Table 6. The CPU time that each test job consumes on the dual-CPU AMD Athlon and AthlonMP systems. The unit of the CPU time is in minutes; the numbers in the parentheses are the step numbers required in the geometry optimization calculations. All jobs utilize only single CPU although two CPUs are installed.

System

TB2

TBMP2

Compiler

pgf77

pgf77

pgf77

pgf77

ifc

ifc

pgf77

pgf77

pgf77

pgf77

ifc

ifc

Arch. opt. a

-tp p6

-tp p6

-tp athlon

-tp athlon

-tpp6

-tpp6

-tp p6

-tp p6

-tp athlon

-tp athlon

-tpp6

-tpp6

Library

BLAS

ATLAS

BLAS

ATLAS

BLAS

ATLAS

BLAS

ATLAS

BLAS

ATLAS

BLAS

ATLAS

322-1

9.65

9.08

10.05

8.87

7.48

7.67

8.81

9.37

8.51

8.47

6.85

6.79

322-2

166

165

171

172

114

108

145

179

167

167

110

112

322-3

305

308

305

317

208

192

296

335

312

312

194

198

338

58

57

57

55

42

35

50

47

48

50

37

36

339

57

59

57

57

42

37

54

52

54

55

39

38

364 d

44

46

44

45

- b

- b

43

43

42

45

- b

- b

397

848

918

847

862

644

634

748

687

692

775

597

617

415-1

40

36

40

34

34

27

51

38

44

44

33

33

415-2 d

- c

- c

- c

- c

- c

- c

- c

- c

- c

- c

- c

- c

420-1

206

195

198

192

161

139

208

188

184

197

163

150

420-2

199

187

198

191

160

139

208

184

184

200

173

148

424-1

161

92

161

94

130

75

179

104

142

106

140

87

424-2

161

95

161

93

131

75

176

104

142

102

145

85

438

61

55

61

53

53

42

60

50

59

51

56

41

439-1 d

297(4)

190(2)

297(4)

175(2)

116(2)

111(2)

286(4)

166(2)

274(4)

164(2)

128(2)

114(2)

439-2

123

131

123

123

86

83

120

126

115

126

92

89

447-1

339

295

340

289

248

202

299

229

266

248

224

207

447-2

165

153

163

156

125

112

146

119

136

128

116

104

559-4

21

23

21

21

14

16

29

23

23

26

19

20

560-4

25

27

24

24

16

19

33

27

27

30

21

21

561-4

56

58

55

55

46

47

66

61

61

65

52

52

Sum

3001

2863

2992

2797

2261

1990

2877

2562

2665

2690

2218

2045

a The optimization to ifc other than the architectural optimization is "-unroll -O2". 

b Bump out due to unknown address error reported by ifc.

c Error due to failure to write scratch files larger than 2 GB. 

d Excluded from sum.