1 /*********************************************************************/
2 /* Copyright 2009, 2010 The University of Texas at Austin. */
3 /* All rights reserved. */
5 /* Redistribution and use in source and binary forms, with or */
6 /* without modification, are permitted provided that the following */
7 /* conditions are met: */
9 /* 1. Redistributions of source code must retain the above */
10 /* copyright notice, this list of conditions and the following */
13 /* 2. Redistributions in binary form must reproduce the above */
14 /* copyright notice, this list of conditions and the following */
15 /* disclaimer in the documentation and/or other materials */
16 /* provided with the distribution. */
18 /* THIS SOFTWARE IS PROVIDED BY THE UNIVERSITY OF TEXAS AT */
19 /* AUSTIN ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, */
20 /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */
21 /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */
22 /* DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY OF TEXAS AT */
23 /* AUSTIN OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, */
24 /* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES */
25 /* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE */
26 /* GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR */
27 /* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF */
28 /* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT */
29 /* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT */
30 /* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */
31 /* POSSIBILITY OF SUCH DAMAGE. */
33 /* The views and conclusions contained in the software and */
34 /* documentation are those of the authors and should not be */
35 /* interpreted as representing official policies, either expressed */
36 /* or implied, of The University of Texas at Austin. */
37 /*********************************************************************/
47 EXTERN gotoblas_t gotoblas_KATMAI;
48 EXTERN gotoblas_t gotoblas_COPPERMINE;
49 EXTERN gotoblas_t gotoblas_NORTHWOOD;
50 EXTERN gotoblas_t gotoblas_BANIAS;
51 EXTERN gotoblas_t gotoblas_ATHLON;
53 extern gotoblas_t gotoblas_PRESCOTT;
54 extern gotoblas_t gotoblas_ATOM;
55 extern gotoblas_t gotoblas_NANO;
56 extern gotoblas_t gotoblas_CORE2;
57 extern gotoblas_t gotoblas_PENRYN;
58 extern gotoblas_t gotoblas_DUNNINGTON;
59 extern gotoblas_t gotoblas_NEHALEM;
60 extern gotoblas_t gotoblas_OPTERON;
61 extern gotoblas_t gotoblas_OPTERON_SSE3;
62 extern gotoblas_t gotoblas_BARCELONA;
63 extern gotoblas_t gotoblas_BOBCAT;
65 extern gotoblas_t gotoblas_SANDYBRIDGE;
67 //Use NEHALEM kernels for sandy bridge
68 #define gotoblas_SANDYBRIDGE gotoblas_NEHALEM
72 #define VENDOR_INTEL 1
74 #define VENDOR_CENTAUR 3
75 #define VENDOR_UNKNOWN 99
77 #define BITMASK(a, b, c) ((((a) >> (b)) & (c)))
80 static inline void xgetbv(int op, int * eax, int * edx){
82 ("xgetbv": "=a" (*eax), "=d" (*edx) : "c" (op) : "cc");
88 int eax, ebx, ecx, edx;
91 cpuid(1, &eax, &ebx, &ecx, &edx);
92 if ((ecx & (1 << 28)) != 0 && (ecx & (1 << 27)) != 0){
93 xgetbv(0, &eax, &edx);
95 ret=1; //OS support AVX
104 static int get_vendor(void){
105 int eax, ebx, ecx, edx;
108 cpuid(0, &eax, &ebx, &ecx, &edx);
110 *(int *)(&vendor[0]) = ebx;
111 *(int *)(&vendor[4]) = edx;
112 *(int *)(&vendor[8]) = ecx;
113 vendor[12] = (char)0;
115 if (!strcmp(vendor, "GenuineIntel")) return VENDOR_INTEL;
116 if (!strcmp(vendor, "AuthenticAMD")) return VENDOR_AMD;
117 if (!strcmp(vendor, "CentaurHauls")) return VENDOR_CENTAUR;
119 if ((eax == 0) || ((eax & 0x500) != 0)) return VENDOR_INTEL;
121 return VENDOR_UNKNOWN;
124 static gotoblas_t *get_coretype(void){
126 int eax, ebx, ecx, edx;
127 int family, exfamily, model, vendor, exmodel;
129 cpuid(1, &eax, &ebx, &ecx, &edx);
131 family = BITMASK(eax, 8, 0x0f);
132 exfamily = BITMASK(eax, 20, 0xff);
133 model = BITMASK(eax, 4, 0x0f);
134 exmodel = BITMASK(eax, 16, 0x0f);
136 vendor = get_vendor();
138 if (vendor == VENDOR_INTEL){
143 if (model <= 0x7) return &gotoblas_KATMAI;
144 if ((model == 0x8) || (model == 0xa) || (model == 0xb)) return &gotoblas_COPPERMINE;
145 if ((model == 0x9) || (model == 0xd)) return &gotoblas_BANIAS;
146 if (model == 14) return &gotoblas_BANIAS;
147 if (model == 15) return &gotoblas_CORE2;
151 if (model == 6) return &gotoblas_CORE2;
152 if (model == 7) return &gotoblas_PENRYN;
153 if (model == 13) return &gotoblas_DUNNINGTON;
154 if ((model == 10) || (model == 11) || (model == 14) || (model == 15)) return &gotoblas_NEHALEM;
155 if (model == 12) return &gotoblas_ATOM;
159 //Intel Core (Clarkdale) / Core (Arrandale)
160 // Pentium (Clarkdale) / Pentium Mobile (Arrandale)
161 // Xeon (Clarkdale), 32nm
162 if (model == 5) return &gotoblas_NEHALEM;
164 //Intel Xeon Processor 5600 (Westmere-EP)
165 //Xeon Processor E7 (Westmere-EX)
166 if (model == 12 || model == 15) return &gotoblas_NEHALEM;
168 //Intel Core i5-2000 /i7-2000 (Sandy Bridge)
169 //Intel Core i7-3000 / Xeon E5
170 if (model == 10 || model == 13) {
172 return &gotoblas_SANDYBRIDGE;
174 fprintf(stderr, "OpenBLAS : Your OS doesn't support AVX. Use Nehalem kernels.\n");
175 return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
180 //Intel Sandy Bridge 22nm (Ivy Bridge?)
183 return &gotoblas_SANDYBRIDGE;
185 fprintf(stderr, "OpenBLAS : Your OS doesn't support AVX. Use Nehalem kernels.\n");
186 return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
192 if (model <= 0x2) return &gotoblas_NORTHWOOD;
193 return &gotoblas_PRESCOTT;
197 if (vendor == VENDOR_AMD){
198 if (family <= 0xe) return &gotoblas_ATHLON;
200 if ((exfamily == 0) || (exfamily == 2)) {
201 if (ecx & (1 << 0)) return &gotoblas_OPTERON_SSE3;
202 else return &gotoblas_OPTERON;
203 } else if (exfamily == 5) {
204 return &gotoblas_BOBCAT;
206 return &gotoblas_BARCELONA;
211 if (vendor == VENDOR_CENTAUR) {
214 return &gotoblas_NANO;
222 static char *corename[] = {
243 char *gotoblas_corename(void) {
245 if (gotoblas == &gotoblas_KATMAI) return corename[ 1];
246 if (gotoblas == &gotoblas_COPPERMINE) return corename[ 2];
247 if (gotoblas == &gotoblas_NORTHWOOD) return corename[ 3];
248 if (gotoblas == &gotoblas_PRESCOTT) return corename[ 4];
249 if (gotoblas == &gotoblas_BANIAS) return corename[ 5];
250 if (gotoblas == &gotoblas_ATOM) return corename[ 6];
251 if (gotoblas == &gotoblas_CORE2) return corename[ 7];
252 if (gotoblas == &gotoblas_PENRYN) return corename[ 8];
253 if (gotoblas == &gotoblas_DUNNINGTON) return corename[ 9];
254 if (gotoblas == &gotoblas_NEHALEM) return corename[10];
255 if (gotoblas == &gotoblas_ATHLON) return corename[11];
256 if (gotoblas == &gotoblas_OPTERON_SSE3) return corename[12];
257 if (gotoblas == &gotoblas_OPTERON) return corename[13];
258 if (gotoblas == &gotoblas_BARCELONA) return corename[14];
259 if (gotoblas == &gotoblas_NANO) return corename[15];
260 if (gotoblas == &gotoblas_SANDYBRIDGE) return corename[16];
261 if (gotoblas == &gotoblas_BOBCAT) return corename[17];
266 void gotoblas_dynamic_init(void) {
268 if (gotoblas) return;
270 gotoblas = get_coretype();
273 if (gotoblas == NULL) gotoblas = &gotoblas_KATMAI;
275 if (gotoblas == NULL) gotoblas = &gotoblas_PRESCOTT;
278 if (gotoblas && gotoblas -> init) {
281 fprintf(stderr, "OpenBLAS : Architecture Initialization failed. No initialization function found.\n");
287 void gotoblas_dynamic_quit(void) {