Hi All
I am trying to analyze the difference between binaries produced using clang
-Oz and Opt -Oz. All the binaries produced are using the -c flag. Right off
the bat, I can see the difference between the IR file produced by clang -Oz
which adds "minsize","optsize" and removes the
"no-inline" function
attribute. To keep things similar the same is done to the default IR
produced for the following code with the optnone attribute removed using
the -disable-O0-optnone flag.
#include <stdio.h>
int num[326] = {
1,
2,
3,
4,
6,
8,
9,
12,
16,
18,
24,
27,
32,
36,
48,
54,
64,
72,
81,
96,
108,
128,
144,
162,
192,
216,
243,
256,
288,
324,
384,
432,
486,
512,
576,
648,
729,
768,
864,
972,
1024,
1152,
1296,
1458,
1536,
1728,
1944,
2048,
2187,
2304,
2592,
2916,
3072,
3456,
3888,
4096,
4374,
4608,
5184,
5832,
6144,
6561,
6912,
7776,
8192,
8748,
9216,
10368,
11664,
12288,
13122,
13824,
15552,
16384,
17496,
18432,
19683,
20736,
23328,
24576,
26244,
27648,
31104,
32768,
34992,
36864,
39366,
41472,
46656,
49152,
52488,
55296,
59049,
62208,
65536,
69984,
73728,
78732,
82944,
93312,
98304,
104976,
110592,
118098,
124416,
131072,
139968,
147456,
157464,
165888,
177147,
186624,
196608,
209952,
221184,
236196,
248832,
262144,
279936,
294912,
314928,
331776,
354294,
373248,
393216,
419904,
442368,
472392,
497664,
524288,
531441,
559872,
589824,
629856,
663552,
708588,
746496,
786432,
839808,
884736,
944784,
995328,
1048576,
1062882,
1119744,
1179648,
1259712,
1327104,
1417176,
1492992,
1572864,
1594323,
1679616,
1769472,
1889568,
1990656,
2097152,
2125764,
2239488,
2359296,
2519424,
2654208,
2834352,
2985984,
3145728,
3188646,
3359232,
3538944,
3779136,
3981312,
4194304,
4251528,
4478976,
4718592,
4782969,
5038848,
5308416,
5668704,
5971968,
6291456,
6377292,
6718464,
7077888,
7558272,
7962624,
8388608,
8503056,
8957952,
9437184,
9565938,
10077696,
10616832,
11337408,
11943936,
12582912,
12754584,
13436928,
14155776,
14348907,
15116544,
15925248,
16777216,
17006112,
17915904,
18874368,
19131876,
20155392,
21233664,
22674816,
23887872,
25165824,
25509168,
26873856,
28311552,
28697814,
30233088,
31850496,
33554432,
34012224,
35831808,
37748736,
38263752,
40310784,
42467328,
43046721,
45349632,
47775744,
50331648,
51018336,
53747712,
56623104,
57395628,
60466176,
63700992,
67108864,
68024448,
71663616,
75497472,
76527504,
80621568,
84934656,
86093442,
90699264,
95551488,
100663296,
102036672,
107495424,
113246208,
114791256,
120932352,
127401984,
129140163,
134217728,
136048896,
143327232,
150994944,
153055008,
161243136,
169869312,
172186884,
181398528,
191102976,
201326592,
204073344,
214990848,
226492416,
229582512,
241864704,
254803968,
258280326,
268435456,
272097792,
286654464,
301989888,
306110016,
322486272,
339738624,
344373768,
362797056,
382205952,
387420489,
402653184,
408146688,
429981696,
452984832,
459165024,
483729408,
509607936,
516560652,
536870912,
544195584,
573308928,
603979776,
612220032,
644972544,
679477248,
688747536,
725594112,
764411904,
774840978,
805306368,
816293376,
859963392,
905969664,
918330048,
967458816,
1019215872,
1033121304,
1073741824,
1088391168,
1146617856,
1162261467,
1207959552,
1224440064,
1289945088,
1358954496,
1377495072,
1451188224,
1528823808,
1549681956,
1610612736,
1632586752,
1719926784,
1811939328,
1836660096,
1934917632
};
int main() {
int l, r;
scanf("%d %d", &l, &r);
int L = -1, R = 326;
for (int i = 0; i < 326; ++i) {
if (num[i] < l) {
L = i;
}
}
for (int i = 325; i >= 0; --i) {
if (num[i] > r) {
R = i;
}
}
printf("%d", (R - L - (long long) 1));
return 0;
}
What follows next is applying opt -S -Oz on the same IR followed by clang
-c <Oz_IR_File> -o binname.out to create the binary
The binary thus formed has a size of 3072 bytes whereas the binary formed
with Clang -Oz on the c file viz clang -c -Oz filename.c has a size of
2944 bytes and the most interesting part is that the out file produced
without any optimization viz clang -c filename.c is 3040 bytes
Are there some other attributes that need to be added to get Opt -Oz
similar to clang -Oz or am I missing something?
P.S: Apologies for the huge chunk of code in the middle of the mail. I was
not sure if attaching it would be viewable or not.
P.S.S: This is one file out of a few more files that result in the
discrepancy.
Regards
Anilava
--
Disclaimer:- This footer text is to convey that this email is sent by one
of the users of IITH. So, do not mark it as SPAM.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20210120/28b64863/attachment-0001.html>