CPU 信息
i7 Tigerlake
| 界限 | 整数加法 | 整数乘法 | 浮点数加法 | 浮点数乘法 |
|---|---|---|---|---|
| 延迟 | 1 | 3 | 4 | 4 |
| 吞吐量 | 0.5 | 1 | 0.5 | 0.5 |
缓存信息
L1 d-cache：
| 高速缓存大小（） | 相连度（） | 块大小（） | 组数（） |
|---|---|---|---|
| 48KB | 12 | 64B | 64 |
家庭作业
5.18
CPE = 1.07 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85double poly(double a[], double x, long degree) {
double result0 = 0;
double result1 = 0;
double result2 = 0;
double result3 = 0;
double result4 = 0;
double result5 = 0;
double result6 = 0;
double result7 = 0;
double result8 = 0;
double result9 = 0;
double resultA = 0;
double resultB = 0;
double resultC = 0;
double resultD = 0;
double resultE = 0;
degree++;
long remaining = degree % 15, align = degree - remaining;
switch (remaining) {
case 14:
resultD = a[align + 13];
case 13:
resultC = a[align + 12];
case 12:
resultB = a[align + 11];
case 11:
resultA = a[align + 10];
case 10:
result9 = a[align + 9];
case 9:
result8 = a[align + 8];
case 8:
result7 = a[align + 7];
case 7:
result6 = a[align + 6];
case 6:
result5 = a[align + 5];
case 5:
result4 = a[align + 4];
case 4:
result3 = a[align + 3];
case 3:
result2 = a[align + 2];
case 2:
result1 = a[align + 1];
case 1:
result0 = a[align];
}
double X = x * x * x;
X *= X * x;
X *= X * x;
for (long i = align; i > 0; i -= 15) {
result0 = result0 * X + a[i - 15];
result1 = result1 * X + a[i - 14];
result2 = result2 * X + a[i - 13];
result3 = result3 * X + a[i - 12];
result4 = result4 * X + a[i - 11];
result5 = result5 * X + a[i - 10];
result6 = result6 * X + a[i - 9];
result7 = result7 * X + a[i - 8];
result8 = result8 * X + a[i - 7];
result9 = result9 * X + a[i - 6];
resultA = resultA * X + a[i - 5];
resultB = resultB * X + a[i - 4];
resultC = resultC * X + a[i - 3];
resultD = resultD * X + a[i - 2];
resultE = resultE * X + a[i - 1];
}
return (((((((((((((
resultE * x +
resultD) * x +
resultC) * x +
resultB) * x +
resultA) * x +
result9) * x +
result8) * x +
result7) * x +
result6) * x +
result5) * x +
result4) * x +
result3) * x +
result2) * x +
result1) * x +
result0;
}
5.19
CPE = 1.28 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34/* Compute prefix sum of vector a */
void psum(float a[], float p[], long n) {
p[0] = a[0];
float sum = a[0];
long limit = n - 14;
long i;
for (i = 1; i < limit; i += 15) {
float a2 = a[i + 1], a4 = a[i + 3], a8 = a[i + 7];
float p1 = sum + a[i], sum2_3 = a2 + a[i + 2];
float sum4_5 = a4 + a[i + 4], sum6_7 = a[i + 5] + a[i + 6];
float sum8_9 = a8 + a[i + 8], sum10_11 = a[i + 9] + a[i + 10];
float sum12_13 = a[i + 11] + a[i + 12], sum14_15 = a[i + 13] + a[i + 14];
float p3 = p1 + sum2_3;
float sum4_6 = sum4_5 + a[i + 5], sum4_7 = sum4_5 + sum6_7;
float sum8_10 = sum8_9 + a[i + 9], sum8_11 = sum8_9 + sum10_11;
float sum12_14 = sum12_13 + a[i + 13], sum12_15 = sum12_13 + sum14_15;
float p7 = p3 + sum4_7;
float sum8_14 = sum8_11 + sum12_14, sum8_15 = sum8_11 + sum12_15;
float sum8_12 = sum8_11 + a[i + 11], sum8_13 = sum8_11 + sum12_13;
p[i] = p1, p[i + 1] = p1 + a2;
p[i + 2] = p3, p[i + 3] = p3 + a4;
p[i + 4] = p3 + sum4_5, p[i + 5] = p3 + sum4_6;
p[i + 6] = p7, p[i + 7] = p7 + a8;
p[i + 8] = p7 + sum8_9, p[i + 9] = p7 + sum8_10;
p[i + 10] = p7 + sum8_11, p[i + 11] = p7 + sum8_12;
p[i + 12] = p7 + sum8_13, p[i + 13] = p7 + sum8_14;
sum = p[i + 14] = p7 + sum8_15;
}
for (; i < n; i++) {
p[i] = a[i] + p[i - 1];
}
}
SIMD.6
CPE = 0.27 点击展开
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239typedef double vec_t __attribute__ ((vector_size(32)));
double poly(double a[], double x, long degree) {
double *b = a;
while ((size_t) a % 32 != 0 && degree) {
a++;
degree--;
}
vec_t vec0, vec1, vec2, vec3,
vec4, vec5, vec6, vec7,
vec8, vec9, vecA, vecB,
vecC, vecD, vecE;
for (int i = 0; i < 4; i++) {
vec0[i] = 0;
vec1[i] = 0;
vec2[i] = 0;
vec3[i] = 0;
vec4[i] = 0;
vec5[i] = 0;
vec6[i] = 0;
vec7[i] = 0;
}
degree++;
long remaining = degree % 60, align = degree - remaining;
switch (remaining) {
case 59:
vecE[2] = a[align + 58];
case 58:
vecE[1] = a[align + 57];
case 57:
vecE[0] = a[align + 56];
case 56:
vecD[3] = a[align + 55];
case 55:
vecD[2] = a[align + 54];
case 54:
vecD[1] = a[align + 53];
case 53:
vecD[0] = a[align + 52];
case 52:
vecC[3] = a[align + 51];
case 51:
vecC[2] = a[align + 50];
case 50:
vecC[1] = a[align + 49];
case 49:
vecC[0] = a[align + 48];
case 48:
vecB[3] = a[align + 47];
case 47:
vecB[2] = a[align + 46];
case 46:
vecB[1] = a[align + 45];
case 45:
vecB[0] = a[align + 44];
case 44:
vecA[3] = a[align + 43];
case 43:
vecA[2] = a[align + 42];
case 42:
vecA[1] = a[align + 41];
case 41:
vecA[0] = a[align + 40];
case 40:
vec9[3] = a[align + 39];
case 39:
vec9[2] = a[align + 38];
case 38:
vec9[1] = a[align + 37];
case 37:
vec9[0] = a[align + 36];
case 36:
vec8[3] = a[align + 35];
case 35:
vec8[2] = a[align + 34];
case 34:
vec8[1] = a[align + 33];
case 33:
vec8[0] = a[align + 32];
case 32:
vec7[3] = a[align + 31];
case 31:
vec7[2] = a[align + 30];
case 30:
vec7[1] = a[align + 29];
case 29:
vec7[0] = a[align + 28];
case 28:
vec6[3] = a[align + 27];
case 27:
vec6[2] = a[align + 26];
case 26:
vec6[1] = a[align + 25];
case 25:
vec6[0] = a[align + 24];
case 24:
vec5[3] = a[align + 23];
case 23:
vec5[2] = a[align + 22];
case 22:
vec5[1] = a[align + 21];
case 21:
vec5[0] = a[align + 20];
case 20:
vec4[3] = a[align + 19];
case 19:
vec4[2] = a[align + 18];
case 18:
vec4[1] = a[align + 17];
case 17:
vec4[0] = a[align + 16];
case 16:
vec3[3] = a[align + 15];
case 15:
vec3[2] = a[align + 14];
case 14:
vec3[1] = a[align + 13];
case 13:
vec3[0] = a[align + 12];
case 12:
vec2[3] = a[align + 11];
case 11:
vec2[2] = a[align + 10];
case 10:
vec2[1] = a[align + 9];
case 9:
vec2[0] = a[align + 8];
case 8:
vec1[3] = a[align + 7];
case 7:
vec1[2] = a[align + 6];
case 6:
vec1[1] = a[align + 5];
case 5:
vec1[0] = a[align + 4];
case 4:
vec0[3] = a[align + 3];
case 3:
vec0[2] = a[align + 2];
case 2:
vec0[1] = a[align + 1];
case 1:
vec0[0] = a[align + 0];
}
double X = x * x * x;
X *= X * x;
X *= X * x;
X *= X;
X *= X;
vec_t mul = {X, X, X, X};
for (long i = align; i > 0; i -= 60) {
vec0 = vec0 * mul + *((vec_t*) &a[i - 60]);
vec1 = vec1 * mul + *((vec_t*) &a[i - 56]);
vec2 = vec2 * mul + *((vec_t*) &a[i - 52]);
vec3 = vec3 * mul + *((vec_t*) &a[i - 48]);
vec4 = vec4 * mul + *((vec_t*) &a[i - 44]);
vec5 = vec5 * mul + *((vec_t*) &a[i - 40]);
vec6 = vec6 * mul + *((vec_t*) &a[i - 36]);
vec7 = vec7 * mul + *((vec_t*) &a[i - 32]);
vec8 = vec8 * mul + *((vec_t*) &a[i - 28]);
vec9 = vec9 * mul + *((vec_t*) &a[i - 24]);
vecA = vecA * mul + *((vec_t*) &a[i - 20]);
vecB = vecB * mul + *((vec_t*) &a[i - 16]);
vecC = vecC * mul + *((vec_t*) &a[i - 12]);
vecD = vecD * mul + *((vec_t*) &a[i - 8]);
vecE = vecE * mul + *((vec_t*) &a[i - 4]);
}
double result = ((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
vecE[3] * x +
vecE[2]) * x +
vecE[1]) * x +
vecE[0]) * x +
vecD[3]) * x +
vecD[2]) * x +
vecD[1]) * x +
vecD[0]) * x +
vecC[3]) * x +
vecC[2]) * x +
vecC[1]) * x +
vecC[0]) * x +
vecB[3]) * x +
vecB[2]) * x +
vecB[1]) * x +
vecB[0]) * x +
vecA[3]) * x +
vecA[2]) * x +
vecA[1]) * x +
vecA[0]) * x +
vec9[3]) * x +
vec9[2]) * x +
vec9[1]) * x +
vec9[0]) * x +
vec8[3]) * x +
vec8[2]) * x +
vec8[1]) * x +
vec8[0]) * x +
vec7[3]) * x +
vec7[2]) * x +
vec7[1]) * x +
vec7[0]) * x +
vec6[3]) * x +
vec6[2]) * x +
vec6[1]) * x +
vec6[0]) * x +
vec5[3]) * x +
vec5[2]) * x +
vec5[1]) * x +
vec5[0]) * x +
vec4[3]) * x +
vec4[2]) * x +
vec4[1]) * x +
vec4[0]) * x +
vec3[3]) * x +
vec3[2]) * x +
vec3[1]) * x +
vec3[0]) * x +
vec2[3]) * x +
vec2[2]) * x +
vec2[1]) * x +
vec2[0]) * x +
vec1[3]) * x +
vec1[2]) * x +
vec1[1]) * x +
vec1[0]) * x +
vec0[3]) * x +
vec0[2]) * x +
vec0[1]) * x +
vec0[0];
while (a > b) {
a--;
result = result * x + *a;
}
return result;
}
perflab
https://git.gzezfisher.top/FISHER_/CSAPP-sol/src/branch/main/perf/kernels.c
cachelab
https://git.gzezfisher.top/FISHER_/CSAPP-sol/src/branch/main/cache