xref: /petsc/src/benchmarks/streams/SSEVersion.c (revision 5c8f6a953e7ed1c81f507d64aebddb11080b60e9)
1 static const char help[] = "STREAM benchmark specialized for SSE2\n\\n";
2 
3 /* Note: this file has been modified significantly from its original version */
4 #include <emmintrin.h>
5 #include <petsctime.h>
6 #include <petscsys.h>
7 #if defined(HAVE_NUMA)
8 #include <numa.h>
9 #endif
10 #include <limits.h>
11 #include <float.h>
12 
13 #if !defined(SSE2)
14 #  define SSE2 1
15 #endif
16 #if !defined(__SSE2__)
17 #  error SSE2 instruction set is not enabled, try adding -march=native to CFLAGS or disable by adding -DSSE2=0
18 #endif
19 #if !defined(PREFETCH_NTA) /* Use software prefetch and set non-temporal policy so that lines evicted from L1D will not subsequently reside in L2 or L3. */
20 #  define PREFETCH_NTA 1
21 #endif
22 #if !defined(STATIC_ALLOC) /* Statically allocate the vectors. Most platforms do not find physical pages when memory is allocated, therefore the faulting strategy still affects performance. */
23 #  define STATIC_ALLOC 0
24 #endif
25 #if !defined(FAULT_TOGETHER) /* Faults all three vectors together which usually interleaves DRAM pages in physical memory. */
26 #  define FAULT_TOGETHER 0
27 #endif
28 #if !defined(USE_MEMCPY) /* Literally call memcpy(3) for the COPY benchmark. Some compilers detect the unoptimized loop as memcpy and call this anyway. */
29 #  define USE_MEMCPY 0
30 #endif
31 
32 /*
33  * Program: Stream
34  * Programmer: Joe R. Zagar
35  * Revision: 4.0-BETA, October 24, 1995
36  * Original code developed by John D. McCalpin
37  *
38  * This program measures memory transfer rates in MB/s for simple
39  * computational kernels coded in C.  These numbers reveal the quality
40  * of code generation for simple uncacheable kernels as well as showing
41  * the cost of floating-point operations relative to memory accesses.
42  *
43  * INSTRUCTIONS:
44  *
45  *       1) Stream requires a good bit of memory to run.  Adjust the
46  *          value of 'N' (below) to give a 'timing calibration' of
47  *          at least 20 clock-ticks.  This will provide rate estimates
48  *          that should be good to about 5% precision.
49  */
50 
51 # define N      4000000
52 # define NTIMES     100
53 # define OFFSET       0
54 
55 # define HLINE "-------------------------------------------------------------\n"
56 
57 # if !defined(MIN)
58 # define MIN(x,y) ((x)<(y)?(x):(y))
59 # endif
60 # if !defined(MAX)
61 # define MAX(x,y) ((x)>(y)?(x):(y))
62 # endif
63 
64 #if STATIC_ALLOC
65   double a[N+OFFSET],b[N+OFFSET],c[N+OFFSET];
66 #endif
67 
68 static int checktick(void);
69 static double Second(void);
70 
71 int main(int argc,char *argv[])
72 {
73   const char *label[4] = {"Copy", "Scale","Add", "Triad"};
74   const double bytes[4] = {2 * sizeof(double) * N,
75                            2 * sizeof(double) * N,
76                            3 * sizeof(double) * N,
77                            3 * sizeof(double) * N};
78   double rmstime[4] = {0},maxtime[4] = {0},mintime[4] = {FLT_MAX,FLT_MAX,FLT_MAX,FLT_MAX};
79   int    quantum;
80   int    BytesPerWord,j,k,size;
81   PetscInt node = -1;
82   double   scalar, t, times[4][NTIMES];
83 #if !STATIC_ALLOC
84   double   *PETSC_RESTRICT a,*PETSC_RESTRICT b,*PETSC_RESTRICT c;
85 #endif
86 
87   PetscInitialize(&argc,&argv,0,help);
88   MPI_Comm_size(PETSC_COMM_WORLD,&size);
89   PetscOptionsGetInt(PETSC_NULL,"-node",&node,PETSC_NULL);
90   /* --- SETUP --- determine precision and check timing --- */
91 
92   PetscPrintf(PETSC_COMM_WORLD,HLINE);
93   BytesPerWord = sizeof(double);
94   PetscPrintf(PETSC_COMM_WORLD,"This system uses %d bytes per DOUBLE PRECISION word.\n",
95          BytesPerWord);
96 
97   PetscPrintf(PETSC_COMM_WORLD,HLINE);
98   PetscPrintf(PETSC_COMM_WORLD,"Array size = %d, Offset = %d\n" , N, OFFSET);
99   PetscPrintf(PETSC_COMM_WORLD,"Total memory required = %.1f MB per process.\n",
100          (3 * N * BytesPerWord) / 1048576.0);
101   PetscPrintf(PETSC_COMM_WORLD,"Each test is run %d times, but only\n", NTIMES);
102   PetscPrintf(PETSC_COMM_WORLD,"the *best* time for each is used.\n");
103 
104   /* Get initial value for system clock. */
105 
106 #if !STATIC_ALLOC
107   if (node == -1) {
108     posix_memalign((void**)&a,64,N*sizeof(double));
109     posix_memalign((void**)&b,64,N*sizeof(double));
110     posix_memalign((void**)&c,64,N*sizeof(double));
111   } else if (node == -2) {
112     a = malloc(N*sizeof(double));
113     b = malloc(N*sizeof(double));
114     c = malloc(N*sizeof(double));
115 #if defined(HAVE_NUMA)
116   } else {
117     a = numa_alloc_onnode(N*sizeof(double),node);
118     b = numa_alloc_onnode(N*sizeof(double),node);
119     c = numa_alloc_onnode(N*sizeof(double),node);
120 #endif
121   }
122 #endif
123 #if FAULT_TOGETHER
124   for (j=0; j<N; j++) {
125     a[j] = 1.0;
126     b[j] = 2.0;
127     c[j] = 0.0;
128   }
129 #else
130   for (j=0; j<N; j++) a[j] = 1.0;
131   for (j=0; j<N; j++) b[j] = 2.0;
132   for (j=0; j<N; j++) c[j] = 0.0;
133 #endif
134 
135   PetscPrintf(PETSC_COMM_WORLD,HLINE);
136 
137   if  ((quantum = checktick()) >= 1) {
138     PetscPrintf(PETSC_COMM_WORLD,"Your clock granularity/precision appears to be "
139            "%d microseconds.\n", quantum);
140   } else {
141     PetscPrintf(PETSC_COMM_WORLD,"Your clock granularity appears to be "
142            "less than one microsecond.\n");
143   }
144 
145   t = Second();
146   for (j = 0; j < N; j++)
147     a[j] = 2.0E0 * a[j];
148   t = 1.0E6 * (Second() - t);
149 
150   PetscPrintf(PETSC_COMM_WORLD,"Each test below will take on the order"
151          " of %d microseconds.\n", (int) t);
152   PetscPrintf(PETSC_COMM_WORLD,"   (= %d clock ticks)\n", (int) (t/quantum));
153   PetscPrintf(PETSC_COMM_WORLD,"Increase the size of the arrays if this shows that\n");
154   PetscPrintf(PETSC_COMM_WORLD,"you are not getting at least 20 clock ticks per test.\n");
155 
156   PetscPrintf(PETSC_COMM_WORLD,HLINE);
157 
158   PetscPrintf(PETSC_COMM_WORLD,"WARNING -- The above is only a rough guideline.\n");
159   PetscPrintf(PETSC_COMM_WORLD,"For best results, please be sure you know the\n");
160   PetscPrintf(PETSC_COMM_WORLD,"precision of your system timer.\n");
161   PetscPrintf(PETSC_COMM_WORLD,HLINE);
162 
163   /* --- MAIN LOOP --- repeat test cases NTIMES times --- */
164 
165   scalar = 3.0;
166   for (k=0; k<NTIMES; k++) {
167     MPI_Barrier(PETSC_COMM_WORLD);
168     /* ### COPY: c <- a ### */
169     times[0][k] = Second();
170     MPI_Barrier(PETSC_COMM_WORLD);
171 #if USE_MEMCPY
172     memcpy(c,a,N*sizeof(double));
173 #elif SSE2
174     for (j=0; j<N; j+=8) {
175       _mm_stream_pd(c+j+0,_mm_load_pd(a+j+0));
176       _mm_stream_pd(c+j+2,_mm_load_pd(a+j+2));
177       _mm_stream_pd(c+j+4,_mm_load_pd(a+j+4));
178       _mm_stream_pd(c+j+6,_mm_load_pd(a+j+6));
179 #  if PREFETCH_NTA
180       _mm_prefetch(a+j+64,_MM_HINT_NTA);
181 #  endif
182     }
183 #else
184     for (j=0; j<N; j++) c[j] = a[j];
185 #endif
186     MPI_Barrier(PETSC_COMM_WORLD);
187     times[0][k] = Second() - times[0][k];
188 
189     /* ### SCALE: b <- scalar * c ### */
190     times[1][k] = Second();
191     MPI_Barrier(PETSC_COMM_WORLD);
192 #if SSE2
193     {
194       __m128d scalar2 = _mm_set1_pd(scalar);
195       for (j=0; j<N; j+=8) {
196         _mm_stream_pd(b+j+0,_mm_mul_pd(scalar2,_mm_load_pd(c+j+0)));
197         _mm_stream_pd(b+j+2,_mm_mul_pd(scalar2,_mm_load_pd(c+j+2)));
198         _mm_stream_pd(b+j+4,_mm_mul_pd(scalar2,_mm_load_pd(c+j+4)));
199         _mm_stream_pd(b+j+6,_mm_mul_pd(scalar2,_mm_load_pd(c+j+6)));
200 #  if PREFETCH_NTA
201         _mm_prefetch(c+j+64,_MM_HINT_NTA);
202 #  endif
203       }
204     }
205 #else
206     for (j=0; j<N; j++) b[j] = scalar*c[j];
207 #endif
208     MPI_Barrier(PETSC_COMM_WORLD);
209     times[1][k] = Second() - times[1][k];
210 
211     /* ### ADD: c <- a + b ### */
212     times[2][k] = Second();
213     MPI_Barrier(PETSC_COMM_WORLD);
214 #if SSE2
215     {
216       for (j=0; j<N; j+=8) {
217         _mm_stream_pd(c+j+0,_mm_add_pd(_mm_load_pd(a+j+0),_mm_load_pd(b+j+0)));
218         _mm_stream_pd(c+j+2,_mm_add_pd(_mm_load_pd(a+j+2),_mm_load_pd(b+j+2)));
219         _mm_stream_pd(c+j+4,_mm_add_pd(_mm_load_pd(a+j+4),_mm_load_pd(b+j+4)));
220         _mm_stream_pd(c+j+6,_mm_add_pd(_mm_load_pd(a+j+6),_mm_load_pd(b+j+6)));
221 #  if PREFETCH_NTA
222         _mm_prefetch(a+j+64,_MM_HINT_NTA);
223         _mm_prefetch(b+j+64,_MM_HINT_NTA);
224 #  endif
225       }
226     }
227 #else
228     for (j=0; j<N; j++) c[j] = a[j]+b[j];
229 #endif
230     MPI_Barrier(PETSC_COMM_WORLD);
231     times[2][k] = Second() - times[2][k];
232 
233     /* ### TRIAD: a <- b + scalar * c ### */
234     times[3][k] = Second();
235     MPI_Barrier(PETSC_COMM_WORLD);
236 #if SSE2
237     {
238       __m128d scalar2 = _mm_set1_pd(scalar);
239       for (j=0; j<N; j+=8) {
240         _mm_stream_pd(a+j+0,_mm_add_pd(_mm_load_pd(b+j+0),_mm_mul_pd(scalar2,_mm_load_pd(c+j+0))));
241         _mm_stream_pd(a+j+2,_mm_add_pd(_mm_load_pd(b+j+2),_mm_mul_pd(scalar2,_mm_load_pd(c+j+2))));
242         _mm_stream_pd(a+j+4,_mm_add_pd(_mm_load_pd(b+j+4),_mm_mul_pd(scalar2,_mm_load_pd(c+j+4))));
243         _mm_stream_pd(a+j+6,_mm_add_pd(_mm_load_pd(b+j+6),_mm_mul_pd(scalar2,_mm_load_pd(c+j+6))));
244 #  if PREFETCH_NTA
245         _mm_prefetch(b+j+64,_MM_HINT_NTA);
246         _mm_prefetch(c+j+64,_MM_HINT_NTA);
247 #  endif
248       }
249     }
250 #else
251     for (j=0; j<N; j++) a[j] = b[j]+scalar*c[j];
252 #endif
253     MPI_Barrier(PETSC_COMM_WORLD);
254     times[3][k] = Second() - times[3][k];
255   }
256 
257   /* --- SUMMARY --- */
258 
259   for (k=0; k<NTIMES; k++) {
260     for (j=0; j<4; j++) {
261       rmstime[j] = rmstime[j] + (times[j][k] * times[j][k]);
262       mintime[j] = MIN(mintime[j], times[j][k]);
263       maxtime[j] = MAX(maxtime[j], times[j][k]);
264     }
265   }
266 
267 
268   PetscPrintf(PETSC_COMM_WORLD,"%8s:  %11s  %11s  %11s  %11s  %11s\n","Function","Rate (MB/s)","Total (MB/s)","RMS time","Min time","Max time");
269   for (j=0; j<4; j++) {
270     rmstime[j] = sqrt(rmstime[j]/(double)NTIMES);
271     PetscPrintf(PETSC_COMM_WORLD,"%8s: %11.4f  %11.4f  %11.4f  %11.4f  %11.4f\n", label[j], 1.0e-06*bytes[j]/mintime[j], size*1.0e-06*bytes[j]/mintime[j], rmstime[j], mintime[j], maxtime[j]);
272   }
273   PetscFinalize();
274   return 0;
275 }
276 
277 static double Second()
278 {
279   double t;
280   PetscTime(t);
281   return t;
282 }
283 
284 #define M 20
285 static int checktick()
286 {
287   int          i, minDelta, Delta;
288   double       t1, t2, timesfound[M];
289 
290   /*  Collect a sequence of M unique time values from the system. */
291 
292   for (i = 0; i < M; i++) {
293     t1 = Second();
294     while ((t2 = Second()) - t1 < 1.0E-6) {}
295     timesfound[i] = t1 = t2;
296   }
297 
298   /*
299    * Determine the minimum difference between these M values.
300    * This result will be our estimate (in microseconds) for the
301    * clock granularity.
302    */
303 
304   minDelta = 1000000;
305   for (i = 1; i < M; i++) {
306     Delta = (int)(1.0E6 * (timesfound[i]-timesfound[i-1]));
307     minDelta = MIN(minDelta, MAX(Delta,0));
308   }
309 
310   return(minDelta);
311 }
312