Vendor compilers typically provide the most information about their optimizations of your code. Some may also provide an inline source listing (Cray and Intel below) where the optimization comments and labels appear next to the code.
Cray compiler (craycc: defaults -O3 and OpenMP enabled)
arnoldg@h2ologin4:~/stream> cc -c -hmsgs -hlist=ai -DTUNED stream.c
arnoldg@h2ologin4:~/stream> cat stream.lst
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
S u m m a r y R e p o r t
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Compilation
-----------
File : /mnt/a/u/staff/arnoldg/stream/stream.c
Compiled : 2021-03-05 09:16:31
Compiler : Version 8.7.7
Ftnlx : Version 8503 (libcif 85008)
Target : x86-64
Command : driver.cc -h cpu=interlagos -h static -D __CRAYXE
-D __CRAY_INTERLAGOS -D __CRAYXT_COMPUTE_LINUX_TARGET
-h network=gemini -c -h msgs -h list=ai -D TUNED stream.c
-isystem /opt/cray/cce/8.7.7/cce/x86_64/include/craylibs
-isystem /opt/cray/cce/8.7.7/cce/x86_64/include/basic
-isystem /opt/gcc/6.1.0/snos/lib/gcc/x86_64-suse-linux/6.1.0/include
-isystem /opt/gcc/6.1.0/snos/lib/gcc/x86_64-suse-linux/6.1.0/include-
fixed -isystem /opt/gcc/6.1.0/snos/include -isystem /usr/include
-I /opt/cray/mpt/7.7.4/gni/mpich-cray/8.6/include
-I /opt/cray/libsci/18.12.1/CRAY/8.6/x86_64/include
-I /opt/cray/rca/1.0.0-2.0502.60530.1.63.gem/include
-I /opt/cray/alps/5.2.4-2.0502.9774.31.12.gem/include
-I /opt/cray/xpmem/0.1-2.0502.64982.5.3.gem/include
-I /opt/cray/gni-headers/4.0-1.0502.10859.7.8.gem/include
-I /opt/cray/dmapp/7.0.1-1.0502.11080.8.74.gem/include
-I /opt/cray/pmi/5.0.14/include
-I /opt/cray/ugni/6.0-1.0502.10863.8.28.gem/include
-I /opt/cray/udreg/2.3.2-1.0502.10518.2.17.gem/include
-I /usr/local/include
-I /opt/cray/wlm_detect/1.0-1.0502.64649.2.2.gem/include
-I /opt/cray/krca/1.0.0-2.0502.63139.4.30.gem/include
-I /opt/cray-hss-devel/7.2.0/include
clx report
------------
Source : /mnt/a/u/staff/arnoldg/stream/stream.c
Date : 03/05/2021 09:16:32
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
O p t i o n s R e p o r t
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Options : -h cache2,scalar2,thread2,vector2,mpi0,ipa3,noaggress
-h autoprefetch,noautothread,fusion2,msgs,nonegmsgs
-h nooverindex,pattern,unroll2,nozeroinc
-h noadd_paren,noupc,dwarf,fma,nofp_trap,nofunc_trace
-h noomp_analyze,noomp_trace,nopat_trace
-h omp,noacc
-h c99,noexceptions,noconform,noinfinitevl
-h notolerant,gnu
-h safe_addr,thread_do_concurrent,fp2=approx,flex_mp=default
-h alias=default:standard_restrict
-h static (or -static)
-h cpu=x86-64,interlagos
-h network=gemini
-K trap=none
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
S o u r c e L i s t i n g
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%% L o o p m a r k L e g e n d %%%
Primary Loop Type Modifiers
------- ---- ---- ---------
A - Pattern matched a - atomic memory operation
b - blocked
C - Collapsed c - conditional and/or computed
D - Deleted
E - Cloned
F - Flat - No calls f - fused
G - Accelerated g - partitioned
I - Inlined i - interchanged
M - Multithreaded m - partitioned
n - non-blocking remote transfer
p - partial
R - Rerolling r - unrolled
s - shortloop
V - Vectorized w - unwound
+ - More messages listed at end of listing
------------------------------------------
1. /*-----------------------------------------------------------------------*/
2. /* Program: Stream */
3. /* Revision: $Id: stream.c,v 5.9 2009/04/11 16:35:00 mccalpin Exp $ */
4. /* Original code developed by John D. McCalpin */
5. /* Programmers: John D. McCalpin */
6. /* Joe R. Zagar */
7. /* */
8. /* This program measures memory transfer rates in MB/s for simple */
9. /* computational kernels coded in C. */
10. /*-----------------------------------------------------------------------*/
11. /* Copyright 1991-2005: John D. McCalpin */
12. /*-----------------------------------------------------------------------*/
13. /* License: */
14. /* 1. You are free to use this program and/or to redistribute */
15. /* this program. */
16. /* 2. You are free to modify this program for your own use, */
17. /* including commercial use, subject to the publication */
18. /* restrictions in item 3. */
19. /* 3. You are free to publish results obtained from running this */
20. /* program, or from works that you derive from this program, */
21. /* with the following limitations: */
22. /* 3a. In order to be referred to as "STREAM benchmark results", */
23. /* published results must be in conformance to the STREAM */
24. /* Run Rules, (briefly reviewed below) published at */
25. /* http://www.cs.virginia.edu/stream/ref.html */
26. /* and incorporated herein by reference. */
27. /* As the copyright holder, John McCalpin retains the */
28. /* right to determine conformity with the Run Rules. */
29. /* 3b. Results based on modified source code or on runs not in */
30. /* accordance with the STREAM Run Rules must be clearly */
31. /* labelled whenever they are published. Examples of */
32. /* proper labelling include: */
33. /* "tuned STREAM benchmark results" */
34. /* "based on a variant of the STREAM benchmark code" */
35. /* Other comparable, clear and reasonable labelling is */
36. /* acceptable. */
37. /* 3c. Submission of results to the STREAM benchmark web site */
38. /* is encouraged, but not required. */
39. /* 4. Use of this program or creation of derived works based on this */
40. /* program constitutes acceptance of these licensing restrictions. */
41. /* 5. Absolutely no warranty is expressed or implied. */
42. /*-----------------------------------------------------------------------*/
43. # include <stdio.h>
44. # include <math.h>
45. # include <float.h>
46. # include <limits.h>
47. # include <sys/time.h>
48.
49. /* INSTRUCTIONS:
50. *
51. * 1) Stream requires a good bit of memory to run. Adjust the
52. * value of 'N' (below) to give a 'timing calibration' of
53. * at least 20 clock-ticks. This will provide rate estimates
54. * that should be good to about 5% precision.
55. */
56.
57. #ifndef N
58. # define N 40000000
59. #endif
60. #ifndef NTIMES
61. # define NTIMES 10
62. #endif
63. #ifndef OFFSET
64. # define OFFSET 0
65. #endif
66.
67. /*
68. * 3) Compile the code with full optimization. Many compilers
69. * generate unreasonably bad code before the optimizer tightens
70. * things up. If the results are unreasonably good, on the
71. * other hand, the optimizer might be too smart for me!
72. *
73. * Try compiling with:
74. * cc -O stream_omp.c -o stream_omp
75. *
76. * This is known to work on Cray, SGI, IBM, and Sun machines.
77. *
78. *
79. * 4) Mail the results to mccalpin@cs.virginia.edu
80. * Be sure to include:
81. * a) computer hardware model number and software revision
82. * b) the compiler flags
83. * c) all of the output from the test case.
84. * Thanks!
85. *
86. */
87.
88. # define HLINE "-------------------------------------------------------------\n"
89.
90. # ifndef MIN
91. # define MIN(x,y) ((x)<(y)?(x):(y))
92. # endif
93. # ifndef MAX
94. # define MAX(x,y) ((x)>(y)?(x):(y))
95. # endif
96.
97. static double a[N+OFFSET],
98. b[N+OFFSET],
99. c[N+OFFSET];
100.
101. static double avgtime[4] = {0}, maxtime[4] = {0},
102. mintime[4] = {FLT_MAX,FLT_MAX,FLT_MAX,FLT_MAX};
103.
104. static char *label[4] = {"Copy: ", "Scale: ",
105. "Add: ", "Triad: "};
106.
107. static double bytes[4] = {
108. 2 * sizeof(double) * N,
109. 2 * sizeof(double) * N,
110. 3 * sizeof(double) * N,
111. 3 * sizeof(double) * N
112. };
113.
114. extern double mysecond();
115. extern void checkSTREAMresults();
116. #ifdef TUNED
117. extern void tuned_STREAM_Copy();
118. extern void tuned_STREAM_Scale(double scalar);
119. extern void tuned_STREAM_Add();
120. extern void tuned_STREAM_Triad(double scalar);
121. #endif
122. #ifdef _OPENMP
123. extern int omp_get_num_threads();
124. #endif
125. int
126. main()
127. {
128. int quantum, checktick();
129. int BytesPerWord;
130. register int j, k;
131. double scalar, t, times[4][NTIMES];
132.
133. /* --- SETUP --- determine precision and check timing --- */
134.
135. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 135, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
136. + printf("STREAM version $Revision: 5.9 $\n");
^
CC-3021 CC: IPA main, File = stream.c, Line = 136, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
137. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 137, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
138. BytesPerWord = sizeof(double);
139. + printf("This system uses %d bytes per DOUBLE PRECISION word.\n",
^
CC-3021 CC: IPA main, File = stream.c, Line = 139, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
140. BytesPerWord);
141.
142. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 142, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
143. #ifdef NO_LONG_LONG
144. printf("Array size = %d, Offset = %d\n" , N, OFFSET);
145. #else
146. + printf("Array size = %llu, Offset = %d\n", (unsigned long long) N, OFFSET);
^
CC-3021 CC: IPA main, File = stream.c, Line = 146, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
147. #endif
148.
149. + printf("Total memory required = %.1f MB.\n",
^
CC-3021 CC: IPA main, File = stream.c, Line = 149, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
150. (3.0 * BytesPerWord) * ( (double) N / 1048576.0));
151. + printf("Each test is run %d times, but only\n", NTIMES);
^
CC-3021 CC: IPA main, File = stream.c, Line = 151, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
152. + printf("the *best* time for each is used.\n");
^
CC-3021 CC: IPA main, File = stream.c, Line = 152, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
153.
154. #ifdef _OPENMP
155. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 155, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
156. #pragma omp parallel
157. M------------< {
CC-6823 CC: THREAD main, File = stream.c, Line = 157
A region starting at line 157 and ending at line 163 was multi-threaded.
158. M #pragma omp master
159. M {
160. + M k = omp_get_num_threads();
^
CC-3021 CC: IPA main, File = stream.c, Line = 160, Column = 6
"omp_get_num_threads" (called from "main") was not inlined because the compiler was unable to locate the routine.
161. + M printf ("Number of Threads requested = %i\n",k);
^
CC-3021 CC: IPA main, File = stream.c, Line = 161, Column = 6
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
162. M }
163. M------------> }
164. #endif
165.
166. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 166, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
167. #pragma omp parallel
168. + M-< {
CC-6831 CC: THREAD main, File = stream.c, Line = 168
An expanded multi-threaded region was created starting near line 168 and ending near line 178.
CC-6824 CC: THREAD main, File = stream.c, Line = 168
A region starting at line 168 and ending at line 170 was multi-threaded and merged with an expanded multi-thread region.
169. + M printf ("Printing one line per active thread....\n");
^
CC-3021 CC: IPA main, File = stream.c, Line = 169, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
170. M }
171. M
172. M /* Get initial value for system clock. */
173. M #pragma omp parallel for
174. + M mA-----------< for (j=0; j<N; j++) {
CC-6230 CC: VECTOR main, File = stream.c, Line = 174
A loop was replaced with multiple library calls.
CC-6824 CC: THREAD main, File = stream.c, Line = 174
A region starting at line 174 and ending at line 178 was multi-threaded and merged with an expanded multi-thread region.
CC-6817 CC: THREAD main, File = stream.c, Line = 174
A loop was partitioned.
175. M mA a[j] = 1.0;
176. M mA b[j] = 2.0;
177. M mA c[j] = 0.0;
178. M->mA-----------> }
179.
180. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 180, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
181.
182. + if ( (quantum = checktick()) >= 1)
^
CC-3118 CC: IPA main, File = stream.c, Line = 182, Column = 5
"checktick" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
183. + printf("Your clock granularity/precision appears to be "
^
CC-3021 CC: IPA main, File = stream.c, Line = 183, Column = 2
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
184. "%d microseconds.\n", quantum);
185. else {
186. + printf("Your clock granularity appears to be "
^
CC-3021 CC: IPA main, File = stream.c, Line = 186, Column = 2
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
187. "less than one microsecond.\n");
188. quantum = 1;
189. }
190.
191. + t = mysecond();
^
CC-3118 CC: IPA main, File = stream.c, Line = 191, Column = 5
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
192. #pragma omp parallel for
193. MmVr4--------< for (j = 0; j < N; j++)
CC-6005 CC: SCALAR main, File = stream.c, Line = 193
A loop was unrolled 4 times.
CC-6823 CC: THREAD main, File = stream.c, Line = 193
A region starting at line 193 and ending at line 194 was multi-threaded.
CC-6204 CC: VECTOR main, File = stream.c, Line = 193
A loop was vectorized.
CC-6817 CC: THREAD main, File = stream.c, Line = 193
A loop was partitioned.
194. MmVr4--------> a[j] = 2.0E0 * a[j];
195. + t = 1.0E6 * (mysecond() - t);
^
CC-3118 CC: IPA main, File = stream.c, Line = 195, Column = 5
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
196.
197. + printf("Each test below will take on the order"
^
CC-3021 CC: IPA main, File = stream.c, Line = 197, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
198. " of %d microseconds.\n", (int) t );
199. + printf(" (= %d clock ticks)\n", (int) (t/quantum) );
^
CC-3021 CC: IPA main, File = stream.c, Line = 199, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
200. + printf("Increase the size of the arrays if this shows that\n");
^
CC-3021 CC: IPA main, File = stream.c, Line = 200, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
201. + printf("you are not getting at least 20 clock ticks per test.\n");
^
CC-3021 CC: IPA main, File = stream.c, Line = 201, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
202.
203. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 203, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
204.
205. + printf("WARNING -- The above is only a rough guideline.\n");
^
CC-3021 CC: IPA main, File = stream.c, Line = 205, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
206. + printf("For best results, please be sure you know the\n");
^
CC-3021 CC: IPA main, File = stream.c, Line = 206, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
207. + printf("precision of your system timer.\n");
^
CC-3021 CC: IPA main, File = stream.c, Line = 207, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
208. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 208, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
209.
210. /* --- MAIN LOOP --- repeat test cases NTIMES times --- */
211.
212. scalar = 3.0;
213. + 1------------< for (k=0; k<NTIMES; k++)
CC-6287 CC: VECTOR main, File = stream.c, Line = 213
A loop was not vectorized because it contains a call to function "mysecond" on line 215.
214. 1 {
215. + 1 times[0][k] = mysecond();
^
CC-3118 CC: IPA main, File = stream.c, Line = 215, Column = 2
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
216. 1 #ifdef TUNED
217. 1 MmA I-----<> tuned_STREAM_Copy();
CC-6202 CC: VECTOR main, File = stream.c, Line = 217
A loop was replaced by a library call.
CC-6823 CC: THREAD main, File = stream.c, Line = 217
A region starting at line 217 and ending at line 217 was multi-threaded.
CC-6817 CC: THREAD main, File = stream.c, Line = 217
A loop was partitioned.
^
CC-3001 CC: IPA main, File = stream.c, Line = 217, Column = 9
The call to tiny leaf routine "tuned_STREAM_Copy" was textually inlined.
218. 1 #else
219. 1 #pragma omp parallel for
220. 1 for (j=0; j<N; j++)
221. 1 c[j] = a[j];
222. 1 #endif
223. + 1 times[0][k] = mysecond() - times[0][k];
^
CC-3118 CC: IPA main, File = stream.c, Line = 223, Column = 2
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
224. 1
225. + 1 times[1][k] = mysecond();
^
CC-3118 CC: IPA main, File = stream.c, Line = 225, Column = 2
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
226. 1 #ifdef TUNED
227. 1 MmVr4 I---<> tuned_STREAM_Scale(scalar);
CC-6005 CC: SCALAR main, File = stream.c, Line = 227
A loop was unrolled 4 times.
CC-6823 CC: THREAD main, File = stream.c, Line = 227
A region starting at line 227 and ending at line 227 was multi-threaded.
CC-6204 CC: VECTOR main, File = stream.c, Line = 227
A loop was vectorized.
CC-6817 CC: THREAD main, File = stream.c, Line = 227
A loop was partitioned.
^
CC-3001 CC: IPA main, File = stream.c, Line = 227, Column = 9
The call to tiny leaf routine "tuned_STREAM_Scale" was textually inlined.
228. 1 #else
229. 1 #pragma omp parallel for
230. 1 for (j=0; j<N; j++)
231. 1 b[j] = scalar*c[j];
232. 1 #endif
233. + 1 times[1][k] = mysecond() - times[1][k];
^
CC-3118 CC: IPA main, File = stream.c, Line = 233, Column = 2
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
234. 1
235. + 1 times[2][k] = mysecond();
^
CC-3118 CC: IPA main, File = stream.c, Line = 235, Column = 2
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
236. 1 #ifdef TUNED
237. 1 MmVr4 I---<> tuned_STREAM_Add();
CC-6005 CC: SCALAR main, File = stream.c, Line = 237
A loop was unrolled 4 times.
CC-6823 CC: THREAD main, File = stream.c, Line = 237
A region starting at line 237 and ending at line 237 was multi-threaded.
CC-6204 CC: VECTOR main, File = stream.c, Line = 237
A loop was vectorized.
CC-6817 CC: THREAD main, File = stream.c, Line = 237
A loop was partitioned.
^
CC-3001 CC: IPA main, File = stream.c, Line = 237, Column = 9
The call to tiny leaf routine "tuned_STREAM_Add" was textually inlined.
238. 1 #else
239. 1 #pragma omp parallel for
240. 1 for (j=0; j<N; j++)
241. 1 c[j] = a[j]+b[j];
242. 1 #endif
243. + 1 times[2][k] = mysecond() - times[2][k];
^
CC-3118 CC: IPA main, File = stream.c, Line = 243, Column = 2
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
244. 1
245. + 1 times[3][k] = mysecond();
^
CC-3118 CC: IPA main, File = stream.c, Line = 245, Column = 2
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
246. 1 #ifdef TUNED
247. 1 MmVr4 I---<> tuned_STREAM_Triad(scalar);
CC-6005 CC: SCALAR main, File = stream.c, Line = 247
A loop was unrolled 4 times.
CC-6823 CC: THREAD main, File = stream.c, Line = 247
A region starting at line 247 and ending at line 247 was multi-threaded.
CC-6204 CC: VECTOR main, File = stream.c, Line = 247
A loop was vectorized.
CC-6817 CC: THREAD main, File = stream.c, Line = 247
A loop was partitioned.
^
CC-3001 CC: IPA main, File = stream.c, Line = 247, Column = 9
The call to tiny leaf routine "tuned_STREAM_Triad" was textually inlined.
248. 1 #else
249. 1 #pragma omp parallel for
250. 1 for (j=0; j<N; j++)
251. 1 a[j] = b[j]+scalar*c[j];
252. 1 #endif
253. + 1 times[3][k] = mysecond() - times[3][k];
^
CC-3118 CC: IPA main, File = stream.c, Line = 253, Column = 2
"mysecond" (called from "main") was not inlined because the call site will not flatten. "gettimeofday" is missing.
254. 1------------> }
255.
256. /* --- SUMMARY --- */
257.
258. + iVw----------< for (k=1; k<NTIMES; k++) /* note -- skip first iteration */
CC-6007 CC: SCALAR main, File = stream.c, Line = 258
A loop was interchanged with the loop starting at line 260.
CC-6373 CC: VECTOR main, File = stream.c, Line = 258
A loop with a trip count of 9 was unwound into 2 vector iterations.
CC-6382 CC: VECTOR main, File = stream.c, Line = 258
A loop was partially vector pipelined.
CC-6204 CC: VECTOR main, File = stream.c, Line = 258
A loop was vectorized.
259. iVw {
260. + iVw i--------< for (j=0; j<4; j++)
CC-6294 CC: VECTOR main, File = stream.c, Line = 260
A loop was not vectorized because a better candidate was found at line 258.
261. iVw i {
262. iVw i avgtime[j] = avgtime[j] + times[j][k];
263. iVw i mintime[j] = MIN(mintime[j], times[j][k]);
264. iVw i maxtime[j] = MAX(maxtime[j], times[j][k]);
265. iVw i--------> }
266. iVw----------> }
267.
268. + printf("Function Rate (MB/s) Avg time Min time Max time\n");
^
CC-3021 CC: IPA main, File = stream.c, Line = 268, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
269. + 1------------< for (j=0; j<4; j++) {
CC-6287 CC: VECTOR main, File = stream.c, Line = 269
A loop was not vectorized because it contains a call to function "printf" on line 272.
270. 1 avgtime[j] = avgtime[j]/(double)(NTIMES-1);
271. 1
272. + 1 printf("%s%11.4f %11.4f %11.4f %11.4f\n", label[j],
^
CC-3021 CC: IPA main, File = stream.c, Line = 272, Column = 2
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
273. 1 1.0E-06 * bytes[j]/mintime[j],
274. 1 avgtime[j],
275. 1 mintime[j],
276. 1 maxtime[j]);
277. 1------------> }
278. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 278, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
279.
280. /* --- Check Results --- */
281. + checkSTREAMresults();
^
CC-3118 CC: IPA main, File = stream.c, Line = 281, Column = 5
"checkSTREAMresults" (called from "main") was not inlined because the call site will not flatten. "printf" is missing.
282. + printf(HLINE);
^
CC-3021 CC: IPA main, File = stream.c, Line = 282, Column = 5
"printf" (called from "main") was not inlined because the compiler was unable to locate the routine.
283.
284. return 0;
285. }
286.
287. # define M 20
288.
289. int
290. checktick()
291. {
292. int i, minDelta, Delta;
293. double t1, t2, timesfound[M];
294.
295. /* Collect a sequence of M unique time values from the system. */
296.
297. + 1------------< for (i = 0; i < M; i++) {
CC-6287 CC: VECTOR checktick, File = stream.c, Line = 297
A loop was not vectorized because it contains a call to function "mysecond" on line 298.
298. + 1 t1 = mysecond();
^
CC-3118 CC: IPA checktick, File = stream.c, Line = 298, Column = 2
"mysecond" (called from "checktick") was not inlined because the call site will not flatten. "gettimeofday" is missing.
299. + 1 2----------< while( ((t2=mysecond()) - t1) < 1.0E-6 )
CC-6287 CC: VECTOR checktick, File = stream.c, Line = 299
A loop was not vectorized because it contains a call to function "mysecond" on line 299.
^
CC-3118 CC: IPA checktick, File = stream.c, Line = 299, Column = 2
"mysecond" (called from "checktick") was not inlined because the call site will not flatten. "gettimeofday" is missing.
^
CC-3118 CC: IPA checktick, File = stream.c, Line = 299, Column = 2
"mysecond" (called from "checktick") was not inlined because the call site will not flatten. "gettimeofday" is missing.
300. 1 2----------> ;
301. 1 timesfound[i] = t1 = t2;
302. 1------------> }
303.
304. /*
305. * Determine the minimum difference between these M values.
306. * This result will be our estimate (in microseconds) for the
307. * clock granularity.
308. */
309.
310. minDelta = 1000000;
311. + Vw-----------< for (i = 1; i < M; i++) {
CC-6373 CC: VECTOR checktick, File = stream.c, Line = 311
A loop with a trip count of 19 was unwound into 4 vector iterations.
CC-6382 CC: VECTOR checktick, File = stream.c, Line = 311
A loop was partially vector pipelined.
CC-6204 CC: VECTOR checktick, File = stream.c, Line = 311
A loop was vectorized.
312. Vw Delta = (int)( 1.0E6 * (timesfound[i]-timesfound[i-1]));
313. Vw minDelta = MIN(minDelta, MAX(Delta,0));
314. Vw-----------> }
315.
316. return(minDelta);
317. }
318.
319.
320.
321. /* A gettimeofday routine to give access to the wall
322. clock timer on most UNIX-like systems. */
323.
324. #include <sys/time.h>
325.
326. double mysecond()
327. {
328. struct timeval tp;
329. struct timezone tzp;
330. int i;
331.
332. + i = gettimeofday(&tp,&tzp);
^
CC-3021 CC: IPA mysecond, File = stream.c, Line = 332, Column = 9
"gettimeofday" (called from "mysecond") was not inlined because the compiler was unable to locate the routine.
333. return ( (double) tp.tv_sec + (double) tp.tv_usec * 1.e-6 );
334. }
335.
336. void checkSTREAMresults ()
337. {
338. double aj,bj,cj,scalar;
339. double asum,bsum,csum;
340. double epsilon;
341. int j,k;
342.
343. /* reproduce initialization */
344. aj = 1.0;
345. bj = 2.0;
346. cj = 0.0;
347. /* a[] is modified during timing check */
348. aj = 2.0E0 * aj;
349. /* now execute timing loop */
350. scalar = 3.0;
351. V------------< for (k=0; k<NTIMES; k++)
CC-6204 CC: VECTOR checkSTREAMresults, File = stream.c, Line = 351
A loop was vectorized.
352. V {
353. V cj = aj;
354. V bj = scalar*cj;
355. V cj = aj+bj;
356. V aj = bj+scalar*cj;
357. V------------> }
358. aj = aj * (double) (N);
359. bj = bj * (double) (N);
360. cj = cj * (double) (N);
361.
362. asum = 0.0;
363. bsum = 0.0;
364. csum = 0.0;
365. Vr4----------< for (j=0; j<N; j++) {
CC-6005 CC: SCALAR checkSTREAMresults, File = stream.c, Line = 365
A loop was unrolled 4 times.
CC-6204 CC: VECTOR checkSTREAMresults, File = stream.c, Line = 365
A loop was vectorized.
366. Vr4 asum += a[j];
367. Vr4 bsum += b[j];
368. Vr4 csum += c[j];
369. Vr4----------> }
370. #ifdef VERBOSE
371. printf ("Results Comparison: \n");
372. printf (" Expected : %f %f %f \n",aj,bj,cj);
373. printf (" Observed : %f %f %f \n",asum,bsum,csum);
374. #endif
375.
376. #ifndef abs
377. #define abs(a) ((a) >= 0 ? (a) : -(a))
378. #endif
379. epsilon = 1.e-8;
380.
381. if (abs(aj-asum)/asum > epsilon) {
382. + printf ("Failed Validation on array a[]\n");
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 382, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
383. + printf (" Expected : %f \n",aj);
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 383, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
384. + printf (" Observed : %f \n",asum);
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 384, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
385. }
386. else if (abs(bj-bsum)/bsum > epsilon) {
387. + printf ("Failed Validation on array b[]\n");
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 387, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
388. + printf (" Expected : %f \n",bj);
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 388, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
389. + printf (" Observed : %f \n",bsum);
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 389, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
390. }
391. else if (abs(cj-csum)/csum > epsilon) {
392. + printf ("Failed Validation on array c[]\n");
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 392, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
393. + printf (" Expected : %f \n",cj);
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 393, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
394. + printf (" Observed : %f \n",csum);
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 394, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
395. }
396. else {
397. + printf ("Solution Validates\n");
^
CC-3021 CC: IPA checkSTREAMresults, File = stream.c, Line = 397, Column = 3
"printf" (called from "checkSTREAMresults") was not inlined because the compiler was unable to locate the routine.
398. }
399. }
400.
401. void tuned_STREAM_Copy()
402. {
403. int j;
404. #pragma omp parallel for
405. MmA----------< for (j=0; j<N; j++)
CC-6202 CC: VECTOR tuned_STREAM_Copy, File = stream.c, Line = 405
A loop was replaced by a library call.
CC-6823 CC: THREAD tuned_STREAM_Copy, File = stream.c, Line = 405
A region starting at line 405 and ending at line 406 was multi-threaded.
CC-6817 CC: THREAD tuned_STREAM_Copy, File = stream.c, Line = 405
A loop was partitioned.
406. MmA----------> c[j] = a[j];
407. }
408.
409. void tuned_STREAM_Scale(double scalar)
410. {
411. int j;
412. #pragma omp parallel for
413. MmVr4--------< for (j=0; j<N; j++)
CC-6005 CC: SCALAR tuned_STREAM_Scale, File = stream.c, Line = 413
A loop was unrolled 4 times.
CC-6823 CC: THREAD tuned_STREAM_Scale, File = stream.c, Line = 413
A region starting at line 413 and ending at line 414 was multi-threaded.
CC-6204 CC: VECTOR tuned_STREAM_Scale, File = stream.c, Line = 413
A loop was vectorized.
CC-6817 CC: THREAD tuned_STREAM_Scale, File = stream.c, Line = 413
A loop was partitioned.
414. MmVr4--------> b[j] = scalar*c[j];
415. }
416.
417. void tuned_STREAM_Add()
418. {
419. int j;
420. #pragma omp parallel for
421. MmVr4--------< for (j=0; j<N; j++)
CC-6005 CC: SCALAR tuned_STREAM_Add, File = stream.c, Line = 421
A loop was unrolled 4 times.
CC-6823 CC: THREAD tuned_STREAM_Add, File = stream.c, Line = 421
A region starting at line 421 and ending at line 422 was multi-threaded.
CC-6204 CC: VECTOR tuned_STREAM_Add, File = stream.c, Line = 421
A loop was vectorized.
CC-6817 CC: THREAD tuned_STREAM_Add, File = stream.c, Line = 421
A loop was partitioned.
422. MmVr4--------> c[j] = a[j]+b[j];
423. }
424.
425. void tuned_STREAM_Triad(double scalar)
426. {
427. int j;
428. #pragma omp parallel for
429. MmVr4--------< for (j=0; j<N; j++)
CC-6005 CC: SCALAR tuned_STREAM_Triad, File = stream.c, Line = 429
A loop was unrolled 4 times.
CC-6823 CC: THREAD tuned_STREAM_Triad, File = stream.c, Line = 429
A region starting at line 429 and ending at line 430 was multi-threaded.
CC-6204 CC: VECTOR tuned_STREAM_Triad, File = stream.c, Line = 429
A loop was vectorized.
CC-6817 CC: THREAD tuned_STREAM_Triad, File = stream.c, Line = 429
A loop was partitioned.
430. MmVr4--------> a[j] = b[j]+scalar*c[j];
431. }
432.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
arnoldg@h2ologin4:~/stream>
PGI compiler (pgcc)
arnoldg@h2ologin4:~/stream> cc -O3 -Minfo=all -mp -c -DTUNED stream.c
main:
157, Parallel region activated
158, Begin master region
163, End master region
166, Parallel region terminated
168, Parallel region activated
173, Parallel region terminated
174, Parallel region activated
Parallel loop activated with static block schedule
Generated an alternate version of the loop
Generated vector simd code for the loop
180, Barrier
Parallel region terminated
182, checktick inlined, size=24 (inline) file stream.c (291)
297, FMA (fused multiply-add) instruction(s) generated
298, mysecond inlined, size=4 (inline) file stream.c (327)
299, Loop not vectorized/parallelized: contains call
FMA (fused multiply-add) instruction(s) generated
191, mysecond inlined, size=4 (inline) file stream.c (327)
191, FMA (fused multiply-add) instruction(s) generated
193, Parallel region activated
Parallel loop activated with static block schedule
Generated vector simd code for the loop
Generated a prefetch instruction for the loop
195, mysecond inlined, size=4 (inline) file stream.c (327)
195, Barrier
Parallel region terminated
FMA (fused multiply-add) instruction(s) generated
213, Loop not vectorized/parallelized: contains call
FMA (fused multiply-add) instruction(s) generated
215, mysecond inlined, size=4 (inline) file stream.c (327)
223, mysecond inlined, size=4 (inline) file stream.c (327)
225, mysecond inlined, size=4 (inline) file stream.c (327)
233, mysecond inlined, size=4 (inline) file stream.c (327)
235, mysecond inlined, size=4 (inline) file stream.c (327)
243, mysecond inlined, size=4 (inline) file stream.c (327)
245, mysecond inlined, size=4 (inline) file stream.c (327)
253, mysecond inlined, size=4 (inline) file stream.c (327)
258, Loop not vectorized: data dependency
260, Loop unrolled 4 times (completely unrolled)
269, Loop not vectorized/parallelized: contains call
281, checkSTREAMresults inlined, size=34 (inline) file stream.c (337)
351, Loop unrolled 4 times
FMA (fused multiply-add) instruction(s) generated
365, Generated vector simd code for the loop containing reductions
Generated 3 prefetch instructions for the loop
checktick:
297, FMA (fused multiply-add) instruction(s) generated
298, mysecond inlined, size=4 (inline) file stream.c (327)
299, mysecond inlined, size=4 (inline) file stream.c (327)
299, Loop not vectorized/parallelized: contains call
FMA (fused multiply-add) instruction(s) generated
mysecond:
332, FMA (fused multiply-add) instruction(s) generated
checkSTREAMresults:
351, Loop unrolled 4 times
FMA (fused multiply-add) instruction(s) generated
365, Generated vector simd code for the loop containing reductions
Generated 3 prefetch instructions for the loop
tuned_STREAM_Copy:
405, Parallel region activated
Parallel loop activated with static block schedule
Memory copy idiom, loop replaced by call to __c_mcopy8
407, Barrier
Parallel region terminated
tuned_STREAM_Scale:
413, Parallel region activated
Parallel loop activated with static block schedule
Generated an alternate version of the loop
Generated vector simd code for the loop
Generated a prefetch instruction for the loop
Generated vector simd code for the loop
Generated a prefetch instruction for the loop
415, Barrier
Parallel region terminated
tuned_STREAM_Add:
421, Parallel region activated
Parallel loop activated with static block schedule
Generated an alternate version of the loop
Generated vector simd code for the loop
Generated 2 prefetch instructions for the loop
Generated vector simd code for the loop
Generated 2 prefetch instructions for the loop
423, Barrier
Parallel region terminated
tuned_STREAM_Triad:
429, Parallel region activated
Parallel loop activated with static block schedule
Generated an alternate version of the loop
Generated vector simd code for the loop
Generated 2 prefetch instructions for the loop
Generated vector simd code for the loop
Generated 2 prefetch instructions for the loop
FMA (fused multiply-add) instruction(s) generated
431, Barrier
Parallel region terminated
arnoldg@h2ologin4:~/stream>
Intel compiler (icc)
arnoldg@h2ologin4:~/stream> cc -c -qopt-report-annotate -DTUNED -O3 -qopenmp stream.c
arnoldg@h2ologin4:~/stream> cat stream.c.annot
//
// ------- Annotated listing with optimization reports for "/mnt/a/u/staff/arnoldg/stream/stream.c" -------
//
//INLINING OPTION VALUES:
// -inline-factor: 100
// -inline-min-size: 30
// -inline-max-size: 230
// -inline-max-total-size: 2000
// -inline-max-per-routine: 10000
// -inline-max-per-compile: 500000
//
1 /*-----------------------------------------------------------------------*/
2 /* Program: Stream */
3 /* Revision: $Id: stream.c,v 5.9 2009/04/11 16:35:00 mccalpin Exp $ */
4 /* Original code developed by John D. McCalpin */
5 /* Programmers: John D. McCalpin */
6 /* Joe R. Zagar */
7 /* */
8 /* This program measures memory transfer rates in MB/s for simple */
9 /* computational kernels coded in C. */
10 /*-----------------------------------------------------------------------*/
11 /* Copyright 1991-2005: John D. McCalpin */
12 /*-----------------------------------------------------------------------*/
13 /* License: */
14 /* 1. You are free to use this program and/or to redistribute */
15 /* this program. */
16 /* 2. You are free to modify this program for your own use, */
17 /* including commercial use, subject to the publication */
18 /* restrictions in item 3. */
19 /* 3. You are free to publish results obtained from running this */
20 /* program, or from works that you derive from this program, */
21 /* with the following limitations: */
22 /* 3a. In order to be referred to as "STREAM benchmark results", */
23 /* published results must be in conformance to the STREAM */
24 /* Run Rules, (briefly reviewed below) published at */
25 /* http://www.cs.virginia.edu/stream/ref.html */
26 /* and incorporated herein by reference. */
27 /* As the copyright holder, John McCalpin retains the */
28 /* right to determine conformity with the Run Rules. */
29 /* 3b. Results based on modified source code or on runs not in */
30 /* accordance with the STREAM Run Rules must be clearly */
31 /* labelled whenever they are published. Examples of */
32 /* proper labelling include: */
33 /* "tuned STREAM benchmark results" */
34 /* "based on a variant of the STREAM benchmark code" */
35 /* Other comparable, clear and reasonable labelling is */
36 /* acceptable. */
37 /* 3c. Submission of results to the STREAM benchmark web site */
38 /* is encouraged, but not required. */
39 /* 4. Use of this program or creation of derived works based on this */
40 /* program constitutes acceptance of these licensing restrictions. */
41 /* 5. Absolutely no warranty is expressed or implied. */
42 /*-----------------------------------------------------------------------*/
43 # include <stdio.h>
44 # include <math.h>
45 # include <float.h>
46 # include <limits.h>
47 # include <sys/time.h>
48
49 /* INSTRUCTIONS:
50 *
51 * 1) Stream requires a good bit of memory to run. Adjust the
52 * value of 'N' (below) to give a 'timing calibration' of
53 * at least 20 clock-ticks. This will provide rate estimates
54 * that should be good to about 5% precision.
55 */
56
57 #ifndef N
58 # define N 40000000
59 #endif
60 #ifndef NTIMES
61 # define NTIMES 10
62 #endif
63 #ifndef OFFSET
64 # define OFFSET 0
65 #endif
66
67 /*
68 * 3) Compile the code with full optimization. Many compilers
69 * generate unreasonably bad code before the optimizer tightens
70 * things up. If the results are unreasonably good, on the
71 * other hand, the optimizer might be too smart for me!
72 *
73 * Try compiling with:
74 * cc -O stream_omp.c -o stream_omp
75 *
76 * This is known to work on Cray, SGI, IBM, and Sun machines.
77 *
78 *
79 * 4) Mail the results to mccalpin@cs.virginia.edu
80 * Be sure to include:
81 * a) computer hardware model number and software revision
82 * b) the compiler flags
83 * c) all of the output from the test case.
84 * Thanks!
85 *
86 */
87
88 # define HLINE "-------------------------------------------------------------\n"
89
90 # ifndef MIN
91 # define MIN(x,y) ((x)<(y)?(x):(y))
92 # endif
93 # ifndef MAX
94 # define MAX(x,y) ((x)>(y)?(x):(y))
95 # endif
96
97 static double a[N+OFFSET],
98 b[N+OFFSET],
99 c[N+OFFSET];
100
101 static double avgtime[4] = {0}, maxtime[4] = {0},
102 mintime[4] = {FLT_MAX,FLT_MAX,FLT_MAX,FLT_MAX};
103
104 static char *label[4] = {"Copy: ", "Scale: ",
105 "Add: ", "Triad: "};
106
107 static double bytes[4] = {
108 2 * sizeof(double) * N,
109 2 * sizeof(double) * N,
110 3 * sizeof(double) * N,
111 3 * sizeof(double) * N
112 };
113
114 extern double mysecond();
115 extern void checkSTREAMresults();
116 #ifdef TUNED
117 extern void tuned_STREAM_Copy();
118 extern void tuned_STREAM_Scale(double scalar);
119 extern void tuned_STREAM_Add();
120 extern void tuned_STREAM_Triad(double scalar);
121 #endif
122 #ifdef _OPENMP
123 extern int omp_get_num_threads();
124 #endif
125 int
126 main()
127 {
//INLINE REPORT: (main()) [1] /mnt/a/u/staff/arnoldg/stream/stream.c(127,5)
// -> INLINE: (182,22) checktick()
// -> INLINE: (298,7) mysecond()
// -> INLINE: (299,14) mysecond()
// -> INLINE: (299,14) mysecond()
// -> INLINE: (191,9) mysecond()
// -> INLINE: (195,18) mysecond()
// -> INLINE: (215,16) mysecond()
// -> INLINE: (217,9) tuned_STREAM_Copy()
// -> INLINE: (223,16) mysecond()
// -> INLINE: (225,16) mysecond()
// -> INLINE: (227,9) tuned_STREAM_Scale(double)
// -> INLINE: (233,16) mysecond()
// -> INLINE: (235,16) mysecond()
// -> INLINE: (237,9) tuned_STREAM_Add()
// -> INLINE: (243,16) mysecond()
// -> INLINE: (245,16) mysecond()
// -> INLINE: (247,9) tuned_STREAM_Triad(double)
// -> INLINE: (253,16) mysecond()
// -> INLINE: (281,5) checkSTREAMresults()
//
///mnt/a/u/staff/arnoldg/stream/stream.c(127,5):remark #34051: REGISTER ALLOCATION : [main] /mnt/a/u/staff/arnoldg/stream/stream.c:127
//
// Hardware registers
// Reserved : 2[ rsp rip]
// Available : 39[ rax rdx rcx rbx rbp rsi rdi r8-r15 mm0-mm7 zmm0-zmm15]
// Callee-save : 6[ rbx rbp r12-r15]
// Assigned : 30[ rax rdx rcx rbx rsi rdi r8-r15 zmm0-zmm15]
//
// Routine temporaries
// Total : 1219
// Global : 168
// Local : 1051
// Regenerable : 448
// Spilled : 11
//
// Routine stack
// Variables : 916 bytes*
// Reads : 124 [3.24e+02 ~ 0.0%]
// Writes : 41 [1.40e+01 ~ 0.0%]
// Spills : 128 bytes*
// Reads : 63 [0.00e+00 ~ 0.0%]
// Writes : 57 [1.00e+01 ~ 0.0%]
//
// Notes
//
// *Non-overlapping variables and spills may share stack space,
// so the total stack size might be less than this.
//
//
128 int quantum, checktick();
129 int BytesPerWord;
130 register int j, k;
131 double scalar, t, times[4][NTIMES];
132
133 /* --- SETUP --- determine precision and check timing --- */
134
135 printf(HLINE);
136 printf("STREAM version $Revision: 5.9 $\n");
137 printf(HLINE);
138 BytesPerWord = sizeof(double);
139 printf("This system uses %d bytes per DOUBLE PRECISION word.\n",
140 BytesPerWord);
141
142 printf(HLINE);
143 #ifdef NO_LONG_LONG
144 printf("Array size = %d, Offset = %d\n" , N, OFFSET);
145 #else
146 printf("Array size = %llu, Offset = %d\n", (unsigned long long) N, OFFSET);
147 #endif
148
149 printf("Total memory required = %.1f MB.\n",
150 (3.0 * BytesPerWord) * ( (double) N / 1048576.0));
151 printf("Each test is run %d times, but only\n", NTIMES);
152 printf("the *best* time for each is used.\n");
153
154 #ifdef _OPENMP
155 printf(HLINE);
156 #pragma omp parallel
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(156,1)
//remark #16201: OpenMP DEFINED REGION WAS PARALLELIZED
157 {
158 #pragma omp master
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(158,1)
//remark #16205: OpenMP multithreaded code generation for MASTER was successful
159 {
160 k = omp_get_num_threads();
161 printf ("Number of Threads requested = %i\n",k);
162 }
163 }
164 #endif
165
166 printf(HLINE);
167 #pragma omp parallel
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(167,1)
//remark #16201: OpenMP DEFINED REGION WAS PARALLELIZED
168 {
169 printf ("Printing one line per active thread....\n");
170 }
171
172 /* Get initial value for system clock. */
173 #pragma omp parallel for
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(173,1)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(173,1)
//<Peeled loop for vectorization>
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(173,1)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(173,1)
//<Remainder loop for vectorization>
//LOOP END
174 for (j=0; j<N; j++) {
175 a[j] = 1.0;
176 b[j] = 2.0;
177 c[j] = 0.0;
178 }
179
180 printf(HLINE);
181
182 if ( (quantum = checktick()) >= 1)
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(297,5) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(182,22)
// remark #15344: loop was not vectorized: vector dependence prevents vectorization. First dependence is shown below. Use level 5 report for details
// remark #15346: vector dependence: assumed OUTPUT dependence between call:gettimeofday(struct timeval *__restrict__, __timezone_ptr_t (332:13) and call:gettimeofday(struct timeval *__restrict__, __timezone_ptr_t (332:13)
//
// LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(299,2) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(182,22)
// remark #15521: loop was not vectorized: loop control variable was not identified. Explicitly compute the iteration count before executing the loop or try using canonical loop form from OpenMP specification
// LOOP END
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(311,5) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(182,22)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(311,5) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(182,22)
//<Remainder loop for vectorization>
// remark #25436: completely unrolled by 3
//LOOP END
183 printf("Your clock granularity/precision appears to be "
184 "%d microseconds.\n", quantum);
185 else {
186 printf("Your clock granularity appears to be "
187 "less than one microsecond.\n");
188 quantum = 1;
189 }
190
191 t = mysecond();
192 #pragma omp parallel for
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(192,1)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(192,1)
//<Peeled loop for vectorization>
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(192,1)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(192,1)
//<Remainder loop for vectorization>
//LOOP END
193 for (j = 0; j < N; j++)
194 a[j] = 2.0E0 * a[j];
195 t = 1.0E6 * (mysecond() - t);
196
197 printf("Each test below will take on the order"
198 " of %d microseconds.\n", (int) t );
199 printf(" (= %d clock ticks)\n", (int) (t/quantum) );
200 printf("Increase the size of the arrays if this shows that\n");
201 printf("you are not getting at least 20 clock ticks per test.\n");
202
203 printf(HLINE);
204
205 printf("WARNING -- The above is only a rough guideline.\n");
206 printf("For best results, please be sure you know the\n");
207 printf("precision of your system timer.\n");
208 printf(HLINE);
209
210 /* --- MAIN LOOP --- repeat test cases NTIMES times --- */
211
212 scalar = 3.0;
213 for (k=0; k<NTIMES; k++)
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(213,5)
// remark #15521: loop was not vectorized: loop control variable was not identified. Explicitly compute the iteration count before executing the loop or try using canonical loop form from OpenMP specification
//LOOP END
214 {
215 times[0][k] = mysecond();
216 #ifdef TUNED
217 tuned_STREAM_Copy();
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(404,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(217,9)
// remark #25399: memcopy generated
// remark #15398: loop was not vectorized: loop was transformed to memset or memcpy
//
// LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(404,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(217,9)
// remark #15335: loop was not vectorized: vectorization possible but seems inefficient. Use vector always directive or -vec-threshold0 to override
// remark #25439: unrolled with remainder by 2
// LOOP END
//
// LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(404,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(217,9)
// <Remainder>
// LOOP END
//LOOP END
218 #else
219 #pragma omp parallel for
220 for (j=0; j<N; j++)
221 c[j] = a[j];
222 #endif
223 times[0][k] = mysecond() - times[0][k];
224
225 times[1][k] = mysecond();
226 #ifdef TUNED
227 tuned_STREAM_Scale(scalar);
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(412,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(227,9)
//<Peeled loop for vectorization>
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(412,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(227,9)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(412,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(227,9)
//<Remainder loop for vectorization>
//LOOP END
228 #else
229 #pragma omp parallel for
230 for (j=0; j<N; j++)
231 b[j] = scalar*c[j];
232 #endif
233 times[1][k] = mysecond() - times[1][k];
234
235 times[2][k] = mysecond();
236 #ifdef TUNED
237 tuned_STREAM_Add();
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(420,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(237,9)
//<Peeled loop for vectorization>
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(420,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(237,9)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(420,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(237,9)
//<Remainder loop for vectorization>
//LOOP END
238 #else
239 #pragma omp parallel for
240 for (j=0; j<N; j++)
241 c[j] = a[j]+b[j];
242 #endif
243 times[2][k] = mysecond() - times[2][k];
244
245 times[3][k] = mysecond();
246 #ifdef TUNED
247 tuned_STREAM_Triad(scalar);
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(428,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(247,9)
//<Peeled loop for vectorization>
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(428,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(247,9)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(428,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(247,9)
//<Remainder loop for vectorization>
//LOOP END
248 #else
249 #pragma omp parallel for
250 for (j=0; j<N; j++)
251 a[j] = b[j]+scalar*c[j];
252 #endif
253 times[3][k] = mysecond() - times[3][k];
254 }
255
256 /* --- SUMMARY --- */
257
258 for (k=1; k<NTIMES; k++) /* note -- skip first iteration */
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(258,5)
// remark #25461: Imperfect Loop Unroll-Jammed by 4 (pre-vector)
// remark #25045: Fused Loops: ( 258 258 )
//
// remark #25084: Preprocess Loopnests: Moving Out Store [ /mnt/a/u/staff/arnoldg/stream/stream.c(258,25) ]
// remark #15335: loop was not vectorized: vectorization possible but seems inefficient. Use vector always directive or -vec-threshold0 to override
// remark #25436: completely unrolled by 9
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(258,5)
//<Distributed chunk2>
// remark #25046: Loop lost in Fusion
//LOOP END
259 {
260 for (j=0; j<4; j++)
261 {
262 avgtime[j] = avgtime[j] + times[j][k];
263 mintime[j] = MIN(mintime[j], times[j][k]);
264 maxtime[j] = MAX(maxtime[j], times[j][k]);
265 }
266 }
267
268 printf("Function Rate (MB/s) Avg time Min time Max time\n");
269 for (j=0; j<4; j++) {
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(269,5)
// remark #15344: loop was not vectorized: vector dependence prevents vectorization. First dependence is shown below. Use level 5 report for details
// remark #25436: completely unrolled by 4
//LOOP END
270 avgtime[j] = avgtime[j]/(double)(NTIMES-1);
271
272 printf("%s%11.4f %11.4f %11.4f %11.4f\n", label[j],
273 1.0E-06 * bytes[j]/mintime[j],
274 avgtime[j],
275 mintime[j],
276 maxtime[j]);
277 }
278 printf(HLINE);
279
280 /* --- Check Results --- */
281 checkSTREAMresults();
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(351,2) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(281,5)
// remark #15344: loop was not vectorized: vector dependence prevents vectorization. First dependence is shown below. Use level 5 report for details
// remark #15346: vector dependence: assumed ANTI dependence between aj (354:13) and aj (356:13)
// remark #25436: completely unrolled by 10
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(365,2) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(281,5)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
282 printf(HLINE);
283
284 return 0;
285 }
286
287 # define M 20
288
289 int
290 checktick()
291 {
//INLINE REPORT: (checktick()) [2] /mnt/a/u/staff/arnoldg/stream/stream.c(291,5)
// -> INLINE: (298,7) mysecond()
// -> INLINE: (299,14) mysecond()
// -> INLINE: (299,14) mysecond()
//
///mnt/a/u/staff/arnoldg/stream/stream.c(291,5):remark #34051: REGISTER ALLOCATION : [checktick] /mnt/a/u/staff/arnoldg/stream/stream.c:291
//
// Hardware registers
// Reserved : 2[ rsp rip]
// Available : 39[ rax rdx rcx rbx rbp rsi rdi r8-r15 mm0-mm7 zmm0-zmm15]
// Callee-save : 6[ rbx rbp r12-r15]
// Assigned : 22[ rax rdx rcx rbp rsi rdi zmm0-zmm15]
//
// Routine temporaries
// Total : 144
// Global : 16
// Local : 128
// Regenerable : 18
// Spilled : 2
//
// Routine stack
// Variables : 232 bytes*
// Reads : 26 [3.00e+02 ~ 12.0%]
// Writes : 1 [2.00e+01 ~ 0.8%]
// Spills : 8 bytes*
// Reads : 2 [1.20e+02 ~ 4.8%]
// Writes : 1 [2.00e+01 ~ 0.8%]
//
// Notes
//
// *Non-overlapping variables and spills may share stack space,
// so the total stack size might be less than this.
//
//
292 int i, minDelta, Delta;
293 double t1, t2, timesfound[M];
294
295 /* Collect a sequence of M unique time values from the system. */
296
297 for (i = 0; i < M; i++) {
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(297,5)
// remark #15344: loop was not vectorized: vector dependence prevents vectorization. First dependence is shown below. Use level 5 report for details
// remark #15346: vector dependence: assumed OUTPUT dependence between call:gettimeofday(struct timeval *__restrict__, __timezone_ptr_t (332:13) and call:gettimeofday(struct timeval *__restrict__, __timezone_ptr_t (332:13)
//
// LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(299,2)
// remark #15521: loop was not vectorized: loop control variable was not identified. Explicitly compute the iteration count before executing the loop or try using canonical loop form from OpenMP specification
// LOOP END
//LOOP END
298 t1 = mysecond();
299 while( ((t2=mysecond()) - t1) < 1.0E-6 )
300 ;
301 timesfound[i] = t1 = t2;
302 }
303
304 /*
305 * Determine the minimum difference between these M values.
306 * This result will be our estimate (in microseconds) for the
307 * clock granularity.
308 */
309
310 minDelta = 1000000;
311 for (i = 1; i < M; i++) {
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(311,5)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(311,5)
//<Remainder loop for vectorization>
// remark #25436: completely unrolled by 3
//LOOP END
312 Delta = (int)( 1.0E6 * (timesfound[i]-timesfound[i-1]));
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
///mnt/a/u/staff/arnoldg/stream/stream.c(312,26):remark #34055: adjacent dense (unit-strided stencil) loads are not optimized. Details: stride { 8 }, step { 8 }, types { F64-V128, F64-V128 }, number of elements { 2 }, select mask { 0x000000003 }.
313 minDelta = MIN(minDelta, MAX(Delta,0));
314 }
315
316 return(minDelta);
317 }
318
319
320
321 /* A gettimeofday routine to give access to the wall
322 clock timer on most UNIX-like systems. */
323
324 #include <sys/time.h>
325
326 double mysecond()
327 {
//INLINE REPORT: (mysecond()) [3] /mnt/a/u/staff/arnoldg/stream/stream.c(327,1)
//
///mnt/a/u/staff/arnoldg/stream/stream.c(327,1):remark #34051: REGISTER ALLOCATION : [mysecond] /mnt/a/u/staff/arnoldg/stream/stream.c:327
//
// Hardware registers
// Reserved : 2[ rsp rip]
// Available : 39[ rax rdx rcx rbx rbp rsi rdi r8-r15 mm0-mm7 zmm0-zmm15]
// Callee-save : 6[ rbx rbp r12-r15]
// Assigned : 4[ rsi rdi zmm0-zmm1]
//
// Routine temporaries
// Total : 15
// Global : 6
// Local : 9
// Regenerable : 4
// Spilled : 0
//
// Routine stack
// Variables : 24 bytes*
// Reads : 2 [2.00e+00 ~ 9.1%]
// Writes : 0 [0.00e+00 ~ 0.0%]
// Spills : 0 bytes*
// Reads : 0 [0.00e+00 ~ 0.0%]
// Writes : 0 [0.00e+00 ~ 0.0%]
//
// Notes
//
// *Non-overlapping variables and spills may share stack space,
// so the total stack size might be less than this.
//
//
328 struct timeval tp;
329 struct timezone tzp;
330 int i;
331
332 i = gettimeofday(&tp,&tzp);
333 return ( (double) tp.tv_sec + (double) tp.tv_usec * 1.e-6 );
334 }
335
336 void checkSTREAMresults ()
337 {
//INLINE REPORT: (checkSTREAMresults()) [4] /mnt/a/u/staff/arnoldg/stream/stream.c(337,1)
//
///mnt/a/u/staff/arnoldg/stream/stream.c(337,1):remark #34051: REGISTER ALLOCATION : [checkSTREAMresults] /mnt/a/u/staff/arnoldg/stream/stream.c:337
//
// Hardware registers
// Reserved : 2[ rsp rip]
// Available : 39[ rax rdx rcx rbx rbp rsi rdi r8-r15 mm0-mm7 zmm0-zmm15]
// Callee-save : 6[ rbx rbp r12-r15]
// Assigned : 9[ rax rdi zmm0-zmm6]
//
// Routine temporaries
// Total : 73
// Global : 17
// Local : 56
// Regenerable : 33
// Spilled : 3
//
// Routine stack
// Variables : 0 bytes*
// Reads : 0 [0.00e+00 ~ 0.0%]
// Writes : 0 [0.00e+00 ~ 0.0%]
// Spills : 24 bytes*
// Reads : 4 [0.00e+00 ~ 0.0%]
// Writes : 3 [0.00e+00 ~ 0.0%]
//
// Notes
//
// *Non-overlapping variables and spills may share stack space,
// so the total stack size might be less than this.
//
//
338 double aj,bj,cj,scalar;
339 double asum,bsum,csum;
340 double epsilon;
341 int j,k;
342
343 /* reproduce initialization */
344 aj = 1.0;
345 bj = 2.0;
346 cj = 0.0;
347 /* a[] is modified during timing check */
348 aj = 2.0E0 * aj;
349 /* now execute timing loop */
350 scalar = 3.0;
351 for (k=0; k<NTIMES; k++)
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(351,2)
// remark #15344: loop was not vectorized: vector dependence prevents vectorization. First dependence is shown below. Use level 5 report for details
// remark #15346: vector dependence: assumed ANTI dependence between aj (354:13) and aj (356:13)
// remark #25436: completely unrolled by 10
//LOOP END
352 {
353 cj = aj;
354 bj = scalar*cj;
355 cj = aj+bj;
356 aj = bj+scalar*cj;
357 }
358 aj = aj * (double) (N);
359 bj = bj * (double) (N);
360 cj = cj * (double) (N);
361
362 asum = 0.0;
363 bsum = 0.0;
364 csum = 0.0;
365 for (j=0; j<N; j++) {
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(365,2)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
366 asum += a[j];
367 bsum += b[j];
368 csum += c[j];
369 }
370 #ifdef VERBOSE
371 printf ("Results Comparison: \n");
372 printf (" Expected : %f %f %f \n",aj,bj,cj);
373 printf (" Observed : %f %f %f \n",asum,bsum,csum);
374 #endif
375
376 #ifndef abs
377 #define abs(a) ((a) >= 0 ? (a) : -(a))
378 #endif
379 epsilon = 1.e-8;
380
381 if (abs(aj-asum)/asum > epsilon) {
382 printf ("Failed Validation on array a[]\n");
383 printf (" Expected : %f \n",aj);
384 printf (" Observed : %f \n",asum);
385 }
386 else if (abs(bj-bsum)/bsum > epsilon) {
387 printf ("Failed Validation on array b[]\n");
388 printf (" Expected : %f \n",bj);
389 printf (" Observed : %f \n",bsum);
390 }
391 else if (abs(cj-csum)/csum > epsilon) {
392 printf ("Failed Validation on array c[]\n");
393 printf (" Expected : %f \n",cj);
394 printf (" Observed : %f \n",csum);
395 }
396 else {
397 printf ("Solution Validates\n");
398 }
399 }
400
401 void tuned_STREAM_Copy()
402 {
//INLINE REPORT: (tuned_STREAM_Copy()) [5] /mnt/a/u/staff/arnoldg/stream/stream.c(402,1)
//
///mnt/a/u/staff/arnoldg/stream/stream.c(402,1):remark #34051: REGISTER ALLOCATION : [tuned_STREAM_Copy] /mnt/a/u/staff/arnoldg/stream/stream.c:402
//
// Hardware registers
// Reserved : 2[ rsp rip]
// Available : 39[ rax rdx rcx rbx rbp rsi rdi r8-r15 mm0-mm7 zmm0-zmm15]
// Callee-save : 6[ rbx rbp r12-r15]
// Assigned : 10[ rax rdx rcx rbx rbp rsi rdi r8-r10]
//
// Routine temporaries
// Total : 94
// Global : 18
// Local : 76
// Regenerable : 32
// Spilled : 0
//
// Routine stack
// Variables : 20 bytes*
// Reads : 4 [0.00e+00 ~ 0.0%]
// Writes : 5 [5.00e+00 ~ 0.0%]
// Spills : 48 bytes*
// Reads : 12 [0.00e+00 ~ 0.0%]
// Writes : 12 [1.20e+01 ~ 0.0%]
//
// Notes
//
// *Non-overlapping variables and spills may share stack space,
// so the total stack size might be less than this.
//
//
403 int j;
404 #pragma omp parallel for
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(404,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(217,9)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
///mnt/a/u/staff/arnoldg/stream/stream.c(404,1):remark #34026: call to memcpy implemented as a call to optimized library version
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(404,1)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(404,1)
// remark #25399: memcopy generated
// remark #15398: loop was not vectorized: loop was transformed to memset or memcpy
//
// LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(404,1)
// remark #15335: loop was not vectorized: vectorization possible but seems inefficient. Use vector always directive or -vec-threshold0 to override
// remark #25439: unrolled with remainder by 2
// LOOP END
//
// LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(404,1)
// <Remainder>
// LOOP END
//LOOP END
///mnt/a/u/staff/arnoldg/stream/stream.c(404,1):remark #34026: call to memcpy implemented as a call to optimized library version
405 for (j=0; j<N; j++)
406 c[j] = a[j];
407 }
408
409 void tuned_STREAM_Scale(double scalar)
410 {
//INLINE REPORT: (tuned_STREAM_Scale(double)) [6] /mnt/a/u/staff/arnoldg/stream/stream.c(410,1)
//
///mnt/a/u/staff/arnoldg/stream/stream.c(410,1):remark #34051: REGISTER ALLOCATION : [tuned_STREAM_Scale] /mnt/a/u/staff/arnoldg/stream/stream.c:410
//
// Hardware registers
// Reserved : 2[ rsp rip]
// Available : 39[ rax rdx rcx rbx rbp rsi rdi r8-r15 mm0-mm7 zmm0-zmm15]
// Callee-save : 6[ rbx rbp r12-r15]
// Assigned : 14[ rax rdx rcx rbx rbp rsi rdi r8-r11 zmm0-zmm2]
//
// Routine temporaries
// Total : 106
// Global : 19
// Local : 87
// Regenerable : 37
// Spilled : 1
//
// Routine stack
// Variables : 28 bytes*
// Reads : 4 [0.00e+00 ~ 0.0%]
// Writes : 6 [6.00e+00 ~ 0.0%]
// Spills : 56 bytes*
// Reads : 13 [1.00e+00 ~ 0.0%]
// Writes : 13 [1.30e+01 ~ 0.0%]
//
// Notes
//
// *Non-overlapping variables and spills may share stack space,
// so the total stack size might be less than this.
//
//
411 int j;
412 #pragma omp parallel for
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(412,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(227,9)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(412,1)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(412,1)
//<Peeled loop for vectorization>
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(412,1)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(412,1)
//<Remainder loop for vectorization>
//LOOP END
413 for (j=0; j<N; j++)
414 b[j] = scalar*c[j];
415 }
416
417 void tuned_STREAM_Add()
418 {
//INLINE REPORT: (tuned_STREAM_Add()) [7] /mnt/a/u/staff/arnoldg/stream/stream.c(418,1)
//
///mnt/a/u/staff/arnoldg/stream/stream.c(418,1):remark #34051: REGISTER ALLOCATION : [tuned_STREAM_Add] /mnt/a/u/staff/arnoldg/stream/stream.c:418
//
// Hardware registers
// Reserved : 2[ rsp rip]
// Available : 39[ rax rdx rcx rbx rbp rsi rdi r8-r15 mm0-mm7 zmm0-zmm15]
// Callee-save : 6[ rbx rbp r12-r15]
// Assigned : 12[ rax rdx rcx rbx rbp rsi rdi r8-r11 zmm0]
//
// Routine temporaries
// Total : 94
// Global : 17
// Local : 77
// Regenerable : 33
// Spilled : 0
//
// Routine stack
// Variables : 20 bytes*
// Reads : 4 [0.00e+00 ~ 0.0%]
// Writes : 5 [5.00e+00 ~ 0.0%]
// Spills : 48 bytes*
// Reads : 12 [0.00e+00 ~ 0.0%]
// Writes : 12 [1.20e+01 ~ 0.0%]
//
// Notes
//
// *Non-overlapping variables and spills may share stack space,
// so the total stack size might be less than this.
//
//
419 int j;
420 #pragma omp parallel for
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(420,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(237,9)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(420,1)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(420,1)
//<Peeled loop for vectorization>
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(420,1)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(420,1)
//<Remainder loop for vectorization>
//LOOP END
421 for (j=0; j<N; j++)
422 c[j] = a[j]+b[j];
423 }
424
425 void tuned_STREAM_Triad(double scalar)
426 {
//INLINE REPORT: (tuned_STREAM_Triad(double)) [8] /mnt/a/u/staff/arnoldg/stream/stream.c(426,1)
//
///mnt/a/u/staff/arnoldg/stream/stream.c(426,1):remark #34051: REGISTER ALLOCATION : [tuned_STREAM_Triad] /mnt/a/u/staff/arnoldg/stream/stream.c:426
//
// Hardware registers
// Reserved : 2[ rsp rip]
// Available : 39[ rax rdx rcx rbx rbp rsi rdi r8-r15 mm0-mm7 zmm0-zmm15]
// Callee-save : 6[ rbx rbp r12-r15]
// Assigned : 14[ rax rdx rcx rbx rbp rsi rdi r8-r11 zmm0-zmm2]
//
// Routine temporaries
// Total : 108
// Global : 19
// Local : 89
// Regenerable : 37
// Spilled : 1
//
// Routine stack
// Variables : 28 bytes*
// Reads : 4 [0.00e+00 ~ 0.0%]
// Writes : 6 [6.00e+00 ~ 0.0%]
// Spills : 56 bytes*
// Reads : 13 [1.00e+00 ~ 0.0%]
// Writes : 13 [1.30e+01 ~ 0.0%]
//
// Notes
//
// *Non-overlapping variables and spills may share stack space,
// so the total stack size might be less than this.
//
//
427 int j;
428 #pragma omp parallel for
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(428,1) inlined into /mnt/a/u/staff/arnoldg/stream/stream.c(247,9)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
//OpenMP Construct at /mnt/a/u/staff/arnoldg/stream/stream.c(428,1)
//remark #16200: OpenMP DEFINED LOOP WAS PARALLELIZED
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(428,1)
//<Peeled loop for vectorization>
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(428,1)
// remark #15300: LOOP WAS VECTORIZED
//LOOP END
//
//LOOP BEGIN at /mnt/a/u/staff/arnoldg/stream/stream.c(428,1)
//<Remainder loop for vectorization>
//LOOP END
429 for (j=0; j<N; j++)
430 a[j] = b[j]+scalar*c[j];
431 }
432
arnoldg@h2ologin4:~/stream>
GNU (gcc)
arnoldg@h2ologin4:~/stream> gcc -c -fopenmp -O3 -fopt-info -DTUNED stream.c stream.c:194:18: note: loop vectorized stream.c:194:18: note: loop peeled for vectorization to enhance alignment stream.c:192:9: note: loop turned into non-loop; it never loops stream.c:192:9: note: loop turned into non-loop; it never loops. stream.c:192:9: note: loop with 3 iterations completely unrolled stream.c:175:7: note: Loop 1 distributed: split to 1 loops and 1 library calls. stream.c:175:7: note: loop vectorized stream.c:175:7: note: loop peeled for vectorization to enhance alignment stream.c:173:9: note: loop turned into non-loop; it never loops stream.c:173:9: note: loop turned into non-loop; it never loops. stream.c:173:9: note: loop with 8 iterations completely unrolled stream.c:406:21: note: Loop 1 distributed: split to 0 loops and 1 library calls. stream.c:414:21: note: loop vectorized stream.c:414:21: note: loop peeled for vectorization to enhance alignment stream.c:412:9: note: loop turned into non-loop; it never loops stream.c:412:9: note: loop turned into non-loop; it never loops. stream.c:412:9: note: loop with 4 iterations completely unrolled stream.c:422:14: note: loop vectorized stream.c:422:14: note: loop peeled for vectorization to enhance alignment stream.c:420:9: note: loop turned into non-loop; it never loops stream.c:420:9: note: loop turned into non-loop; it never loops. stream.c:420:9: note: loop with 3 iterations completely unrolled stream.c:430:14: note: loop vectorized stream.c:430:14: note: loop peeled for vectorization to enhance alignment stream.c:428:9: note: loop turned into non-loop; it never loops stream.c:428:9: note: loop turned into non-loop; it never loops. stream.c:428:9: note: loop with 3 iterations completely unrolled stream.c:311:5: note: loop vectorized stream.c:311:5: note: loop turned into non-loop; it never loops. stream.c:311:5: note: loop with 3 iterations completely unrolled stream.c:290:1: note: loop turned into non-loop; it never loops stream.c:290:1: note: loop turned into non-loop; it never loops. stream.c:290:1: note: loop with 4 iterations completely unrolled stream.c:351:2: note: loop turned into non-loop; it never loops. stream.c:351:2: note: loop with 10 iterations completely unrolled stream.c:260:2: note: loop turned into non-loop; it never loops. stream.c:260:2: note: loop with 5 iterations completely unrolled stream.c:258:5: note: loop turned into non-loop; it never loops. stream.c:258:5: note: loop with 9 iterations completely unrolled arnoldg@h2ologin4:~/stream>
To confirm vectorization, compile to assembly code (gcc -S or similar ) and look for vector instructions. This may change with optimization levels.
counting vector instructions
arnoldg@h2ologin4:~/stream> gcc -c -fopenmp -O1 -fopt-info -DTUNED -S stream.c arnoldg@h2ologin4:~/stream> grep xmm stream.s | wc -l 138 arnoldg@h2ologin4:~/stream> gcc -c -fopenmp -O3 -fopt-info -DTUNED -S stream.c stream.c:194:18: note: loop vectorized stream.c:194:18: note: loop peeled for vectorization to enhance alignment stream.c:192:9: note: loop turned into non-loop; it never loops stream.c:192:9: note: loop turned into non-loop; it never loops. stream.c:192:9: note: loop with 3 iterations completely unrolled stream.c:175:7: note: Loop 1 distributed: split to 1 loops and 1 library calls. stream.c:175:7: note: loop vectorized stream.c:175:7: note: loop peeled for vectorization to enhance alignment stream.c:173:9: note: loop turned into non-loop; it never loops stream.c:173:9: note: loop turned into non-loop; it never loops. stream.c:173:9: note: loop with 8 iterations completely unrolled stream.c:406:21: note: Loop 1 distributed: split to 0 loops and 1 library calls. stream.c:414:21: note: loop vectorized stream.c:414:21: note: loop peeled for vectorization to enhance alignment stream.c:412:9: note: loop turned into non-loop; it never loops stream.c:412:9: note: loop turned into non-loop; it never loops. stream.c:412:9: note: loop with 4 iterations completely unrolled stream.c:422:14: note: loop vectorized stream.c:422:14: note: loop peeled for vectorization to enhance alignment stream.c:420:9: note: loop turned into non-loop; it never loops stream.c:420:9: note: loop turned into non-loop; it never loops. stream.c:420:9: note: loop with 3 iterations completely unrolled stream.c:430:14: note: loop vectorized stream.c:430:14: note: loop peeled for vectorization to enhance alignment stream.c:428:9: note: loop turned into non-loop; it never loops stream.c:428:9: note: loop turned into non-loop; it never loops. stream.c:428:9: note: loop with 3 iterations completely unrolled stream.c:311:5: note: loop vectorized stream.c:311:5: note: loop turned into non-loop; it never loops. stream.c:311:5: note: loop with 3 iterations completely unrolled stream.c:290:1: note: loop turned into non-loop; it never loops stream.c:290:1: note: loop turned into non-loop; it never loops. stream.c:290:1: note: loop with 4 iterations completely unrolled stream.c:351:2: note: loop turned into non-loop; it never loops. stream.c:351:2: note: loop with 10 iterations completely unrolled stream.c:260:2: note: loop turned into non-loop; it never loops. stream.c:260:2: note: loop with 5 iterations completely unrolled stream.c:258:5: note: loop turned into non-loop; it never loops. stream.c:258:5: note: loop with 9 iterations completely unrolled arnoldg@h2ologin4:~/stream> grep xmm stream.s | wc -l 524 arnoldg@h2ologin4:~/stream>