From: Rick Hohensee <humbubba@smarty.smart.net>
To: linux-kernel@vger.kernel.org
Subject: What's a return stack worth?
Date: Sun, 18 Mar 2001 00:38:38 -0500 (EST) [thread overview]
Message-ID: <200103180538.AAA14598@smarty.smart.net> (raw)
These are the occurance counts of the 256 byte values in a big monolithic
stripped 2.4.0 UP 386 vmlinux. 1/256 is .39% .
count dec hex octal/ascii % possible meaning
334170 0 0 [ 0] 17.05% ADD
63889 255 ff [377] 3.26% 2-byte escape
63536 36 24 $ 3.24% AND with AL
52652 139 8b [213] 2.68% MOV
50862 192 c0 [300] 2.59% 2 byte escape
44628 131 83 [203] 2.27% 2-byte escape
42419 137 89 [211] 2.16% MOV
39630 116 74 t 2.02% JZ
37832 32 20 1.93% AND
27742 1 1 [ 1] 1.41% ADD
22879 4 4 [ 4] 1.16% ADD with AL
22307 232 e8 [350] 1.13% CALL
21945 196 c4 [304] 1.11% LES
21769 68 44 D 1.11% INC eSP
19960 102 66 f 1.01% other operand size prefix
19793 8 8 [ 10] 1.01% OR
19212 15 f [ 17] .98% 2-byte escape
18234 117 75 u .93% JNZ
17335 101 65 e .88% GS prefix
17204 141 8d [215] .87% LEA
17066 16 10 [ 20] .87% ADC
16673 133 85 [205] .85% TEST
14666 12 c [ 14] .74% OR with AL
13710 115 73 s .69% JNB
13642 195 c3 [303] .69% RET near
13293 246 f6 [366] .67% 2-byte escape
13103 83 53 S .66% PUSH eBX
12969 100 64 d .66% FS prefix
12895 104 68 h .65% PUSH immediate
12820 2 2 [ 2] .65% ADD
12525 37 25 % .63% AND with eAX
12412 20 14 [ 24] .63% ADC with AL
11661 114 72 r .59% JB
11619 80 50 P .59% PUSH eAX
10791 128 80 [200] .55% 2-byte escape
10776 67 43 C .54% INC eBX
10702 199 c7 [307] .54% MOV
10640 111 6f o .54% OUTSW/D
10577 24 18 [ 30] .53% SBB
10546 124 7c | .53% JL
10517 105 69 i .53% IMUL
10256 95 5f _ .52% POP eDI
10073 84 54 T .51% PUSH eSP
10055 110 6e n .51% OUTSB
9965 118 76 v .50% JBE
9812 97 61 a .50% POPA
9667 40 28 ( .49% SUB
9435 49 31 1 .48% XOR
9410 76 4c L .48% DEC eSP
9304 108 6c l .47% INSB
9277 3 3 [ 3] .47% ADD
8988 64 40 @ .45% INC eAX
8922 91 5b [ .45% POP eBX
8903 193 c1 [301] .45% 2 byte escape
8638 132 84 [204] .44% TEST
8580 144 90 [220] .43% NOP
8423 86 56 V .42% PUSH eSI
8422 48 30 0 .42% XOR
8417 5 5 [ 5] .42% ADD with eAX
8267 28 1c [ 34] .42% SBB with AL
8116 235 eb [353] .41% JNP
7963 46 2e . .40% CS prefix
7882 94 5e ^ .40% POP eSI
7856 99 63 c .40% ARPL
7723 136 88 [210] .39% MOV
7667 41 29 ) .39% SUB
7404 10 a [ 12] .37% OR
7226 112 70 p .36% JO
7220 129 81 [201] .36% 2-byte escape
6818 57 39 9 .34% CMP
6732 138 8a [212] .34% MOV
6680 194 c2 [302] .34% RET near with stacked data
6518 44 2c , .33% SUB with AL
6509 92 5c \ .33% POP eSP
6363 184 b8 [270] .32% MOV immediate
6246 224 e0 [340] .31% LOOPNE
6139 85 55 U .31% PUSH eBP
6117 42 2a * .31% SUB
6001 87 57 W .30% PUSH eDI
5937 254 fe [376] .30% INC/DEC
5839 106 6a j .29% PUSH immediate
5829 7 7 [ 7] .29% POP ES
5703 182 b6 [266] .29% MOV immediate
5654 198 c6 [306] .28% MOV
5553 70 46 F .28% INC eSI
5327 93 5d ] .27% POP eBP
5318 60 3c < .27% CMP with AL
5301 58 3a : .27% CMP
5281 236 ec [354] .26% IN
5174 52 34 4 .26% XOR with AL
5165 120 78 x .26% JS
5156 9 9 [ 11] .26% OR
5121 233 e9 [351] .26% JNP
5014 6 6 [ 6] .25% PUSH ES
4921 56 38 8 .25% CMP
4915 66 42 B .25% INC eDX
4913 65 41 A .25% INC eCX
4890 69 45 E .24% INC eBP
4846 13 d [ 15] .24% OR with eAX
4825 240 f0 [360] .24% LOCK
4768 248 f8 [370] .24% CLC
4623 208 d0 [320] .23% 2-byte escape
4513 109 6d m .23% INSW/D
4488 82 52 R .22% PUSH eDX
4459 61 3d = .22% CMP with eAX
4444 252 fc [374] .22% CLD
4411 43 2b + .22% SUB
4335 71 47 G .22% INC eDI
4235 210 d2 [322] .21% 2-byte escape
3918 238 ee [356] .19% OUT
3841 39 27 ' .19% DAA
3797 21 15 [ 25] .19% ADC with eAX
3778 250 fa [372] .19% CLI
3698 103 67 g .18% other address size prefix
3629 72 48 H .18% DEC eAX
3612 219 db [333] .18% coprocessor
3525 38 26 & .17% ES prefix
3431 50 32 2 .17% XOR
3349 73 49 I .17% DEC eCX
3340 33 21 ! .17% AND
3330 134 86 [206] .16% XCNG
3257 119 77 w .16% JNBE
3185 216 d8 [330] .16% coprocessor
3142 200 c8 [310] .16% ENTER
3126 247 f7 [367] .15% 2-byte escape
2980 17 11 [ 21] .15% ADC
2977 121 79 y .15% JNS
2963 226 e2 [342] .15% LOOP
2953 176 b0 [260] .15% MOV immediate
2930 251 fb [373] .14% STI
2891 243 f3 [363] .14% REP
2885 11 b [ 13] .14% OR
2875 126 7e ~ .14% JLE
2857 183 b7 [267] .14% MOV immediate
2856 78 4e N .14% DEC eSI
2834 98 62 b .14% BOUND
2812 88 58 X .14% POP eAX
2740 14 e [ 16] .13% PUSH CS
2734 18 12 [ 22] .13% ADC
2725 242 f2 [362] .13% REPNE
2693 127 7f [177] .13% JNLE
2692 156 9c [234] .13% PUSHF
2650 81 51 Q .13% PUSH eCX
2636 161 a1 [241] .13% MOV
2618 47 2f / .13% DAS
2585 25 19 [ 31] .13% SBB
2563 201 c9 [311] .13% LEAVE
2537 237 ed [355] .12% IN
2528 22 16 [ 26] .12% PUSH SS
2510 249 f9 [371] .12% STC
2505 107 6b k .12% IMUL
2485 19 13 [ 23] .12% ADC
2484 135 87 [207] .12% XCNG
2467 185 b9 [271] .12% MOV immediate
2419 140 8c [214] .12% MOV
2384 45 2d - .12% SUB with eAX
2372 77 4d M .12% DEC eBP
2372 51 33 3 .12% XOR
2341 168 a8 [250] .11% TEST
2325 62 3e > .11% CS prefix
2305 253 fd [375] .11% STD
2283 218 da [332] .11% coprocessor
2282 202 ca [312] .11% RET far
2262 148 94 [224] .11% XCHG eSP
2241 53 35 5 .11% XOR with eAX
2224 125 7d } .11% JNL
2197 180 b4 [264] .11% MOV immediate
2161 130 82 [202] .11%
2128 26 1a [ 32] .10% SBB
2122 172 ac [254] .10% LODSB
2118 197 c5 [305] .10% LDS
2101 29 1d [ 35] .10% SBB with eAX
2078 23 17 [ 27] .10% POP SS
2068 186 ba [272] .10% MOV immediate
2035 34 22 " .10% AND
2024 31 1f [ 37] .10% POP DS
2016 79 4f O .10% DEC eDI
1994 75 4b K .10% DEC eBX
1981 188 bc [274] .10% MOV immediate
1976 239 ef [357] .10% OUT
1951 244 f4 [364] .09% HLT
1933 96 60 ` .09% PUSHA
1918 160 a0 [240] .09% MOV
1886 35 23 # .09% AND
1884 187 bb [273] .09% MOV immediate
1870 54 36 6 .09% SS prefix
1855 55 37 7 .09% AAA
1849 234 ea [352] .09% JNP
1811 190 be [276] .09% MOV immediate
1797 225 e1 [341] .09% LOOPE
1788 30 1e [ 36] .09% PUSH DS
1719 211 d3 [323] .08% 2-byte escape
1678 228 e4 [344] .08% IN AL
1636 191 bf [277] .08% MOV immediate
1621 123 7b { .08% JNP
1619 230 e6 [346] .08% OUT AL
1596 27 1b [ 33] .08% SBB
1590 157 9d [235] .08% POPF
1582 165 a5 [245] .08% MOVSW/D
1545 152 98 [230] .07% CBW
1506 164 a4 [244] .07% MOVSB
1427 245 f5 [365] .07% CMC
1406 227 e3 [343] .07% JCXZ
1333 222 de [336] .06% coprocessor
1331 209 d1 [321] .06% 2-byte escape
1311 63 3f ? .06% AAS
1305 171 ab [253] .06% STOSW/D
1291 113 71 q .06% JNO
1271 241 f1 [361] .06%
1270 74 4a J .06% DEC eDX
1214 220 dc [334] .06% coprocessor
1184 122 7a z .06% JP
1123 189 bd [275] .05% MOV immediate
1120 142 8e [216] .05% MOV
1104 212 d4 [324] .05% AAM
1100 89 59 Y .05% POP eCX
1098 223 df [337] .05% coprocessor
1094 163 a3 [243] .05% MOV
1037 149 95 [225] .05% XCHG eBP
1006 147 93 [223] .05% XCHG eBX
994 179 b3 [263] .05% MOV immediate
972 204 cc [314] .04% INT 3
966 217 d9 [331] .04% coprocessor
955 90 5a Z .04% POP eDX
930 231 e7 [347] .04% OUT eAX
905 59 3b ; .04% CMP
875 174 ae [256] .04% SCASB
874 145 91 [221] .04% XCHG eCX
857 229 e5 [345] .04% IN eAX
854 203 cb [313] .04% RET far
836 214 d6 [326] .04%
833 181 b5 [265] .04% MOV immediate
833 143 8f [217] .04% POP
825 207 cf [317] .04% IRET
824 146 92 [222] .04% XCHG eDX
809 206 ce [316] .04% INTO
806 213 d5 [325] .04% AAD
777 153 99 [231] .03% CWD
758 158 9e [236] .03% SAHF
758 150 96 [226] .03% XCHG eSI
755 159 9f [237] .03% LAHF
751 215 d7 [327] .03% XLAT
748 177 b1 [261] .03% MOV immediate
735 178 b2 [262] .03% MOV immediate
735 173 ad [255] .03% LODSW/D
731 170 aa [252] .03% STOSB
722 205 cd [315] .03% INT
684 221 dd [335] .03% coprocessor
669 155 9b [233] .03% WAIT
638 151 97 [227] .03% XCHG eDI
577 166 a6 [246] .02% CMPSB
567 175 af [257] .02% SCASW/D
529 154 9a [232] .02% CALL
515 169 a9 [251] .02% TEST
513 162 a2 [242] .02% MOV
386 167 a7 [247] .01% CMPSW/D
Considering 386 opcodes as "signal" (and throwing out the zeros), decimal
36 looks like a spike in the noise due to a modR/M byte value. 36 decimal
is the pmode modR/M byte saying "this instruction uses a memory reference
with a one-byte literal displacement value." gcc does a lot of that, for
locals I believe. LES, octal 304, is probably also actually a prevalent
modR/M value. Otherwise things look mostly as one might expect. ff, 1, 2,
4 and 8 are probably frequent as literals. The fact that there are more
4's than 2's meshes with the fact that it's a 4 byte machine as far as
vmlinux is concerned. There are a lot of 4's affiliated with that locals
addressing mode. ASCII text in the kernel seems to be a relatively light
factor. The apparent ratio between CALLs and RETs is about 1.6:1, which is
the low end of what I expected. I suspect 1.6:1 is representative though.
That interests me quite a lot. Is the exact ratio known? Has anyone ever
built a 100% inlined kernel?
Rick Hohensee
www.clienux.com
reply other threads:[~2001-03-18 5:36 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=200103180538.AAA14598@smarty.smart.net \
--to=humbubba@smarty.smart.net \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox