Search Results

Search found 2224 results on 89 pages for 'scientific computing'.

Page 17/89 | < Previous Page | 13 14 15 16 17 18 19 20 21 22 23 24  | Next Page >

  • How John Got 15x Improvement Without Really Trying

    - by rchrd
    The following article was published on a Sun Microsystems website a number of years ago by John Feo. It is still useful and worth preserving. So I'm republishing it here.  How I Got 15x Improvement Without Really Trying John Feo, Sun Microsystems Taking ten "personal" program codes used in scientific and engineering research, the author was able to get from 2 to 15 times performance improvement easily by applying some simple general optimization techniques. Introduction Scientific research based on computer simulation depends on the simulation for advancement. The research can advance only as fast as the computational codes can execute. The codes' efficiency determines both the rate and quality of results. In the same amount of time, a faster program can generate more results and can carry out a more detailed simulation of physical phenomena than a slower program. Highly optimized programs help science advance quickly and insure that monies supporting scientific research are used as effectively as possible. Scientific computer codes divide into three broad categories: ISV, community, and personal. ISV codes are large, mature production codes developed and sold commercially. The codes improve slowly over time both in methods and capabilities, and they are well tuned for most vendor platforms. Since the codes are mature and complex, there are few opportunities to improve their performance solely through code optimization. Improvements of 10% to 15% are typical. Examples of ISV codes are DYNA3D, Gaussian, and Nastran. Community codes are non-commercial production codes used by a particular research field. Generally, they are developed and distributed by a single academic or research institution with assistance from the community. Most users just run the codes, but some develop new methods and extensions that feed back into the general release. The codes are available on most vendor platforms. Since these codes are younger than ISV codes, there are more opportunities to optimize the source code. Improvements of 50% are not unusual. Examples of community codes are AMBER, CHARM, BLAST, and FASTA. Personal codes are those written by single users or small research groups for their own use. These codes are not distributed, but may be passed from professor-to-student or student-to-student over several years. They form the primordial ocean of applications from which community and ISV codes emerge. Government research grants pay for the development of most personal codes. This paper reports on the nature and performance of this class of codes. Over the last year, I have looked at over two dozen personal codes from more than a dozen research institutions. The codes cover a variety of scientific fields, including astronomy, atmospheric sciences, bioinformatics, biology, chemistry, geology, and physics. The sources range from a few hundred lines to more than ten thousand lines, and are written in Fortran, Fortran 90, C, and C++. For the most part, the codes are modular, documented, and written in a clear, straightforward manner. They do not use complex language features, advanced data structures, programming tricks, or libraries. I had little trouble understanding what the codes did or how data structures were used. Most came with a makefile. Surprisingly, only one of the applications is parallel. All developers have access to parallel machines, so availability is not an issue. Several tried to parallelize their applications, but stopped after encountering difficulties. Lack of education and a perception that parallelism is difficult prevented most from trying. I parallelized several of the codes using OpenMP, and did not judge any of the codes as difficult to parallelize. Even more surprising than the lack of parallelism is the inefficiency of the codes. I was able to get large improvements in performance in a matter of a few days applying simple optimization techniques. Table 1 lists ten representative codes [names and affiliation are omitted to preserve anonymity]. Improvements on one processor range from 2x to 15.5x with a simple average of 4.75x. I did not use sophisticated performance tools or drill deep into the program's execution character as one would do when tuning ISV or community codes. Using only a profiler and source line timers, I identified inefficient sections of code and improved their performance by inspection. The changes were at a high level. I am sure there is another factor of 2 or 3 in each code, and more if the codes are parallelized. The study’s results show that personal scientific codes are running many times slower than they should and that the problem is pervasive. Computational scientists are not sloppy programmers; however, few are trained in the art of computer programming or code optimization. I found that most have a working knowledge of some programming language and standard software engineering practices; but they do not know, or think about, how to make their programs run faster. They simply do not know the standard techniques used to make codes run faster. In fact, they do not even perceive that such techniques exist. The case studies described in this paper show that applying simple, well known techniques can significantly increase the performance of personal codes. It is important that the scientific community and the Government agencies that support scientific research find ways to better educate academic scientific programmers. The inefficiency of their codes is so bad that it is retarding both the quality and progress of scientific research. # cacheperformance redundantoperations loopstructures performanceimprovement 1 x x 15.5 2 x 2.8 3 x x 2.5 4 x 2.1 5 x x 2.0 6 x 5.0 7 x 5.8 8 x 6.3 9 2.2 10 x x 3.3 Table 1 — Area of improvement and performance gains of 10 codes The remainder of the paper is organized as follows: sections 2, 3, and 4 discuss the three most common sources of inefficiencies in the codes studied. These are cache performance, redundant operations, and loop structures. Each section includes several examples. The last section summaries the work and suggests a possible solution to the issues raised. Optimizing cache performance Commodity microprocessor systems use caches to increase memory bandwidth and reduce memory latencies. Typical latencies from processor to L1, L2, local, and remote memory are 3, 10, 50, and 200 cycles, respectively. Moreover, bandwidth falls off dramatically as memory distances increase. Programs that do not use cache effectively run many times slower than programs that do. When optimizing for cache, the biggest performance gains are achieved by accessing data in cache order and reusing data to amortize the overhead of cache misses. Secondary considerations are prefetching, associativity, and replacement; however, the understanding and analysis required to optimize for the latter are probably beyond the capabilities of the non-expert. Much can be gained simply by accessing data in the correct order and maximizing data reuse. 6 out of the 10 codes studied here benefited from such high level optimizations. Array Accesses The most important cache optimization is the most basic: accessing Fortran array elements in column order and C array elements in row order. Four of the ten codes—1, 2, 4, and 10—got it wrong. Compilers will restructure nested loops to optimize cache performance, but may not do so if the loop structure is too complex, or the loop body includes conditionals, complex addressing, or function calls. In code 1, the compiler failed to invert a key loop because of complex addressing do I = 0, 1010, delta_x IM = I - delta_x IP = I + delta_x do J = 5, 995, delta_x JM = J - delta_x JP = J + delta_x T1 = CA1(IP, J) + CA1(I, JP) T2 = CA1(IM, J) + CA1(I, JM) S1 = T1 + T2 - 4 * CA1(I, J) CA(I, J) = CA1(I, J) + D * S1 end do end do In code 2, the culprit is conditionals do I = 1, N do J = 1, N If (IFLAG(I,J) .EQ. 0) then T1 = Value(I, J-1) T2 = Value(I-1, J) T3 = Value(I, J) T4 = Value(I+1, J) T5 = Value(I, J+1) Value(I,J) = 0.25 * (T1 + T2 + T5 + T4) Delta = ABS(T3 - Value(I,J)) If (Delta .GT. MaxDelta) MaxDelta = Delta endif enddo enddo I fixed both programs by inverting the loops by hand. Code 10 has three-dimensional arrays and triply nested loops. The structure of the most computationally intensive loops is too complex to invert automatically or by hand. The only practical solution is to transpose the arrays so that the dimension accessed by the innermost loop is in cache order. The arrays can be transposed at construction or prior to entering a computationally intensive section of code. The former requires all array references to be modified, while the latter is cost effective only if the cost of the transpose is amortized over many accesses. I used the second approach to optimize code 10. Code 5 has four-dimensional arrays and loops are nested four deep. For all of the reasons cited above the compiler is not able to restructure three key loops. Assume C arrays and let the four dimensions of the arrays be i, j, k, and l. In the original code, the index structure of the three loops is L1: for i L2: for i L3: for i for l for l for j for k for j for k for j for k for l So only L3 accesses array elements in cache order. L1 is a very complex loop—much too complex to invert. I brought the loop into cache alignment by transposing the second and fourth dimensions of the arrays. Since the code uses a macro to compute all array indexes, I effected the transpose at construction and changed the macro appropriately. The dimensions of the new arrays are now: i, l, k, and j. L3 is a simple loop and easily inverted. L2 has a loop-carried scalar dependence in k. By promoting the scalar name that carries the dependence to an array, I was able to invert the third and fourth subloops aligning the loop with cache. Code 5 is by far the most difficult of the four codes to optimize for array accesses; but the knowledge required to fix the problems is no more than that required for the other codes. I would judge this code at the limits of, but not beyond, the capabilities of appropriately trained computational scientists. Array Strides When a cache miss occurs, a line (64 bytes) rather than just one word is loaded into the cache. If data is accessed stride 1, than the cost of the miss is amortized over 8 words. Any stride other than one reduces the cost savings. Two of the ten codes studied suffered from non-unit strides. The codes represent two important classes of "strided" codes. Code 1 employs a multi-grid algorithm to reduce time to convergence. The grids are every tenth, fifth, second, and unit element. Since time to convergence is inversely proportional to the distance between elements, coarse grids converge quickly providing good starting values for finer grids. The better starting values further reduce the time to convergence. The downside is that grids of every nth element, n > 1, introduce non-unit strides into the computation. In the original code, much of the savings of the multi-grid algorithm were lost due to this problem. I eliminated the problem by compressing (copying) coarse grids into continuous memory, and rewriting the computation as a function of the compressed grid. On convergence, I copied the final values of the compressed grid back to the original grid. The savings gained from unit stride access of the compressed grid more than paid for the cost of copying. Using compressed grids, the loop from code 1 included in the previous section becomes do j = 1, GZ do i = 1, GZ T1 = CA(i+0, j-1) + CA(i-1, j+0) T4 = CA1(i+1, j+0) + CA1(i+0, j+1) S1 = T1 + T4 - 4 * CA1(i+0, j+0) CA(i+0, j+0) = CA1(i+0, j+0) + DD * S1 enddo enddo where CA and CA1 are compressed arrays of size GZ. Code 7 traverses a list of objects selecting objects for later processing. The labels of the selected objects are stored in an array. The selection step has unit stride, but the processing steps have irregular stride. A fix is to save the parameters of the selected objects in temporary arrays as they are selected, and pass the temporary arrays to the processing functions. The fix is practical if the same parameters are used in selection as in processing, or if processing comprises a series of distinct steps which use overlapping subsets of the parameters. Both conditions are true for code 7, so I achieved significant improvement by copying parameters to temporary arrays during selection. Data reuse In the previous sections, we optimized for spatial locality. It is also important to optimize for temporal locality. Once read, a datum should be used as much as possible before it is forced from cache. Loop fusion and loop unrolling are two techniques that increase temporal locality. Unfortunately, both techniques increase register pressure—as loop bodies become larger, the number of registers required to hold temporary values grows. Once register spilling occurs, any gains evaporate quickly. For multiprocessors with small register sets or small caches, the sweet spot can be very small. In the ten codes presented here, I found no opportunities for loop fusion and only two opportunities for loop unrolling (codes 1 and 3). In code 1, unrolling the outer and inner loop one iteration increases the number of result values computed by the loop body from 1 to 4, do J = 1, GZ-2, 2 do I = 1, GZ-2, 2 T1 = CA1(i+0, j-1) + CA1(i-1, j+0) T2 = CA1(i+1, j-1) + CA1(i+0, j+0) T3 = CA1(i+0, j+0) + CA1(i-1, j+1) T4 = CA1(i+1, j+0) + CA1(i+0, j+1) T5 = CA1(i+2, j+0) + CA1(i+1, j+1) T6 = CA1(i+1, j+1) + CA1(i+0, j+2) T7 = CA1(i+2, j+1) + CA1(i+1, j+2) S1 = T1 + T4 - 4 * CA1(i+0, j+0) S2 = T2 + T5 - 4 * CA1(i+1, j+0) S3 = T3 + T6 - 4 * CA1(i+0, j+1) S4 = T4 + T7 - 4 * CA1(i+1, j+1) CA(i+0, j+0) = CA1(i+0, j+0) + DD * S1 CA(i+1, j+0) = CA1(i+1, j+0) + DD * S2 CA(i+0, j+1) = CA1(i+0, j+1) + DD * S3 CA(i+1, j+1) = CA1(i+1, j+1) + DD * S4 enddo enddo The loop body executes 12 reads, whereas as the rolled loop shown in the previous section executes 20 reads to compute the same four values. In code 3, two loops are unrolled 8 times and one loop is unrolled 4 times. Here is the before for (k = 0; k < NK[u]; k++) { sum = 0.0; for (y = 0; y < NY; y++) { sum += W[y][u][k] * delta[y]; } backprop[i++]=sum; } and after code for (k = 0; k < KK - 8; k+=8) { sum0 = 0.0; sum1 = 0.0; sum2 = 0.0; sum3 = 0.0; sum4 = 0.0; sum5 = 0.0; sum6 = 0.0; sum7 = 0.0; for (y = 0; y < NY; y++) { sum0 += W[y][0][k+0] * delta[y]; sum1 += W[y][0][k+1] * delta[y]; sum2 += W[y][0][k+2] * delta[y]; sum3 += W[y][0][k+3] * delta[y]; sum4 += W[y][0][k+4] * delta[y]; sum5 += W[y][0][k+5] * delta[y]; sum6 += W[y][0][k+6] * delta[y]; sum7 += W[y][0][k+7] * delta[y]; } backprop[k+0] = sum0; backprop[k+1] = sum1; backprop[k+2] = sum2; backprop[k+3] = sum3; backprop[k+4] = sum4; backprop[k+5] = sum5; backprop[k+6] = sum6; backprop[k+7] = sum7; } for one of the loops unrolled 8 times. Optimizing for temporal locality is the most difficult optimization considered in this paper. The concepts are not difficult, but the sweet spot is small. Identifying where the program can benefit from loop unrolling or loop fusion is not trivial. Moreover, it takes some effort to get it right. Still, educating scientific programmers about temporal locality and teaching them how to optimize for it will pay dividends. Reducing instruction count Execution time is a function of instruction count. Reduce the count and you usually reduce the time. The best solution is to use a more efficient algorithm; that is, an algorithm whose order of complexity is smaller, that converges quicker, or is more accurate. Optimizing source code without changing the algorithm yields smaller, but still significant, gains. This paper considers only the latter because the intent is to study how much better codes can run if written by programmers schooled in basic code optimization techniques. The ten codes studied benefited from three types of "instruction reducing" optimizations. The two most prevalent were hoisting invariant memory and data operations out of inner loops. The third was eliminating unnecessary data copying. The nature of these inefficiencies is language dependent. Memory operations The semantics of C make it difficult for the compiler to determine all the invariant memory operations in a loop. The problem is particularly acute for loops in functions since the compiler may not know the values of the function's parameters at every call site when compiling the function. Most compilers support pragmas to help resolve ambiguities; however, these pragmas are not comprehensive and there is no standard syntax. To guarantee that invariant memory operations are not executed repetitively, the user has little choice but to hoist the operations by hand. The problem is not as severe in Fortran programs because in the absence of equivalence statements, it is a violation of the language's semantics for two names to share memory. Codes 3 and 5 are C programs. In both cases, the compiler did not hoist all invariant memory operations from inner loops. Consider the following loop from code 3 for (y = 0; y < NY; y++) { i = 0; for (u = 0; u < NU; u++) { for (k = 0; k < NK[u]; k++) { dW[y][u][k] += delta[y] * I1[i++]; } } } Since dW[y][u] can point to the same memory space as delta for one or more values of y and u, assignment to dW[y][u][k] may change the value of delta[y]. In reality, dW and delta do not overlap in memory, so I rewrote the loop as for (y = 0; y < NY; y++) { i = 0; Dy = delta[y]; for (u = 0; u < NU; u++) { for (k = 0; k < NK[u]; k++) { dW[y][u][k] += Dy * I1[i++]; } } } Failure to hoist invariant memory operations may be due to complex address calculations. If the compiler can not determine that the address calculation is invariant, then it can hoist neither the calculation nor the associated memory operations. As noted above, code 5 uses a macro to address four-dimensional arrays #define MAT4D(a,q,i,j,k) (double *)((a)->data + (q)*(a)->strides[0] + (i)*(a)->strides[3] + (j)*(a)->strides[2] + (k)*(a)->strides[1]) The macro is too complex for the compiler to understand and so, it does not identify any subexpressions as loop invariant. The simplest way to eliminate the address calculation from the innermost loop (over i) is to define a0 = MAT4D(a,q,0,j,k) before the loop and then replace all instances of *MAT4D(a,q,i,j,k) in the loop with a0[i] A similar problem appears in code 6, a Fortran program. The key loop in this program is do n1 = 1, nh nx1 = (n1 - 1) / nz + 1 nz1 = n1 - nz * (nx1 - 1) do n2 = 1, nh nx2 = (n2 - 1) / nz + 1 nz2 = n2 - nz * (nx2 - 1) ndx = nx2 - nx1 ndy = nz2 - nz1 gxx = grn(1,ndx,ndy) gyy = grn(2,ndx,ndy) gxy = grn(3,ndx,ndy) balance(n1,1) = balance(n1,1) + (force(n2,1) * gxx + force(n2,2) * gxy) * h1 balance(n1,2) = balance(n1,2) + (force(n2,1) * gxy + force(n2,2) * gyy)*h1 end do end do The programmer has written this loop well—there are no loop invariant operations with respect to n1 and n2. However, the loop resides within an iterative loop over time and the index calculations are independent with respect to time. Trading space for time, I precomputed the index values prior to the entering the time loop and stored the values in two arrays. I then replaced the index calculations with reads of the arrays. Data operations Ways to reduce data operations can appear in many forms. Implementing a more efficient algorithm produces the biggest gains. The closest I came to an algorithm change was in code 4. This code computes the inner product of K-vectors A(i) and B(j), 0 = i < N, 0 = j < M, for most values of i and j. Since the program computes most of the NM possible inner products, it is more efficient to compute all the inner products in one triply-nested loop rather than one at a time when needed. The savings accrue from reading A(i) once for all B(j) vectors and from loop unrolling. for (i = 0; i < N; i+=8) { for (j = 0; j < M; j++) { sum0 = 0.0; sum1 = 0.0; sum2 = 0.0; sum3 = 0.0; sum4 = 0.0; sum5 = 0.0; sum6 = 0.0; sum7 = 0.0; for (k = 0; k < K; k++) { sum0 += A[i+0][k] * B[j][k]; sum1 += A[i+1][k] * B[j][k]; sum2 += A[i+2][k] * B[j][k]; sum3 += A[i+3][k] * B[j][k]; sum4 += A[i+4][k] * B[j][k]; sum5 += A[i+5][k] * B[j][k]; sum6 += A[i+6][k] * B[j][k]; sum7 += A[i+7][k] * B[j][k]; } C[i+0][j] = sum0; C[i+1][j] = sum1; C[i+2][j] = sum2; C[i+3][j] = sum3; C[i+4][j] = sum4; C[i+5][j] = sum5; C[i+6][j] = sum6; C[i+7][j] = sum7; }} This change requires knowledge of a typical run; i.e., that most inner products are computed. The reasons for the change, however, derive from basic optimization concepts. It is the type of change easily made at development time by a knowledgeable programmer. In code 5, we have the data version of the index optimization in code 6. Here a very expensive computation is a function of the loop indices and so cannot be hoisted out of the loop; however, the computation is invariant with respect to an outer iterative loop over time. We can compute its value for each iteration of the computation loop prior to entering the time loop and save the values in an array. The increase in memory required to store the values is small in comparison to the large savings in time. The main loop in Code 8 is doubly nested. The inner loop includes a series of guarded computations; some are a function of the inner loop index but not the outer loop index while others are a function of the outer loop index but not the inner loop index for (j = 0; j < N; j++) { for (i = 0; i < M; i++) { r = i * hrmax; R = A[j]; temp = (PRM[3] == 0.0) ? 1.0 : pow(r, PRM[3]); high = temp * kcoeff * B[j] * PRM[2] * PRM[4]; low = high * PRM[6] * PRM[6] / (1.0 + pow(PRM[4] * PRM[6], 2.0)); kap = (R > PRM[6]) ? high * R * R / (1.0 + pow(PRM[4]*r, 2.0) : low * pow(R/PRM[6], PRM[5]); < rest of loop omitted > }} Note that the value of temp is invariant to j. Thus, we can hoist the computation for temp out of the loop and save its values in an array. for (i = 0; i < M; i++) { r = i * hrmax; TEMP[i] = pow(r, PRM[3]); } [N.B. – the case for PRM[3] = 0 is omitted and will be reintroduced later.] We now hoist out of the inner loop the computations invariant to i. Since the conditional guarding the value of kap is invariant to i, it behooves us to hoist the computation out of the inner loop, thereby executing the guard once rather than M times. The final version of the code is for (j = 0; j < N; j++) { R = rig[j] / 1000.; tmp1 = kcoeff * par[2] * beta[j] * par[4]; tmp2 = 1.0 + (par[4] * par[4] * par[6] * par[6]); tmp3 = 1.0 + (par[4] * par[4] * R * R); tmp4 = par[6] * par[6] / tmp2; tmp5 = R * R / tmp3; tmp6 = pow(R / par[6], par[5]); if ((par[3] == 0.0) && (R > par[6])) { for (i = 1; i <= imax1; i++) KAP[i] = tmp1 * tmp5; } else if ((par[3] == 0.0) && (R <= par[6])) { for (i = 1; i <= imax1; i++) KAP[i] = tmp1 * tmp4 * tmp6; } else if ((par[3] != 0.0) && (R > par[6])) { for (i = 1; i <= imax1; i++) KAP[i] = tmp1 * TEMP[i] * tmp5; } else if ((par[3] != 0.0) && (R <= par[6])) { for (i = 1; i <= imax1; i++) KAP[i] = tmp1 * TEMP[i] * tmp4 * tmp6; } for (i = 0; i < M; i++) { kap = KAP[i]; r = i * hrmax; < rest of loop omitted > } } Maybe not the prettiest piece of code, but certainly much more efficient than the original loop, Copy operations Several programs unnecessarily copy data from one data structure to another. This problem occurs in both Fortran and C programs, although it manifests itself differently in the two languages. Code 1 declares two arrays—one for old values and one for new values. At the end of each iteration, the array of new values is copied to the array of old values to reset the data structures for the next iteration. This problem occurs in Fortran programs not included in this study and in both Fortran 77 and Fortran 90 code. Introducing pointers to the arrays and swapping pointer values is an obvious way to eliminate the copying; but pointers is not a feature that many Fortran programmers know well or are comfortable using. An easy solution not involving pointers is to extend the dimension of the value array by 1 and use the last dimension to differentiate between arrays at different times. For example, if the data space is N x N, declare the array (N, N, 2). Then store the problem’s initial values in (_, _, 2) and define the scalar names new = 2 and old = 1. At the start of each iteration, swap old and new to reset the arrays. The old–new copy problem did not appear in any C program. In programs that had new and old values, the code swapped pointers to reset data structures. Where unnecessary coping did occur is in structure assignment and parameter passing. Structures in C are handled much like scalars. Assignment causes the data space of the right-hand name to be copied to the data space of the left-hand name. Similarly, when a structure is passed to a function, the data space of the actual parameter is copied to the data space of the formal parameter. If the structure is large and the assignment or function call is in an inner loop, then copying costs can grow quite large. While none of the ten programs considered here manifested this problem, it did occur in programs not included in the study. A simple fix is always to refer to structures via pointers. Optimizing loop structures Since scientific programs spend almost all their time in loops, efficient loops are the key to good performance. Conditionals, function calls, little instruction level parallelism, and large numbers of temporary values make it difficult for the compiler to generate tightly packed, highly efficient code. Conditionals and function calls introduce jumps that disrupt code flow. Users should eliminate or isolate conditionls to their own loops as much as possible. Often logical expressions can be substituted for if-then-else statements. For example, code 2 includes the following snippet MaxDelta = 0.0 do J = 1, N do I = 1, M < code omitted > Delta = abs(OldValue ? NewValue) if (Delta > MaxDelta) MaxDelta = Delta enddo enddo if (MaxDelta .gt. 0.001) goto 200 Since the only use of MaxDelta is to control the jump to 200 and all that matters is whether or not it is greater than 0.001, I made MaxDelta a boolean and rewrote the snippet as MaxDelta = .false. do J = 1, N do I = 1, M < code omitted > Delta = abs(OldValue ? NewValue) MaxDelta = MaxDelta .or. (Delta .gt. 0.001) enddo enddo if (MaxDelta) goto 200 thereby, eliminating the conditional expression from the inner loop. A microprocessor can execute many instructions per instruction cycle. Typically, it can execute one or more memory, floating point, integer, and jump operations. To be executed simultaneously, the operations must be independent. Thick loops tend to have more instruction level parallelism than thin loops. Moreover, they reduce memory traffice by maximizing data reuse. Loop unrolling and loop fusion are two techniques to increase the size of loop bodies. Several of the codes studied benefitted from loop unrolling, but none benefitted from loop fusion. This observation is not too surpising since it is the general tendency of programmers to write thick loops. As loops become thicker, the number of temporary values grows, increasing register pressure. If registers spill, then memory traffic increases and code flow is disrupted. A thick loop with many temporary values may execute slower than an equivalent series of thin loops. The biggest gain will be achieved if the thick loop can be split into a series of independent loops eliminating the need to write and read temporary arrays. I found such an occasion in code 10 where I split the loop do i = 1, n do j = 1, m A24(j,i)= S24(j,i) * T24(j,i) + S25(j,i) * U25(j,i) B24(j,i)= S24(j,i) * T25(j,i) + S25(j,i) * U24(j,i) A25(j,i)= S24(j,i) * C24(j,i) + S25(j,i) * V24(j,i) B25(j,i)= S24(j,i) * U25(j,i) + S25(j,i) * V25(j,i) C24(j,i)= S26(j,i) * T26(j,i) + S27(j,i) * U26(j,i) D24(j,i)= S26(j,i) * T27(j,i) + S27(j,i) * V26(j,i) C25(j,i)= S27(j,i) * S28(j,i) + S26(j,i) * U28(j,i) D25(j,i)= S27(j,i) * T28(j,i) + S26(j,i) * V28(j,i) end do end do into two disjoint loops do i = 1, n do j = 1, m A24(j,i)= S24(j,i) * T24(j,i) + S25(j,i) * U25(j,i) B24(j,i)= S24(j,i) * T25(j,i) + S25(j,i) * U24(j,i) A25(j,i)= S24(j,i) * C24(j,i) + S25(j,i) * V24(j,i) B25(j,i)= S24(j,i) * U25(j,i) + S25(j,i) * V25(j,i) end do end do do i = 1, n do j = 1, m C24(j,i)= S26(j,i) * T26(j,i) + S27(j,i) * U26(j,i) D24(j,i)= S26(j,i) * T27(j,i) + S27(j,i) * V26(j,i) C25(j,i)= S27(j,i) * S28(j,i) + S26(j,i) * U28(j,i) D25(j,i)= S27(j,i) * T28(j,i) + S26(j,i) * V28(j,i) end do end do Conclusions Over the course of the last year, I have had the opportunity to work with over two dozen academic scientific programmers at leading research universities. Their research interests span a broad range of scientific fields. Except for two programs that relied almost exclusively on library routines (matrix multiply and fast Fourier transform), I was able to improve significantly the single processor performance of all codes. Improvements range from 2x to 15.5x with a simple average of 4.75x. Changes to the source code were at a very high level. I did not use sophisticated techniques or programming tools to discover inefficiencies or effect the changes. Only one code was parallel despite the availability of parallel systems to all developers. Clearly, we have a problem—personal scientific research codes are highly inefficient and not running parallel. The developers are unaware of simple optimization techniques to make programs run faster. They lack education in the art of code optimization and parallel programming. I do not believe we can fix the problem by publishing additional books or training manuals. To date, the developers in questions have not studied the books or manual available, and are unlikely to do so in the future. Short courses are a possible solution, but I believe they are too concentrated to be much use. The general concepts can be taught in a three or four day course, but that is not enough time for students to practice what they learn and acquire the experience to apply and extend the concepts to their codes. Practice is the key to becoming proficient at optimization. I recommend that graduate students be required to take a semester length course in optimization and parallel programming. We would never give someone access to state-of-the-art scientific equipment costing hundreds of thousands of dollars without first requiring them to demonstrate that they know how to use the equipment. Yet the criterion for time on state-of-the-art supercomputers is at most an interesting project. Requestors are never asked to demonstrate that they know how to use the system, or can use the system effectively. A semester course would teach them the required skills. Government agencies that fund academic scientific research pay for most of the computer systems supporting scientific research as well as the development of most personal scientific codes. These agencies should require graduate schools to offer a course in optimization and parallel programming as a requirement for funding. About the Author John Feo received his Ph.D. in Computer Science from The University of Texas at Austin in 1986. After graduate school, Dr. Feo worked at Lawrence Livermore National Laboratory where he was the Group Leader of the Computer Research Group and principal investigator of the Sisal Language Project. In 1997, Dr. Feo joined Tera Computer Company where he was project manager for the MTA, and oversaw the programming and evaluation of the MTA at the San Diego Supercomputer Center. In 2000, Dr. Feo joined Sun Microsystems as an HPC application specialist. He works with university research groups to optimize and parallelize scientific codes. Dr. Feo has published over two dozen research articles in the areas of parallel parallel programming, parallel programming languages, and application performance.

    Read the article

  • Computing MD5SUM of large files in C#

    - by spkhaira
    I am using following code to compute MD5SUM of a file - byte[] b = System.IO.File.ReadAllBytes(file); string sum = BitConverter.ToString(new MD5CryptoServiceProvider().ComputeHash(b)); This works fine normally, but if I encounter a large file (~1GB) - e.g. an iso image or a DVD VOB file - I get an Out of Memory exception. Though, I am able to compute the MD5SUM in cygwin for the same file in about 10secs. Please suggest how can I get this to work for big files in my program. Thanks

    Read the article

  • running a python script where dependencies are not avail: distributed computing

    - by sadhu_
    Hi, I have access to a grid (running condor) that would (potentially) allow to very substantially reduce how long by nltk based nlp tasks take. unfortunately, i dont have root access on the cluster so cannot install new packages, only run whatever is available on the linux boxes. python is of course available, but nltk isnt - i was wondering however, if there might be a way around this somehow ? is there a way i can somehow still distribute the task in a self-contained 'package' of some sort? Thanks for your hel

    Read the article

  • computing "node closure" of graph with removal

    - by Fakrudeen
    Given a directed graph, the goal is to combine the node with the nodes it is pointing to and come up with minimum number of these [lets give the name] super nodes. The catch is once you combine the nodes you can't use those nodes again. [first node as well as all the combined nodes - that is all the members of one super node] The greedy approach would be to pick the node with maximum out degree and combine that node with nodes it is pointing to and remove all of them. Do this every time with the nodes which are not removed yet from graph. The greedy is O(V), but this won't necessarily output minimum number super nodes. So what is the best algorithm to do this?

    Read the article

  • Computing orientation of a square and displaying an object with the same orientation

    - by Robin
    Hi, I wrote an application which detects a square within an image. To give you a good understanding of how such an image containing such a square, in this case a marker, could look like: What I get, after the detection, are the coordinates of the four corners of my marker. Now I don't know how to display an object on my marker. The object should have the same rotation/angle/direction as the marker. Are there any papers on how to achieve that, any algorithms that I can use that proofed to be pretty solid/working? It doesn't need to be a working solution, it could be a simple description on how to achieve that or something similar. If you point me at a library or something, it should work under linux, windows is not needed but would be great in case I need to port the application at some point. I already looked at the ARToolkit but they you camera parameter files and more complex matrices while I only got the four corner points and a single image instead of a whole video / camera stream.

    Read the article

  • computing z-scores for 2D matrices in scipy/numpy in Python

    - by user248237
    How can I compute the z-score for matrices in Python? Suppose I have the array: a = array([[ 1, 2, 3], [ 30, 35, 36], [2000, 6000, 8000]]) and I want to compute the z-score for each row. The solution I came up with is: array([zs(item) for item in a]) where zs is in scipy.stats.stats. Is there a better built-in vectorized way to do this? Also, is it always good to z-score numbers before using hierarchical clustering with euclidean or seuclidean distance? Can anyone discuss the relative advantages/disadvantages? thanks.

    Read the article

  • silverlight for .NET / CLR based numerical computing on osx

    - by Jonathan Shore
    I'm interested in using F# for numerical work, but my platforms are not windows based. Mono still has a significant performance penalty for programs that generate a significant amount of short-lived objects (as would be typical for functional languages). Silverlight is available on OSX. I had seen some reference indicating that assemblies compiled in the usual way could not be referenced, but not clear on the details. I'm not interested in UIs, but wondering whether could use the VM bundled with silverlight effectively for execution? I would want to be able to reference a large library of numerical models I already have in java (cross-compiled via IKVM to .NET assemblies) and a new codebase written in F#. My hope would be that the silverlight VM on OSX has good performance and can reference external assemblies and native libraries. Is this doable?

    Read the article

  • Computing unique index for every poker starting hand

    - by Aly
    As there are 52 cards in a deck we know there are 52 choose 2 = 1326 distinct matchups, however in preflop poker this can be bucketed into 169 different hands such as AK offsuit and AK suited as whether it is A hearts K hearts or A spade K spades it makes no difference preflop. My question is, is there a nice mathematical property in which I can uniquely index each of these 169 hands (from 0 to 168 preferably). I am trying to create a look up table as a double[][] = new double [169][169] but have no way of changing a hand representation such as AKs (an Ace and a King of the same suit) to a unique index in this array.

    Read the article

  • The speed of .NET in numerical computing

    - by Yin Zhu
    In my experience, .net is 2 to 3 times slower than native code. (I implemented L-BFGS for multivariate optimization). I have traced the ads on stackoverflow to http://www.centerspace.net/products/ the speed is really amazing, the speed is close to native code. How can they do that? They said that: Q. Is NMath "pure" .NET? A. The answer depends somewhat on your definition of "pure .NET". NMath is written in C#, plus a small Managed C++ layer. For better performance of basic linear algebra operations, however, NMath does rely on the native Intel Math Kernel Library (included with NMath). But there are no COM components, no DLLs--just .NET assemblies. Also, all memory allocated in the Managed C++ layer and used by native code is allocated from the managed heap. Can someone explain more to me? Thanks!

    Read the article

  • Grails - Recursive computing call "StackOverflowError" and don't update on show

    - by Kedare
    Hello, I have a little problem, I have made a Category domain on Grails, but I want to "cache" the string representation on the database by generating it when the representation field is empty or NULL (to avoid recursive queries at each toString), here is the code : package devkedare class Category { String title; String description; String representation; Date dateCreated; Date lastUpdate; Category parent; static constraints = { title(blank:false, nullable:false, size:2..100); description(blank:true, nullable:false); dateCreated(nullable:true); lastUpdate(nullable:true); autoTimestamp: true; } static hasMany = [childs:Category] @Override String toString() { if((!this.representation) || (this.representation == "")) { this.computeRepresentation(true) } return this.representation; } String computeRepresentation(boolean updateChilds) { if(updateChilds) { for(child in this.childs) { child.representation = child.computeRepresentation(true); } } if(this.parent) { this.representation = "${this.parent.computeRepresentation(false)}/${this.title}"; } else { this.representation = this.title } return this.representation; } void beforeUpdate() { this.computeRepresentation(true); } } There is 2 problems : It don't update the representation when the representation if NULL or empty and toString id called, it only update it on save, how to fix that ? On some category, updating it throw a StackOverflowError, here is an example of my actual category table as CSV : Uploaded the csv file here, pasting csv looks buggy: http://kedare.free.fr/Dist/01.csv Updating the representation works everywhere except on "IT" that throw a StackOverlowError (this is the only category that has many childs). Any idea of how can I fix those 2 problems ? Thank you !

    Read the article

  • Parallel computing for integrals

    - by Iman
    I want to reduce the calculation time for a time-consuming integral by splitting the integration range. I'm using C++, Windows, and a quad-core Intel i7 CPU. How can I split it into 4 parallel computations?

    Read the article

  • Code computing the cross-product

    - by WizardOfOdds
    This is not a dupe of my question: http://stackoverflow.com/questions/2532810/detecting-one-points-location-compared-to-two-other-points If I have the following piece of pseudo-C/Java/C# code: int a[]= { 30, 20 }; int b[] = { 40, 50 }; int c[] = {12, 12}; How do I compute the sign of the cross-product ABxAC? I'm only interested in the sign, so I have: boolean signABxAC = ? Now concretely what do I write to get the sign of the cross-product ABxAC?

    Read the article

  • operators computing direction

    - by amiad
    Hi all! I enqunterd something that I can't understand. I have this code: cout << "f1 * f1 + f2 * f1 - f1 / f2 is: "<< f1 * f1 + f2 * f1 - f1 / f2 << endl; All the "f"s are objects, and all the operators are overloaded. The weird this is that the first computarion is of the "/" operator, then the second "" and then the first "", after that - the operator "+" and at last - operator "-". So basicly - the "/" and "*" worked from right to left, and the "+" and "-" operators worked from left to right. I made another test... I checked this code: cout << "f1 * f1 / f2 is: " << f1 * f1 / f2 << endl; Now, the first operator was "*" and only then oerator "/". So now, it worked from left to right. Can someone help me underatand why is there diffrence in the directions? 10X!

    Read the article

  • Computing "average" of two colors

    - by Francisco P.
    This is only marginally programming related - has much more to do w/ colors and their representation. I am working on a very low level app. I have an array of bytes in memory. Those are characters. They were rendered with anti-aliasing: they have values from 0 to 255, 0 being fully transparent and 255 totally opaque (alpha, if you wish). I am having trouble conceiving an algorithm for the rendering of this font. I'm doing the following for each pixel: // intensity is the weight I talked about: 0 to 255 intensity = glyphs[text[i]][x + GLYPH_WIDTH*y]; if (intensity == 255) continue; // Don't draw it, fully transparent else if (intensity == 0) setPixel(x + xi, y + yi, color, base); // Fully opaque, can draw original color else { // Here's the tricky part // Get the pixel in the destination for averaging purposes pixel = getPixel(x + xi, y + yi, base); // transfer is an int for calculations transfer = (int) ((float)((float) (255.0 - (float) intensity/255.0) * (float) color.red + (float) pixel.red)/2); // This is my attempt at averaging newPixel.red = (Byte) transfer; transfer = (int) ((float)((float) (255.0 - (float) intensity/255.0) * (float) color.green + (float) pixel.green)/2); newPixel.green = (Byte) transfer; // transfer = (int) ((float) ((float) 255.0 - (float) intensity)/255.0 * (((float) color.blue) + (float) pixel.blue)/2); transfer = (int) ((float)((float) (255.0 - (float) intensity/255.0) * (float) color.blue + (float) pixel.blue)/2); newPixel.blue = (Byte) transfer; // Set the newpixel in the desired mem. position setPixel(x+xi, y+yi, newPixel, base); } The results, as you can see, are less than desirable. That is a very zoomed in image, at 1:1 scale it looks like the text has a green "aura". Any idea for how to properly compute this would be greatly appreciated. Thanks for your time!

    Read the article

  • Arrays/Lists and computing hashvalues (VB, C#)

    - by Jeffrey Kern
    I feel bad asking this question but I am currently not able to program and test this as I'm writing this on my cell-phone and not on my dev machine :P (Easy rep points if someone answers! XD ) Anyway, I've had experience with using hashvalues from String objects. E.g., if I have StringA and StringB both equal to "foo", they'll both compute out the same hashvalue, because they're set to equal values. Now what if I have a List, with T being a native data type. If I tried to compute the hashvalue of ListA and ListB, assuming that they'd both be the same size and contain the same information, wouldn't they have equal hashvalues as well? Assuming as sample dataset of 'byte' with a length of 5 {5,2,0,1,3}

    Read the article

  • Whats the most common waste of computing power in javascipt

    - by qwertymk
    We've all seen people who do this jQuery('a').each(function(){ jQuery(this)[0].innerHTML += ' proccessed'; }); function letsPoluteNS() { polute = ''; for (morePolution = 0; morePolution < arguments.length; morePolution++) polute.join(arguments[morePolution]); return polute; } and so on. I was wondering what people have seen the most common javascript/jQuery technique that is slowing down the page and/or wasting time for the javascript engine. PS I know that this question may not seem to fit into whats an accepted question, yet I'm asking for what the most common accepted waste is

    Read the article

  • Native arrays and computing hashvalues (VB, C#)

    - by Jeffrey Kern
    I feel bad asking this question but I am currently not able to program and test this as I'm writing this on my cell-phone and not on my dev machine :P (Easy rep points if someone answers! XD ) Anyway, I've had experience with using hashvalues from String objects. E.g., if I have StringA and StringB both equal to "foo", they'll both compute out the same hashvalue, because they're set to equal values. Now what if I have a List, with T being a native data type. If I tried to compute the hashvalue of ListA and ListB, assuming that they'd both be the same size and contain the same information, wouldn't they have equal hashvalues as well?

    Read the article

  • How to Design Programs: An Introduction to Programming and Computing -- teacher guide access

    - by user295683
    Hello -- I'm a biologist switching careers, and trying to learn programming as a result. I stumbled upon the aforementioned book on Amazon, which jived with my liberal arts background. Despite my great satisfaction with the didactic approach, I was frustrated to see that the answers to the exercises are restricted to teachers only. As I am pursuing this endeavor on my own, this restriction dramatically cripples the value of this book. My request to the author's website for access to the answers has not been answered, and I would desperately like to continue with this book. Anyone have any experience dealing with the book's website, or at the very least a torrent of the answers? Otherwise, I suspect I will be relegated to using JavaScript for everything! Thanks!

    Read the article

  • Computing, storing, and retrieving values to and from an N-Dimensional matrix

    - by Adam S
    This question is probably quite different from what you are used to reading here - I hope it can provide a fun challenge. Essentially I have an algorithm that uses 5(or more) variables to compute a single value, called outcome. Now I have to implement this algorithm on an embedded device which has no memory limitations, but has very harsh processing constraints. Because of this, I would like to run a calculation engine which computes outcome for, say, 20 different values of each variable and stores this information in a file. You may think of this as a 5(or more)-dimensional matrix or 5(or more)-dimensional array, each dimension being 20 entries long. In any modern language, filling this array is as simple as having 5(or more) nested for loops. The tricky part is that I need to dump these values into a file that can then be placed onto the embedded device so that the device can use it as a lookup table. The questions now, are: What format(s) might be acceptable for storing the data? What programs (MATLAB, C#, etc) might be best suited to compute the data? C# must be used to import the data on the device - is this possible given your answer to #1?

    Read the article

  • python parallel computing: split keyspace to give each node a range to work on

    - by MatToufoutu
    My question is rather complicated for me to explain, as i'm not really good at maths, but i'll try to be as clear as possible. I'm trying to code a cluster in python, which will generate words given a charset (i.e. with lowercase: aaaa, aaab, aaac, ..., zzzz) and make various operations on them. I'm searching how to calculate, given the charset and the number of nodes, what range each node should work on (i.e.: node1: aaaa-azzz, node2: baaa-czzz, node3: daaa-ezzz, ...). Is it possible to make an algorithm that could compute this, and if it is, how could i implement this in python? I really don't know how to do that, so any help would be much appreciated

    Read the article

  • Computing the scalar product of two vectors in C++

    - by HowardRoark
    I am trying to write a program with a function double_product(vector<double> a, vector<double> b) that computes the scalar product of two vectors. The scalar product is $a_{0}b_{0}+a_{1}b_{1}+...+a_{n-1}b_{n-1}$. Here is what I have. It is a mess, but I am trying! #include <iostream> #include <vector> using namespace std; class Scalar_product { public: Scalar_product(vector<double> a, vector<double> b); }; double scalar_product(vector<double> a, vector<double> b) { double product = 0; for (int i = 0; i <= a.size()-1; i++) for (int i = 0; i <= b.size()-1; i++) product = product + (a[i])*(b[i]); return product; } int main() { cout << product << endl; return 0; }

    Read the article

< Previous Page | 13 14 15 16 17 18 19 20 21 22 23 24  | Next Page >