHING(@"SEL$A8B7A3F4" "O"@"SEL$13") USE_HASH(@"SEL$A8B7A3F4" "D"@"SEL$13") USE_NL(@"SEL$A8B7A3F4" "DF"@"SEL$13") USE_NL(@"SEL$A8B7A3F4" "H7"@"SEL$2") NLJ_BATCHING(@"SEL$A8B7A3F4" "H7"@"SEL$2") USE_HASH(@"SEL$A8B7A3F4" "DS"@"SEL$9") USE_HASH(@"SEL$A8B7A3F4" "RST"@"SEL$11") USE_HASH(@"SEL$A8B7A3F4" "ITT"@"SEL$3") USE_HASH(@"SEL$A8B7A3F4" "ST"@"SEL$5") USE_HASH(@"SEL$A8B7A3F4" "PT"@"SEL$7") SWAP_JOIN_INPUTS(@"SEL$A8B7A3F4" "D"@"SEL$13") INDEX_RS_ASC(@"SEL$13512960" "DF1"@"SEL$14" ("DSCLR_FLNG"."DSCLR_PK" "DSCLR_FLNG"."BLLBL_FL")) END_OUTLINE_DATA */This plan gives you an idea of why your initial synchronization has been long - most of your readings are one-piece readings and you have read 53 724 blocks. If you estimate 1 centisecond by read (which could be on a reasonable charge system) then it is 537 seconds. to 5 milliseconds per read, it's still 268 seconds or 4 minutes and 30 seconds.
A quick break down of some (the worst case) potential time - only you can say whether or not some of these readings will be always/often/never being cached in the file or at the SAN level.
Line 11 - physical reads 6 400 done - it's a tree bitmap then the readings are in the order of physical table with no. rereads. Call 5ms by read-online 32 seconds. L
| * 11 | TABLE ACCESS BY INDEX ROWID | DSCLR | 1. 5079. 158K | 533 (1) | 00:00:08 | 91395 | 00:00:01.27 | 6460. 6400 | | | |
The hash join takes us from 6 400 to 11 788 because it joins, lines 21 to 25 (nested loop) in which most of the work is the random access table with some automatic caching: 26 seconds. Do a search here rather than indexed 22 000 access? Which would be faster, which would charge less.
| * 25. TABLE ACCESS BY INDEX ROWID | OCRN | 22050. 1. 21. 1 (0) | 00:00:01 | 21059 | 00:01:04.74 | 15634 | 5280. | | |
Three nested loops that introduce more readings:
11788 9-> 26386:14 600 of the predicate view pushed which some lies but a very large index and a table of random access 73 seconds
8 26386-> 33783:7 400 index line 30 - big clue about 37 seconds Access hides
33783 7-> 53691:19 900 31 table access - large table line, not much caching, 100 seconds.
In each case, you can ask if an analysis would take less time than all of these random reads.
NOTE - all I'm pointing our here where you potentially use a lot of time so that you can ask whether or not you should do little of the query in this way, or if there is an alternative, or if the work you're doing is absolutely necessary. If you want to optimize the query you should consider the intent of the query and think your way through a reasonable join order to see if you can imagine an effective way to get the same data. I have written a few articles and presentation on an approach that begins by drawing - here is a link to an article I wrote for a group of SQL Server: https://www.simple-talk.com/sql/performance/designing-efficient-sql-a-visual-approach/
and a video I did with Kyle Hailey, while he was working for Embarcadero:Tune in for the ultimate SQL Tune-off | Landing pages
Concerning
Jonathan Lewis
-
Optimize for speed = > error runtime LV in the CCGArrSupport2.c file (LV2009 / MCB2400)
Hello
I have a MCB2400 and LV2009 and normal Code generation works fine. But if I try the Run-Time Option: "Optimize for Speed" I always this error:
"Runtime error LV in the fichier...\CCGArrSupport2.c at 2621:6 line 3.
I already tried to work around the problem, which has helped someone black fins: http://forums.ni.com/ni/board/message?board.id=420&thread.id=811
But it helps didn t in my case.
Good bye
Amin
The runtime error is rooted in constant initialization.
When you choose "Optimize for speed" or simply 'Generate C function' calls for applications with subVIs that contains the constant table initialization always use the option "allocate constants" on "first use".Other workarounds may apply as well: initialiser initialize everything in Top VI, use Array initialize instead of constant tables, etc.
-
Need help for the conversion of the nested loops to use the stream
I'm trying to teach me how to effectively use the java stream, but finds it difficult to find the correct syntax for a problem of nested loop. I would like to convert the following code to use the Java 8 stream instead of nested loops. Can anyone provide an example or a solution?
/** * Return all members of stringListOne that begin with any of the strings in * stringListTwo * * @return all members of stringListOne that begin with any of the strings in * stringListTwo */ Set<String> getFilteredStrings(Set<String> stringListOne, Set<String> stringListTwo) { Set<String> stringListOneFiltered = new HashSet<String>(); for (String next : stringListOne) { for (String nextPrefix : stringListTwo) { if (next.startsWith(nextPrefix)) { stringListOneFiltered.add(next); } } } return stringListOneFiltered; }Here's a test case for the above code:
/** * Test method for {@link scratchpad.MyClass#getFilteredStrings(java.util.Set, java.util.Set)}. */ @Test public final void testGetFilteredStrings() { MyClass uut = new MyClass(); Set<String> stringListOne = new HashSet<String>(); stringListOne.add("until"); stringListOne.add("unite"); stringListOne.add("under"); stringListOne.add("prepare"); stringListOne.add("pretend"); stringListOne.add("prefix"); stringListOne.add("prepend"); stringListOne.add("post"); stringListOne.add("positive"); stringListOne.add("postgame"); stringListOne.add("postpone"); Set<String> stringListTwo = new HashSet<String>(); stringListTwo.add("post"); stringListTwo.add("pre"); stringListTwo.add("und"); Set<String> result= uut.getFilteredStrings(stringListOne, stringListTwo); System.out.println(result); assertEquals(8, result.size()); }Thanks for any help you can provide! I get the hang of using the course of water, but this particular code was difficult.
tmbc
I found the following solution to my question:
GetFilteredStrings (Set stringListOne, the value {Set
stringListTwo) Predicate startsWith
= p-> stringListTwo.stream () .anyMatch (prefix-> p.startsWith (prefix)); Return stringListOne.stream () .filter (startsWith) .collect (Collectors.toSet ());
}
-
Create a table in a loop 'for' (reset the number of loops)
Hello
I try to incorporate a signal of acceleration from an analog input CRIO to get speed and then further integrate to get the moving of a vibration system.
I understand that to achieve integration, a table must be fed in. My problem is that I could not buffer the signal in a buffer block, as I couldn't find buffer blocks. so I decided to create my data table by using a loop "for". The nests of the loop 'for' inside a timed loop. The problem that I'm stuck on is how to reset the loop 'for' in order to fill the buffer with each iteration of the outer loop timed.
I enclose the code if that helps. Note that I use labview in 2013.
If anyone who has encountered a similar problem before can help me, I would be gratefull.
Concerning
Ali
You can use a shift register initialized as inputs to your curls, to build your table of one. However, you realize, don't you, that the execution of your program as written loops as soon as the computer can spit out responses, then the output array may contain several points that took place before the entry has updated? Assuming that your thing cRIO (I've not worked with these) has integrated into its routine acquisition of calendar data, I suggest you put that inside the loop to get a data point by reading.
And do you really want your graph to distance for update only every thousand points (points of acceleration million) speed?
Cameron
-
Not the same State (<>) of nested loops
Hi gurus and Experts.
I need advice/suggestion on the query plan below. This plan is go for Nested Loops, am not able to force to go to the hash join or Merge Join with tips.
Can someone help me please what is happening, and it is possible to escape from nested loops.
I don't want to take the help of index for this scenario.
D/B: Oracle 11 g 2,
create table parent as with main as (select --+materialize round(dbms_random.value(low=>1,high=>100)) SAMPLE_ID,dbms_random.string(opt=>'U',len=>10) SAMPLE_TEXT,rownum id from dual connect by level<=100000) select sample_id,rpad(sample_text,10,'*') padding,sample_text,id from main order by idcreate table child as with main as (select --+materialize round(dbms_random.value(low=>1,high=>100)) SAMPLE_ID,dbms_random.string(opt=>'U',len=>10) SAMPLE_TEXT,rownum id from dual connect by level<=10000) select sample_id,rpad(sample_text,10,'*') padding,sample_text,id from main order by idselect --+use_hash(p c) p.id from parent p,child c where p.id<>c.idPlan SELECT STATEMENT ALL_ROWSCost: 973,076 Bytes: 8,999,910,000 Cardinality: 999,990,000 3 NESTED LOOPS Cost: 973,076 Bytes: 8,999,910,000 Cardinality: 999,990,000 1 TABLE ACCESS FULL TABLE APP.CHILD Cost: 12 Bytes: 40,000 Cardinality: 10,000 2 TABLE ACCESS FULL TABLE APP.PARENT Cost: 97 Bytes: 499,995 Cardinality: 99,999902629 wrote:
I need advice/suggestion on the query plan below. This plan is go for Nested Loops, am not able to force to go to the hash join or Merge Join with tips.
You will not be able to use a hash join
http://docs.Oracle.com/CD/E11882_01/server.112/e16638/optimops.htm#i36043
>
The optimizer uses hash join to join two tables if they are added by using an equijoin and if one of the following conditions are met:
>You may be able to force a sort merge join, it does not mention if different will work, I don't see how it would be
http://docs.Oracle.com/CD/E11882_01/server.112/e16638/optimops.htm#i49183
>
Sort merge joins are useful when the join condition between two tables is a condition of inequality such as <,> <=,>, or > =. Fusion type joints are more successful than joins loops nested for large sets of data. You cannot use hash joins unless there is a condition of equality.
>I can't imagine this treatment an antijoin more efficiently than nested If loops.
-
Hello
A fundamental question about Nested Loops, what is the number of rows in the inner table means (4 in this example)? This is the average number of lines that oracle had to read by loop?
Thank you.
SYS> SHOW PARAMETER optimizer_features_enable NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ optimizer_features_enable string 11.2.0.1 HR> EXEC DBMS_STATS.GATHER_TABLE_STATS (USER , 'EMPLOYEES' , CASCADE => TRUE ) PL/SQL procedure successfully completed. HR> EXEC DBMS_STATS.GATHER_TABLE_STATS (USER , 'DEPARTMENTS' , CASCADE => TRUE ) PL/SQL procedure successfully completed. HR> SELECT COUNT(*) FROM employees ; COUNT(*) ---------- 107 HR> SELECT COUNT(*) FROM departments ; COUNT(*) ---------- 27 SELECT /*+ USE_NL(e d) */ e.last_name , d.department_name FROM employees e , departments d WHERE e.department_id = d.department_id; Execution Plan ---------------------------------------------------------- Plan hash value: 2968905875 ------------------------------------------------------------------ | Id | Operation | Name | Rows | Bytes | Cost | ------------------------------------------------------------------ | 0 | SELECT STATEMENT | | 106 | 2862 | 24 | | 1 | NESTED LOOPS | | 106 | 2862 | 24 | | 2 | TABLE ACCESS FULL| DEPARTMENTS | 27 | 432 | 2 | |* 3 | TABLE ACCESS FULL| EMPLOYEES | *4* | 44 | 1 | ------------------------------------------------------------------Hello
A fundamental question about Nested Loops, what is the number of rows in the inner table means (4 in this example)? This is the average number of lines that oracle had to read by loop?
Hello
in fact, it's exactly what it is, with the exception of is not the true Oracle number was read, it is an estimate.
You can track 10053 and see how the optimizer reach these numbers. The optimizer calculates the cardinality of the join using the following formula:
cardinality (employees) * *(join selectivity) cardinality (departments).
and join selectivity is essentially 1/greater (num_distinct (departments.department_id), (departments.department_id) num_distinct), adjusted for null values (not NULL values in this particular case, so no adjustment is necessary).
If you plug the values you get for the cardinality of the join:
107 * 27 * join selectivity, which join selectivity = 1/greater(11, 27) = 1/27 = 0.036691, so 107 * 27 * 0.036 ~ = 106.
4 comes from this number by dividing the cardinality expected join (106) by the estimated number of iterations of the nested loop (27).
It is not true the optimizer assumes an even distribution here - if it did, then it should get to 11 employees by Department, not 4 (because num_distinct (employees.department_id) = 11, not 4).
If you are interested in these issues, please refer to the excellent book by Jonathan Lewis "Cost based fundamentals", is all explained quite well in there.
Best regards
Nikolai -
When oracle uses a nested loop join full?
Hello!
Oracle database uses a complete nested loop join? I mean, imagine 2 tables without any index... is there a case where Oracle made a full scan of the inner table for each row in the outer table? I know it's the original algorithm of the nested loop join, but some databases prefer to do a temp table to autoindex the internal table and never does a full scan of the inner table...
Thank you!!http://download.Oracle.com/docs/CD/B10501_01/server.920/a96533/optimops.htm#49548
is 100% clear in its language. Simply click on it.
A nested loop join involves the following steps:The optimizer determines the driving table and it designates as the outer table.
The other table is referred to as the inner table.
For each row in the outer table, Oracle access all rows from the inner table. The outer loop is for each row in the outer table, and the inner loop is for each line of the internal table. The outer loop appears before the inner loop in the execution plan, as follows:
NESTED LOOPS
outer_loop
inner_loopConcerning
Girish Sharma -
Partition pruning, nested loops
Hello
I'm having a problem with obtaining the size of partition in a query. I managed to dumb down the problem to the two tables and a minimum query (see below).
Basically I have a partitioned table "Made each year" and a table of assistance 'Current year' that contain always 1 row. The sole purpose of this line is to say which is the year curent, it was the previous year and that next year will be. (In the real problem, there is no standard timeperiods, so we can't calculate just previous and next by adding/subtraction 1 as would be possible in this example).
The following query is executed as I want.
It runs a scan on current_year and then nested loop on the facts. And the size of the partition that was happening.
The following query is where I have my problem.select sum(decode(a.year_key, b.curr_year, some_measure)) as curr_year_measure from yearly_fact_t a ,current_year b where a.year_key = b.curr_year; ------------------------------------------------------------------------------------------------------------ | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop | ------------------------------------------------------------------------------------------------------------ | 0 | SELECT STATEMENT | | 1 | 39 | 4 (0)| 00:00:01 | | | | 1 | SORT AGGREGATE | | 1 | 39 | | | | | | 2 | NESTED LOOPS | | 1 | 39 | 4 (0)| 00:00:01 | | | | 3 | INDEX FAST FULL SCAN | SYS_C00247890 | 1 | 13 | 2 (0)| 00:00:01 | | | | 4 | PARTITION RANGE ITERATOR| | 1 | 26 | 2 (0)| 00:00:01 | KEY | KEY | |* 5 | TABLE ACCESS FULL | YEARLY_FACT_T | 1 | 26 | 2 (0)| 00:00:01 | KEY | KEY | ------------------------------------------------------------------------------------------------------------ Predicate Information (identified by operation id): --------------------------------------------------- 5 - filter("A"."YEAR_KEY"="B"."CURR_YEAR")
The xplan base is the same, but for some reason any cbo abandoned and decide to analyze all the partitions, which was not highly scalable data on production :)
I would have thought that the plan would be the same.
select sum(decode(a.year_key, b.curr_year, some_measure)) as curr_year_measure ,sum(decode(a.year_key, b.prev_year, some_measure)) as prev_year_measure from yearly_fact_t a ,current_year b where a.year_key = b.curr_year or a.year_key = b.prev_year; ------------------------------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop | ------------------------------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1 | 52 | 13 (0)| 00:00:01 | | | | 1 | SORT AGGREGATE | | 1 | 52 | | | | | | 2 | NESTED LOOPS | | 4 | 208 | 13 (0)| 00:00:01 | | | | 3 | TABLE ACCESS FULL | CURRENT_YEAR | 1 | 26 | 4 (0)| 00:00:01 | | | | 4 | PARTITION RANGE ALL| | 4 | 104 | 9 (0)| 00:00:01 | 1 | 6 | |* 5 | TABLE ACCESS FULL | YEARLY_FACT_T | 4 | 104 | 9 (0)| 00:00:01 | 1 | 6 | ------------------------------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 5 - filter("A"."YEAR_KEY"="B"."CURR_YEAR" OR "A"."YEAR_KEY"="B"."PREV_YEAR")
What can I do to get the size of partition that you want to occur in the second query?-- drop table yearly_fact_t purge; -- drop table current_year purge; create table current_year( curr_year number(4) not null ,prev_year number(4) not null ,next_year number(4) not null ,primary key(curr_year) ,unique(prev_year) ,unique(next_year) ); insert into current_year(curr_year, prev_year, next_year) values(2010, 2009, 2011); commit; create table yearly_fact_t( year_key number(4) not null ,some_dim_key number not null ,some_measure number not null ) partition by range(year_key)( partition p2007 values less than(2008) ,partition p2008 values less than(2009) ,partition p2009 values less than(2010) ,partition p2010 values less than(2011) ,partition p2011 values less than(2012) ,partition pmax values less than(maxvalue) ); insert into yearly_fact_t(year_key, some_dim_key, some_measure) values(2007,1, 10); insert into yearly_fact_t(year_key, some_dim_key, some_measure) values(2008,1, 20); insert into yearly_fact_t(year_key, some_dim_key, some_measure) values(2009,1, 30); insert into yearly_fact_t(year_key, some_dim_key, some_measure) values(2010,1, 40); commit;
Or better yet, what is it in my query that prevents it from happening?
We do not have Oracle Database 10 g Enterprise Edition Release 10.2.0.4.0 - 64 bit.
Best regards
RonnieI think that the decision of condition influenced gold optimizer. You can rewrite your query to use UNION ALL to see if it helps.
select sum(decode(year_key, curr_year, some_measure)) as curr_year_measure ,sum(decode(year_key, prev_year, some_measure)) as prev_year_measure from (select a.year_key, b.curr_year, some_measure from yearly_fact_t a ,current_year b where a.year_key = b.curr_year UNION ALL select a.year_key, b.curr_year, some_measure from yearly_fact_t a ,current_year b where a.year_key = b.prev_year);Published by: user503699 on February 17, 2010 14:59
-
Hi, I'm a relatively new to programming of LV, so be gentle.
In a Word, I am running nested loops than the material test power on and off cycles. The parameters for each loop is entered on the front panel of the user. For example, the power supply at the time, how many times to read in the data, power supply off time and the number of cycles power is a typical customer task ask me to do. Therefore, I decided to do a generic as possible executable vi instead of writing code (in which I request is trivial), that my technicians can easily use.
I want to put a counter - down in this vi, even a journey of one. I'll take the more the off time multiplied by the number of cycles and display on the front panel. For some reason that I can't find a simple way such that when the last cycle ended if poster zero before the program ends.
For example, the total is equal to 60 minutes and 10-minute cycle. I run the program and the display shows 60 during the first cycle, after the first cycle is complete and the second cycle begins the display turns 50, and so on until that last cycle where you can read 10. Then the program ends and the screen 10. How can I get a last beat of this kind it says 0 and 0 poster before the end of the program?
Thank you!
Wire a zero to a local variable of this indicator so that it can turn when the while loop ends.
-
Value of the loop 'For' pass before the complete loop - FPGA
Hello
8.5 LV, LV FPGA, PCI 7831-R FPGA Board
I got a cramp of brain on this one. I have a function (Arb. GIS read) that Im using to generate an arbitrary signal which I created in memory. I can't move to the value of data however. I don't know why I can not, its because it is nested in a "loop For" who runs indefinetly and updates only the value whenever it loops back to 'zero '. Ideas quick and dirty on how to use this value as it is being updated in the loop in my hand vi?
I've seen messages on the use of local variable 'Files' and property nodes, but I can not just give a sense the. Maybe because it's FPGA, something is different/no supported?
* My principal is 'control MicroMirror Arb. SP", look in the #4 case and the condition of"false. "
The 'Sub - VI' is called 'Arb Sig read RevB', and I'm trying to pass the variable 'Data' to hand while the loop For always runs.
Thank you!
-
Control of the double loop for a positioning system. - movement ocntrol
I am currently a control of the double loop for a positioning system, I use the speedometer to read the speed and the encoder for position on the position of the platform.
the problem I'm meeting now, is that I'm not sure weather the computer have the PID toolkit, because I can't find them under the control design and simulation module. so I did a PID using the basic features of labview, related information in the attached documents.
I'm not sure of the part of the dt, the labview has this kind of service as of the timing of simpling?
do I have to install the labview real_time potion? I'm just using the labview 8.5
Another issue is that, how I'm going to let the system remember the position instead of back to 0 whenever I restarted the program? If not is there any other way I can control the motor to move the position of the target that is initially set?
the last question is that is it possible for me to generate a profile about the current and desired position? to set a maximum speed and acceleration.
I do not have a motion controller, so I have not installed the module version and motion.
can someone suggest me a way to do it?
Hey, 14 I never used TDMS files. but maybe this will help. [zip file]
and if it isn't, I found this (which you can probably read into its equivalent in English)
http://digital.NI.com/public.nsf/allkb/099731B9B31936B08625745A0076A610
[don't forget to Congrats]
Maybe you are looking for
-
my itunes purchase history for some year
I really appreciate Drop Gummy, but here to keep track of what I spend (not much). But to do this, I need to know what I spent the beginning of 2016. May not know how to do this!
-
How can I change the number on my macbook pro 10.11.2 characters?
I've always been able to adjust my fractions and now since they have repaired and restored my macbook pro to its factory settings, I can access is no longer a viewer of characters as I used to. I even restored to my original my time machine settings
-
Trying to figure out how to use the Windows calculator to convert a basic number of 50pico.
I can't find out how to enter to the power 10 - 12. When I type exp it evokes 10th + 0, I want 10-12. I also can't figure out how to use the brackets and parentheses, I tried to use help and can not find an answer for what I am trying to do. I need t
-
External DVD for Inspiron 7459?
Just bought a desktop Inspiron 7459 all-in-one PC and need an external DVD/CD drive compatible with it. Any suggestions? Thank you
-
Window update causing problems of printer driver
Hi, I have a laptop running XP and router wireless computer (WAG54G) and a set of HP6380 in one. About 2 nights ago an update to Windows Installer by himself and rebooted the laptop (even though I have some turned off automatic updates what they seem