News:

Masm32 SDK description, downloads and other helpful links
Message to All Guests
NB: Posting URL's See here: Posted URL Change

Main Menu

Sorting strings

Started by RuiLoureiro, May 29, 2014, 06:15:48 AM

Previous topic - Next topic
Print
Go Down Pages 1 ... 7 8 9 10 11
User actions

RuiLoureiro

#120
July 09, 2014, 08:17:51 PM Last Edit: July 09, 2014, 09:51:33 PM by RuiLoureiro
Hi,
        Now, i used crt_memcpy, memcpy_1, memcpy_2, memcpy_3
        procedures.
        Copying 16 bytes at a time we don't get the best results in my P4.

        The versions YZZE, YZZG and YZZH copy 16 bytes at a time using
        the 32 bit registers.
        The strings are not aligned.
       
        Gunther, could you post your results, if you dont mind ?
        Thank you.
       
        These are my results.
Quote
NOT ALIGNED
-----------------------------------------------------
Intel(R) Pentium(R) 4 CPU 3.40GHz (SSE3)
-----------------------------------------------------
***** Time table *****
  32 cycles, COPYAtoB_SSEDS-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
  32 cycles, COPYAtoB_SSEES-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
  32 cycles, COPYAtoB_XZZFS-  15 bytes- copy lenght DWORDS+MOVZX- uses ESP
  33 cycles, COPYAtoB_XZES-   15 bytes- copy lenght DWORDS+MOVZX- uses ESP
  34 cycles, COPYAtoB_XZZES-  15 bytes- copy lenght DWORDS+MOVZX- uses ESP
  35 cycles, COPYAtoB_XZZCS-  15 bytes- copy lenght DWORDS+MOVZX- uses ESP
  35 cycles, COPYAtoB_SSEFS-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
  35 cycles, crt_memcpy-      15 bytes- copy crt_memcpy
  35 cycles, COPYAtoB_SSED-   15 bytes- copy 16 BYTES+MOVZX
  35 cycles, COPYAtoB_XZZF-   15 bytes- copy lenght DWORDS+MOVZX
  36 cycles, COPYAtoB_XZZC-   15 bytes- copy lenght DWORDS+MOVZX

  37 cycles, COPYAtoB_YZZH-   15 bytes- copy length 16 BYTES+MOVZX  <<<<---
 
  37 cycles, COPYAtoB_SSEF-   15 bytes- copy 16 BYTES+MOVZX
  38 cycles, COPYAtoB_SSEE-   15 bytes- copy 16 BYTES+MOVZX
  39 cycles, COPYAtoB_XZZE-   15 bytes- copy lenght DWORDS+MOVZX
  41 cycles, COPYAtoB_YZZG-   15 bytes- copy length 16 BYTES+MOVZX  <<<<---
 
  41 cycles, COPYAtoB_YZZE-   15 bytes- copy length 16 BYTES+MOVZX
  45 cycles, COPYAtoB_XZE-    15 bytes- copy lenght DWORDS+MOVZX
  50 cycles, COPYAtoB_WZZE-   15 bytes- copy 16 BYTES+MOVZX
  63 cycles, COPYAtoB_XZES-   53 bytes- copy lenght DWORDS+MOVZX- uses ESP
  65 cycles, COPYAtoB_XZZFS-  53 bytes- copy lenght DWORDS+MOVZX- uses ESP

  69 cycles, memcpy_1-        53 bytes- copy regcopy                >>>>---
  72 cycles, COPYAtoB_XZZES-  53 bytes- copy lenght DWORDS+MOVZX- uses ESP
  74 cycles, COPYAtoB_XZZCS-  53 bytes- copy lenght DWORDS+MOVZX- uses ESP

  74 cycles, COPYAtoB_YZZG-   53 bytes- copy length 16 BYTES+MOVZX  <<<<---

  75 cycles, COPYAtoB_XZZE-   53 bytes- copy lenght DWORDS+MOVZX
  75 cycles, COPYAtoB_XZE-    53 bytes- copy lenght DWORDS+MOVZX
  75 cycles, COPYAtoB_XZZF-   53 bytes- copy lenght DWORDS+MOVZX
  76 cycles, COPYAtoB_YZZE-   53 bytes- copy length 16 BYTES+MOVZX
  77 cycles, COPYAtoB_XZZC-   53 bytes- copy lenght DWORDS+MOVZX

  78 cycles, memcpy_1-        15 bytes- copy regcopy                >>>>---
 
  78 cycles, COPYAtoB_WZZE-   53 bytes- copy 16 BYTES+MOVZX       

  82 cycles, COPYAtoB_YZZH-   53 bytes- copy length 16 BYTES+MOVZX  <<<<---
 
  85 cycles, COPYAtoB_SSEDS-  53 bytes- copy 16 BYTES+MOVZX- uses ESP
  90 cycles, COPYAtoB_SSEFS-  53 bytes- copy 16 BYTES+MOVZX- uses ESP
  91 cycles, COPYAtoB_SSEES-  53 bytes- copy 16 BYTES+MOVZX- uses ESP
  93 cycles, COPYAtoB_SSEF-   53 bytes- copy 16 BYTES+MOVZX
  93 cycles, COPYAtoB_SSED-   53 bytes- copy 16 BYTES+MOVZX
  93 cycles, COPYAtoB_SSEE-   53 bytes- copy 16 BYTES+MOVZX
106 cycles, COPYAtoB_SSEHS-  53 bytes- copy 16 BYTES+MOVZX- uses ESP

106 cycles, memcpy_1-       103 bytes- copy regcopy                >>>>---
107 cycles, crt_memcpy-      53 bytes- copy crt_memcpy
109 cycles, COPYAtoB_YZZH-  103 bytes- copy length 16 BYTES+MOVZX  <<<<---
109 cycles, COPYAtoB_YZZG-  103 bytes- copy length 16 BYTES+MOVZX  <<<<---

110 cycles, COPYAtoB_SSEH-   53 bytes- copy 16 BYTES+MOVZX
110 cycles, COPYAtoB_SSEG-   53 bytes- copy 16 BYTES+MOVZX
110 cycles, COPYAtoB_SSEGS-  53 bytes- copy 16 BYTES+MOVZX- uses ESP
111 cycles, COPYAtoB_YZZE-  103 bytes- copy length 16 BYTES+MOVZX
113 cycles, COPYAtoB_WZZE-  103 bytes- copy 16 BYTES+MOVZX
117 cycles, COPYAtoB_XZZCS- 103 bytes- copy lenght DWORDS+MOVZX- uses ESP
117 cycles, COPYAtoB_XZZFS- 103 bytes- copy lenght DWORDS+MOVZX- uses ESP
117 cycles, COPYAtoB_XZZES- 103 bytes- copy lenght DWORDS+MOVZX- uses ESP
117 cycles, COPYAtoB_XZES-  103 bytes- copy lenght DWORDS+MOVZX- uses ESP
119 cycles, COPYAtoB_XZZE-  103 bytes- copy lenght DWORDS+MOVZX
120 cycles, COPYAtoB_XZZC-  103 bytes- copy lenght DWORDS+MOVZX
120 cycles, COPYAtoB_XZE-   103 bytes- copy lenght DWORDS+MOVZX
122 cycles, COPYAtoB_XZZF-  103 bytes- copy lenght DWORDS+MOVZX
141 cycles, crt_memcpy-     103 bytes- copy crt_memcpy

173 cycles, COPYAtoB_YZZH-  203 bytes- copy length 16 BYTES+MOVZX  <<<<---
173 cycles, COPYAtoB_YZZG-  203 bytes- copy length 16 BYTES+MOVZX  <<<<---

174 cycles, COPYAtoB_YZZE-  203 bytes- copy length 16 BYTES+MOVZX
175 cycles, COPYAtoB_SSEH-   15 bytes- copy 16 BYTES+MOVZX
176 cycles, COPYAtoB_SSEG-   15 bytes- copy 16 BYTES+MOVZX
177 cycles, COPYAtoB_SSEHS-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
177 cycles, memcpy_1-       203 bytes- copy regcopy                >>>>---

177 cycles, COPYAtoB_SSEGS-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
188 cycles, COPYAtoB_SSEDS- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
205 cycles, COPYAtoB_XZES-  203 bytes- copy lenght DWORDS+MOVZX- uses ESP
206 cycles, COPYAtoB_XZZFS- 203 bytes- copy lenght DWORDS+MOVZX- uses ESP
207 cycles, COPYAtoB_XZZCS- 203 bytes- copy lenght DWORDS+MOVZX- uses ESP
208 cycles, COPYAtoB_SSEFS- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
209 cycles, COPYAtoB_SSEHS- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
210 cycles, COPYAtoB_SSED-  103 bytes- copy 16 BYTES+MOVZX
210 cycles, COPYAtoB_XZZC-  203 bytes- copy lenght DWORDS+MOVZX
211 cycles, COPYAtoB_SSEES- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
211 cycles, COPYAtoB_XZZE-  203 bytes- copy lenght DWORDS+MOVZX
211 cycles, COPYAtoB_XZZF-  203 bytes- copy lenght DWORDS+MOVZX
212 cycles, COPYAtoB_XZE-   203 bytes- copy lenght DWORDS+MOVZX
213 cycles, COPYAtoB_SSEF-  103 bytes- copy 16 BYTES+MOVZX
213 cycles, COPYAtoB_SSEE-  103 bytes- copy 16 BYTES+MOVZX
217 cycles, COPYAtoB_SSEH-  103 bytes- copy 16 BYTES+MOVZX
217 cycles, COPYAtoB_WZZE-  203 bytes- copy 16 BYTES+MOVZX
223 cycles, crt_memcpy-     203 bytes- copy crt_memcpy
224 cycles, COPYAtoB_SSEGS- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
229 cycles, COPYAtoB_SSEG-  103 bytes- copy 16 BYTES+MOVZX
248 cycles, memcpy_2-        15 bytes- copy memcpy SSE
252 cycles, memcpy_3-        15 bytes- copy memcpyxmmU SSE
259 cycles, COPYAtoB_XZZES- 203 bytes- copy lenght DWORDS+MOVZX- uses ESP
276 cycles, memcpy_3-        53 bytes- copy memcpyxmmU SSE
300 cycles, memcpy_2-        53 bytes- copy memcpy SSE
379 cycles, memcpy_3-       103 bytes- copy memcpyxmmU SSE

395 cycles, COPYAtoB_YZZH-  503 bytes- copy length 16 BYTES+MOVZX  <<<<---
397 cycles, COPYAtoB_YZZG-  503 bytes- copy length 16 BYTES+MOVZX  <<<<---

403 cycles, COPYAtoB_YZZE-  503 bytes- copy length 16 BYTES+MOVZX
406 cycles, COPYAtoB_SSEE-  203 bytes- copy 16 BYTES+MOVZX
406 cycles, COPYAtoB_SSEFS- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
407 cycles, COPYAtoB_SSEF-  203 bytes- copy 16 BYTES+MOVZX
412 cycles, memcpy_2-       103 bytes- copy memcpy SSE
417 cycles, COPYAtoB_SSEES- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
419 cycles, COPYAtoB_XZZES- 503 bytes- copy lenght DWORDS+MOVZX- uses ESP
419 cycles, COPYAtoB_XZZF-  503 bytes- copy lenght DWORDS+MOVZX
420 cycles, COPYAtoB_XZZCS- 503 bytes- copy lenght DWORDS+MOVZX- uses ESP
421 cycles, COPYAtoB_XZES-  503 bytes- copy lenght DWORDS+MOVZX- uses ESP
421 cycles, COPYAtoB_XZZC-  503 bytes- copy lenght DWORDS+MOVZX
422 cycles, COPYAtoB_XZE-   503 bytes- copy lenght DWORDS+MOVZX
424 cycles, COPYAtoB_XZZFS- 503 bytes- copy lenght DWORDS+MOVZX- uses ESP
425 cycles, COPYAtoB_WZZE-  503 bytes- copy 16 BYTES+MOVZX
427 cycles, COPYAtoB_XZZE-  503 bytes- copy lenght DWORDS+MOVZX
428 cycles, COPYAtoB_SSEDS- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
435 cycles, COPYAtoB_SSED-  203 bytes- copy 16 BYTES+MOVZX
443 cycles, crt_memcpy-     503 bytes- copy crt_memcpy

447 cycles, memcpy_1-       503 bytes- copy regcopy                >>>>---

466 cycles, COPYAtoB_SSEGS- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
466 cycles, COPYAtoB_SSEHS- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
466 cycles, COPYAtoB_SSEG-  203 bytes- copy 16 BYTES+MOVZX
472 cycles, COPYAtoB_SSEH-  203 bytes- copy 16 BYTES+MOVZX
624 cycles, memcpy_2-       203 bytes- copy memcpy SSE
634 cycles, memcpy_3-       203 bytes- copy memcpyxmmU SSE

728 cycles, memcpy_1-      1027 bytes- copy regcopy                >>>>---
730 cycles, COPYAtoB_YZZH- 1027 bytes- copy length 16 BYTES+MOVZX  <<<<---
760 cycles, COPYAtoB_YZZG- 1027 bytes- copy length 16 BYTES+MOVZX  <<<<---

784 cycles, COPYAtoB_YZZE- 1027 bytes- copy length 16 BYTES+MOVZX
792 cycles, COPYAtoB_XZZES-1027 bytes- copy lenght DWORDS+MOVZX- uses ESP
793 cycles, COPYAtoB_XZE-  1027 bytes- copy lenght DWORDS+MOVZX
793 cycles, COPYAtoB_XZZCS-1027 bytes- copy lenght DWORDS+MOVZX- uses ESP
796 cycles, COPYAtoB_XZZC- 1027 bytes- copy lenght DWORDS+MOVZX
799 cycles, COPYAtoB_XZZE- 1027 bytes- copy lenght DWORDS+MOVZX
801 cycles, COPYAtoB_XZZF- 1027 bytes- copy lenght DWORDS+MOVZX
808 cycles, COPYAtoB_XZES- 1027 bytes- copy lenght DWORDS+MOVZX- uses ESP
812 cycles, COPYAtoB_WZZE- 1027 bytes- copy 16 BYTES+MOVZX
817 cycles, COPYAtoB_XZZFS-1027 bytes- copy lenght DWORDS+MOVZX- uses ESP
831 cycles, crt_memcpy-    1027 bytes- copy crt_memcpy
962 cycles, COPYAtoB_SSEDS- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
968 cycles, COPYAtoB_SSEGS- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
969 cycles, COPYAtoB_SSED-  503 bytes- copy 16 BYTES+MOVZX
980 cycles, COPYAtoB_SSEFS- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
981 cycles, COPYAtoB_SSEE-  503 bytes- copy 16 BYTES+MOVZX
982 cycles, COPYAtoB_SSEG-  503 bytes- copy 16 BYTES+MOVZX
984 cycles, COPYAtoB_SSEES- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
987 cycles, COPYAtoB_SSEF-  503 bytes- copy 16 BYTES+MOVZX
991 cycles, COPYAtoB_SSEH-  503 bytes- copy 16 BYTES+MOVZX
995 cycles, COPYAtoB_SSEHS- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
1140 cycles, memcpy_2-       503 bytes- copy memcpy SSE
1162 cycles, memcpy_3-       503 bytes- copy memcpyxmmU SSE

1527 cycles, COPYAtoB_YZZH- 2062 bytes- copy length 16 BYTES+MOVZX  <<<<---
1546 cycles, COPYAtoB_YZZG- 2062 bytes- copy length 16 BYTES+MOVZX  <<<<---

1557 cycles, COPYAtoB_XZZE- 2062 bytes- copy lenght DWORDS+MOVZX
1558 cycles, memcpy_1-      2062 bytes- copy regcopy                >>>>---

1568 cycles, COPYAtoB_XZZC- 2062 bytes- copy lenght DWORDS+MOVZX
1569 cycles, COPYAtoB_XZZF- 2062 bytes- copy lenght DWORDS+MOVZX
1570 cycles, COPYAtoB_XZZFS-2062 bytes- copy lenght DWORDS+MOVZX- uses ESP
1571 cycles, COPYAtoB_WZZE- 2062 bytes- copy 16 BYTES+MOVZX
1573 cycles, COPYAtoB_XZZCS-2062 bytes- copy lenght DWORDS+MOVZX- uses ESP
1576 cycles, COPYAtoB_XZE-  2062 bytes- copy lenght DWORDS+MOVZX
1577 cycles, COPYAtoB_XZZES-2062 bytes- copy lenght DWORDS+MOVZX- uses ESP
1580 cycles, COPYAtoB_XZES- 2062 bytes- copy lenght DWORDS+MOVZX- uses ESP
1589 cycles, COPYAtoB_YZZE- 2062 bytes- copy length 16 BYTES+MOVZX
1635 cycles, crt_memcpy-    2062 bytes- copy crt_memcpy
1908 cycles, COPYAtoB_SSEES-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
2026 cycles, COPYAtoB_SSEFS-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
2030 cycles, COPYAtoB_SSEF- 1027 bytes- copy 16 BYTES+MOVZX
2043 cycles, COPYAtoB_SSEHS-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
2045 cycles, COPYAtoB_SSEE- 1027 bytes- copy 16 BYTES+MOVZX
2046 cycles, COPYAtoB_SSED- 1027 bytes- copy 16 BYTES+MOVZX
2078 cycles, COPYAtoB_SSEH- 1027 bytes- copy 16 BYTES+MOVZX
2087 cycles, COPYAtoB_SSEDS-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
2109 cycles, memcpy_3-      1027 bytes- copy memcpyxmmU SSE
2118 cycles, memcpy_2-      1027 bytes- copy memcpy SSE
2123 cycles, COPYAtoB_SSEGS-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
2150 cycles, COPYAtoB_SSEG- 1027 bytes- copy 16 BYTES+MOVZX
3949 cycles, COPYAtoB_SSEGS-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
4010 cycles, COPYAtoB_SSEG- 2062 bytes- copy 16 BYTES+MOVZX
4020 cycles, COPYAtoB_SSEES-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
4046 cycles, COPYAtoB_SSEE- 2062 bytes- copy 16 BYTES+MOVZX
4052 cycles, COPYAtoB_SSEDS-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
4054 cycles, COPYAtoB_SSEFS-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
4060 cycles, COPYAtoB_SSED- 2062 bytes- copy 16 BYTES+MOVZX
4081 cycles, COPYAtoB_SSEHS-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
4086 cycles, COPYAtoB_SSEF- 2062 bytes- copy 16 BYTES+MOVZX
4099 cycles, COPYAtoB_SSEH- 2062 bytes- copy 16 BYTES+MOVZX
4116 cycles, memcpy_2-      2062 bytes- copy memcpy SSE
4368 cycles, memcpy_3-      2062 bytes- copy memcpyxmmU SSE
********** END III **********
_YZZH  mean:  436.14
_YZZG               444.57
memcpy mean: 451.86

Gunther

#121
July 09, 2014, 09:09:41 PM
Hi Rui,

the results of #47 and #48 are attached. copy.zip is the archive's name.

Gunther
You have to know the facts before you can distort them.

RuiLoureiro

#122
July 09, 2014, 09:44:04 PM
Thank you Gunther  :t

In your powerful i7, _SSEE is the best

Now, i would like to know the results for CopyString53
if you dont mind.

Quote
--------------------------------------------------------------
Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz (SSE4)
--------------------------------------------------------------
***** Time table *****

  5 cycles, MOVEAtoB_SSEF-  13 bytes- copy 128 BITS+MOVZX
  5 cycles, MOVEAtoB_SSEE-  13 bytes- copy 128 BITS+MOVZX
  6 cycles, MOVEAtoB_SSED-  13 bytes- copy 128 BITS+MOVZX
  7 cycles, MOVEAtoB_XZZF-  13 bytes- copy lenght DWORDS+MOVZX
  7 cycles, MOVEAtoB_XZZE-  13 bytes- copy lenght DWORDS+MOVZX
  7 cycles, MOVEAtoB_XZZC-  13 bytes- copy lenght DWORDS+MOVZX
 
11 cycles, MOVEAtoB_SSEE-  53 bytes- copy 128 BITS+MOVZX
12 cycles, MOVEAtoB_SSEF-  53 bytes- copy 128 BITS+MOVZX
14 cycles, MOVEAtoB_XZE-   13 bytes- copy lenght DWORDS+MOVZX
15 cycles, MOVEAtoB_SSED-  53 bytes- copy 128 BITS+MOVZX
19 cycles, MOVEAtoB_SSEE- 103 bytes- copy 128 BITS+MOVZX
20 cycles, MOVEAtoB_SSEF- 103 bytes- copy 128 BITS+MOVZX
21 cycles, MOVEAtoB_XZZE-  53 bytes- copy lenght DWORDS+MOVZX
22 cycles, MOVEAtoB_XZZC-  53 bytes- copy lenght DWORDS+MOVZX

25 cycles, MOVEAtoB_SSED- 103 bytes- copy 128 BITS+MOVZX
29 cycles, MOVEAtoB_XZE-   53 bytes- copy lenght DWORDS+MOVZX
30 cycles, MOVEAtoB_SSEH-  53 bytes- copy 128 BITS+MOVZX
32 cycles, MOVEAtoB_SSEG-  53 bytes- copy 128 BITS+MOVZX
34 cycles, MOVEAtoB_XZZF-  53 bytes- copy lenght DWORDS+MOVZX

37 cycles, MOVEAtoB_SSEE- 203 bytes- copy 128 BITS+MOVZX
37 cycles, MOVEAtoB_SSEF- 203 bytes- copy 128 BITS+MOVZX
47 cycles, MOVEAtoB_XZE-  103 bytes- copy lenght DWORDS+MOVZX
47 cycles, MOVEAtoB_XZZF- 103 bytes- copy lenght DWORDS+MOVZX
47 cycles, MOVEAtoB_XZZE- 103 bytes- copy lenght DWORDS+MOVZX
48 cycles, MOVEAtoB_SSEH- 103 bytes- copy 128 BITS+MOVZX
49 cycles, MOVEAtoB_SSED- 203 bytes- copy 128 BITS+MOVZX
50 cycles, MOVEAtoB_XZZC- 103 bytes- copy lenght DWORDS+MOVZX
52 cycles, MOVEAtoB_SSEG- 103 bytes- copy 128 BITS+MOVZX
63 cycles, MOVEAtoB_SSEH-  13 bytes- copy 128 BITS+MOVZX

63 cycles, MOVEAtoB_SSEE- 503 bytes- copy 128 BITS+MOVZX
63 cycles, MOVEAtoB_SSEF- 503 bytes- copy 128 BITS+MOVZX
66 cycles, MOVEAtoB_SSEG-  13 bytes- copy 128 BITS+MOVZX
72 cycles, MOVEAtoB_SSEG- 203 bytes- copy 128 BITS+MOVZX
78 cycles, MOVEAtoB_SSEH- 203 bytes- copy 128 BITS+MOVZX
90 cycles, MOVEAtoB_SSED- 503 bytes- copy 128 BITS+MOVZX
105 cycles, MOVEAtoB_XZZC- 203 bytes- copy lenght DWORDS+MOVZX
106 cycles, MOVEAtoB_XZE-  203 bytes- copy lenght DWORDS+MOVZX
106 cycles, MOVEAtoB_XZZE- 203 bytes- copy lenght DWORDS+MOVZX
110 cycles, MOVEAtoB_SSEH- 503 bytes- copy 128 BITS+MOVZX
128 cycles, MOVEAtoB_SSEG- 503 bytes- copy 128 BITS+MOVZX

129 cycles, MOVEAtoB_SSEF-1027 bytes- copy 128 BITS+MOVZX
129 cycles, MOVEAtoB_SSEE-1027 bytes- copy 128 BITS+MOVZX
130 cycles, MOVEAtoB_SSEG-1027 bytes- copy 128 BITS+MOVZX
136 cycles, MOVEAtoB_XZZF- 203 bytes- copy lenght DWORDS+MOVZX
158 cycles, MOVEAtoB_SSED-1027 bytes- copy 128 BITS+MOVZX
165 cycles, MOVEAtoB_SSEH-1027 bytes- copy 128 BITS+MOVZX
237 cycles, MOVEAtoB_XZZE- 503 bytes- copy lenght DWORDS+MOVZX
238 cycles, MOVEAtoB_XZZC- 503 bytes- copy lenght DWORDS+MOVZX
238 cycles, MOVEAtoB_XZE-  503 bytes- copy lenght DWORDS+MOVZX
240 cycles, MOVEAtoB_XZZF- 503 bytes- copy lenght DWORDS+MOVZX
465 cycles, MOVEAtoB_XZZC-1027 bytes- copy lenght DWORDS+MOVZX
466 cycles, MOVEAtoB_XZE- 1027 bytes- copy lenght DWORDS+MOVZX
466 cycles, MOVEAtoB_XZZE-1027 bytes- copy lenght DWORDS+MOVZX
495 cycles, MOVEAtoB_XZZF-1027 bytes- copy lenght DWORDS+MOVZX
********** END III **********

Gunther

#123
July 09, 2014, 10:10:24 PM
Hi Rui,

Quote from: RuiLoureiro on July 09, 2014, 09:44:04 PM
Now, i would like to know the results for CopyString53
if you dont mind.

no problem. Results are attached.

Gunther
You have to know the facts before you can distort them.

nidud

#124
July 10, 2014, 12:45:41 AM Last Edit: February 24, 2022, 10:23:53 PM by nidud
deleted

RuiLoureiro

#125
July 10, 2014, 01:34:05 AM
Thank you so much Gunther  :t

In this test, to copy a large amount of bytes
(.../1027/2062 bytes) the best seems to be memcpy_3

nidud,
        I will add your new procedure
        in the next post.

--------------------------
Results from Gunther
---------------------------
---------------------------------------------
    names used by nidud
---------------------------------------------
memcpy_1  -> is regcopy
memcpy_2  -> is memcpy SSE
memcpy_3  -> is memcpyxmmU SSE
---------------------------------------------
Quote
--------------------------------------------------------------
Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz (SSE4)
--------------------------------------------------------------
***** Time table *****

13 cycles, COPYAtoB_YZZH-   15 bytes- copy length 16 BYTES+MOVZX
13 cycles, crt_memcpy-      15 bytes- copy crt_memcpy
14 cycles, COPYAtoB_SSEE-   15 bytes- copy 16 BYTES+MOVZX

15 cycles, COPYAtoB_SSED-   15 bytes- copy 16 BYTES+MOVZX
15 cycles, COPYAtoB_SSEES-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
16 cycles, COPYAtoB_SSEDS-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
16 cycles, COPYAtoB_SSEFS-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
16 cycles, COPYAtoB_WZZE-   15 bytes- copy 16 BYTES+MOVZX
17 cycles, COPYAtoB_YZZE-   15 bytes- copy length 16 BYTES+MOVZX
17 cycles, COPYAtoB_SSED-   53 bytes- copy 16 BYTES+MOVZX
17 cycles, COPYAtoB_SSEES-  53 bytes- copy 16 BYTES+MOVZX- uses ESP
17 cycles, COPYAtoB_SSEF-   53 bytes- copy 16 BYTES+MOVZX
17 cycles, COPYAtoB_SSEF-   15 bytes- copy 16 BYTES+MOVZX
17 cycles, COPYAtoB_SSEFS-  53 bytes- copy 16 BYTES+MOVZX- uses ESP
17 cycles, COPYAtoB_XZZE-   15 bytes- copy lenght DWORDS+MOVZX
17 cycles, COPYAtoB_XZZC-   15 bytes- copy lenght DWORDS+MOVZX
18 cycles, COPYAtoB_XZZCS-  15 bytes- copy lenght DWORDS+MOVZX- uses ESP
18 cycles, COPYAtoB_XZES-   15 bytes- copy lenght DWORDS+MOVZX- uses ESP
18 cycles, COPYAtoB_SSEDS-  53 bytes- copy 16 BYTES+MOVZX- uses ESP
18 cycles, COPYAtoB_YZZG-   15 bytes- copy length 16 BYTES+MOVZX
18 cycles, COPYAtoB_XZZES-  15 bytes- copy lenght DWORDS+MOVZX- uses ESP
19 cycles, COPYAtoB_XZZFS-  15 bytes- copy lenght DWORDS+MOVZX- uses ESP
19 cycles, COPYAtoB_XZZF-   15 bytes- copy lenght DWORDS+MOVZX
21 cycles, COPYAtoB_SSEE-   53 bytes- copy 16 BYTES+MOVZX
22 cycles, COPYAtoB_YZZH-   53 bytes- copy length 16 BYTES+MOVZX
22 cycles, COPYAtoB_YZZG-   53 bytes- copy length 16 BYTES+MOVZX
24 cycles, COPYAtoB_WZZE-   53 bytes- copy 16 BYTES+MOVZX
25 cycles, memcpy_1-        53 bytes- copy regcopy
25 cycles, COPYAtoB_XZZFS-  53 bytes- copy lenght DWORDS+MOVZX- uses ESP
26 cycles, COPYAtoB_YZZE-   53 bytes- copy length 16 BYTES+MOVZX
26 cycles, COPYAtoB_XZZES-  53 bytes- copy lenght DWORDS+MOVZX- uses ESP
26 cycles, COPYAtoB_XZZF-   53 bytes- copy lenght DWORDS+MOVZX
27 cycles, COPYAtoB_XZZE-   53 bytes- copy lenght DWORDS+MOVZX
28 cycles, COPYAtoB_XZZC-   53 bytes- copy lenght DWORDS+MOVZX
28 cycles, memcpy_1-        15 bytes- copy regcopy
31 cycles, COPYAtoB_SSEGS-  53 bytes- copy 16 BYTES+MOVZX- uses ESP
32 cycles, COPYAtoB_SSEG-   53 bytes- copy 16 BYTES+MOVZX
32 cycles, memcpy_2-        15 bytes- copy memcpy SSE
32 cycles, COPYAtoB_SSEH-   53 bytes- copy 16 BYTES+MOVZX
34 cycles, memcpy_3-        15 bytes- copy memcpyxmmU SSE
34 cycles, COPYAtoB_SSEHS-  53 bytes- copy 16 BYTES+MOVZX- uses ESP
36 cycles, COPYAtoB_XZZCS-  53 bytes- copy lenght DWORDS+MOVZX- uses ESP
37 cycles, crt_memcpy-      53 bytes- copy crt_memcpy

37 cycles, COPYAtoB_YZZG-  103 bytes- copy length 16 BYTES+MOVZX
37 cycles, COPYAtoB_XZES-   53 bytes- copy lenght DWORDS+MOVZX- uses ESP
37 cycles, COPYAtoB_YZZH-  103 bytes- copy length 16 BYTES+MOVZX
38 cycles, memcpy_2-        53 bytes- copy memcpy SSE
38 cycles, COPYAtoB_SSEES- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
38 cycles, COPYAtoB_SSEDS- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
38 cycles, COPYAtoB_SSEFS- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
39 cycles, memcpy_1-       103 bytes- copy regcopy
39 cycles, memcpy_3-        53 bytes- copy memcpyxmmU SSE
40 cycles, COPYAtoB_SSED-  103 bytes- copy 16 BYTES+MOVZX
40 cycles, COPYAtoB_SSEF-  103 bytes- copy 16 BYTES+MOVZX
40 cycles, COPYAtoB_WZZE-  103 bytes- copy 16 BYTES+MOVZX
40 cycles, COPYAtoB_SSEE-  103 bytes- copy 16 BYTES+MOVZX
41 cycles, COPYAtoB_XZE-    15 bytes- copy lenght DWORDS+MOVZX
42 cycles, COPYAtoB_YZZE-  103 bytes- copy length 16 BYTES+MOVZX
47 cycles, crt_memcpy-     103 bytes- copy crt_memcpy
48 cycles, COPYAtoB_SSEGS- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
48 cycles, COPYAtoB_SSEH-  103 bytes- copy 16 BYTES+MOVZX
49 cycles, memcpy_2-       103 bytes- copy memcpy SSE
49 cycles, COPYAtoB_XZZES- 103 bytes- copy lenght DWORDS+MOVZX- uses ESP
50 cycles, COPYAtoB_XZZFS- 103 bytes- copy lenght DWORDS+MOVZX- uses ESP
50 cycles, COPYAtoB_SSEHS- 103 bytes- copy 16 BYTES+MOVZX- uses ESP
51 cycles, memcpy_3-       103 bytes- copy memcpyxmmU SSE  <<<<<<<-----

51 cycles, COPYAtoB_SSEES- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
51 cycles, COPYAtoB_XZZE-  103 bytes- copy lenght DWORDS+MOVZX
51 cycles, COPYAtoB_SSEFS- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
52 cycles, COPYAtoB_SSED-  203 bytes- copy 16 BYTES+MOVZX
52 cycles, COPYAtoB_SSEDS- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
52 cycles, COPYAtoB_SSEG-  103 bytes- copy 16 BYTES+MOVZX
52 cycles, COPYAtoB_XZZC-  103 bytes- copy lenght DWORDS+MOVZX
52 cycles, COPYAtoB_XZZF-  103 bytes- copy lenght DWORDS+MOVZX
52 cycles, COPYAtoB_SSEF-  203 bytes- copy 16 BYTES+MOVZX
52 cycles, COPYAtoB_XZE-   103 bytes- copy lenght DWORDS+MOVZX
52 cycles, COPYAtoB_SSEE-  203 bytes- copy 16 BYTES+MOVZX
55 cycles, memcpy_3-       203 bytes- copy memcpyxmmU SSE      <<<<<<<-----
59 cycles, COPYAtoB_XZE-    53 bytes- copy lenght DWORDS+MOVZX
59 cycles, memcpy_2-       203 bytes- copy memcpy SSE
64 cycles, COPYAtoB_YZZH-  203 bytes- copy length 16 BYTES+MOVZX
64 cycles, COPYAtoB_YZZG-  203 bytes- copy length 16 BYTES+MOVZX
65 cycles, COPYAtoB_WZZE-  203 bytes- copy 16 BYTES+MOVZX
68 cycles, COPYAtoB_YZZE-  203 bytes- copy length 16 BYTES+MOVZX
68 cycles, memcpy_1-       203 bytes- copy regcopy
69 cycles, crt_memcpy-     203 bytes- copy crt_memcpy

69 cycles, memcpy_3-       503 bytes- copy memcpyxmmU SSE
70 cycles, COPYAtoB_SSEG-  203 bytes- copy 16 BYTES+MOVZX
72 cycles, COPYAtoB_XZZCS- 103 bytes- copy lenght DWORDS+MOVZX- uses ESP
72 cycles, COPYAtoB_SSEE-  503 bytes- copy 16 BYTES+MOVZX
72 cycles, COPYAtoB_XZES-  103 bytes- copy lenght DWORDS+MOVZX- uses ESP
75 cycles, COPYAtoB_SSEGS- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
76 cycles, COPYAtoB_SSEH-   15 bytes- copy 16 BYTES+MOVZX
79 cycles, COPYAtoB_SSEGS-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
80 cycles, COPYAtoB_SSEG-   15 bytes- copy 16 BYTES+MOVZX
82 cycles, COPYAtoB_SSEH-  203 bytes- copy 16 BYTES+MOVZX
84 cycles, COPYAtoB_SSEGS- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
87 cycles, COPYAtoB_SSEHS- 203 bytes- copy 16 BYTES+MOVZX- uses ESP
88 cycles, COPYAtoB_SSEG-  503 bytes- copy 16 BYTES+MOVZX
92 cycles, COPYAtoB_SSEHS-  15 bytes- copy 16 BYTES+MOVZX- uses ESP
93 cycles, crt_memcpy-     503 bytes- copy crt_memcpy
93 cycles, COPYAtoB_SSED-  503 bytes- copy 16 BYTES+MOVZX
94 cycles, COPYAtoB_SSEFS- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
94 cycles, COPYAtoB_SSEDS- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
94 cycles, COPYAtoB_SSEF-  503 bytes- copy 16 BYTES+MOVZX
107 cycles, COPYAtoB_SSEES- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
108 cycles, COPYAtoB_XZZFS- 203 bytes- copy lenght DWORDS+MOVZX- uses ESP
109 cycles, COPYAtoB_XZZES- 203 bytes- copy lenght DWORDS+MOVZX- uses ESP
110 cycles, COPYAtoB_XZZF-  203 bytes- copy lenght DWORDS+MOVZX
110 cycles, COPYAtoB_XZZC-  203 bytes- copy lenght DWORDS+MOVZX
111 cycles, COPYAtoB_XZZE-  203 bytes- copy lenght DWORDS+MOVZX
111 cycles, COPYAtoB_SSEH-  503 bytes- copy 16 BYTES+MOVZX
111 cycles, COPYAtoB_XZE-   203 bytes- copy lenght DWORDS+MOVZX

115 cycles, memcpy_3-      1027 bytes- copy memcpyxmmU SSE
129 cycles, COPYAtoB_SSEHS- 503 bytes- copy 16 BYTES+MOVZX- uses ESP
130 cycles, memcpy_2-       503 bytes- copy memcpy SSE
135 cycles, COPYAtoB_SSEE- 1027 bytes- copy 16 BYTES+MOVZX

137 cycles, crt_memcpy-    1027 bytes- copy crt_memcpy
137 cycles, COPYAtoB_XZES-  203 bytes- copy lenght DWORDS+MOVZX- uses ESP
137 cycles, COPYAtoB_XZZCS- 203 bytes- copy lenght DWORDS+MOVZX- uses ESP
143 cycles, COPYAtoB_YZZE-  503 bytes- copy length 16 BYTES+MOVZX
145 cycles, memcpy_1-       503 bytes- copy regcopy
150 cycles, COPYAtoB_YZZH-  503 bytes- copy length 16 BYTES+MOVZX
150 cycles, COPYAtoB_YZZG-  503 bytes- copy length 16 BYTES+MOVZX
153 cycles, COPYAtoB_SSEG- 1027 bytes- copy 16 BYTES+MOVZX
153 cycles, COPYAtoB_SSEGS-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
160 cycles, COPYAtoB_SSEDS-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
161 cycles, COPYAtoB_SSED- 1027 bytes- copy 16 BYTES+MOVZX
161 cycles, COPYAtoB_SSEF- 1027 bytes- copy 16 BYTES+MOVZX
162 cycles, COPYAtoB_SSEFS-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
162 cycles, COPYAtoB_SSEES-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
165 cycles, COPYAtoB_SSEHS-1027 bytes- copy 16 BYTES+MOVZX- uses ESP
169 cycles, COPYAtoB_WZZE-  503 bytes- copy 16 BYTES+MOVZX
170 cycles, COPYAtoB_SSEH- 1027 bytes- copy 16 BYTES+MOVZX
185 cycles, memcpy_2-      1027 bytes- copy memcpy SSE

192 cycles, memcpy_3-      2062 bytes- copy memcpyxmmU SSE
216 cycles, crt_memcpy-    2062 bytes- copy crt_memcpy
239 cycles, COPYAtoB_XZZES- 503 bytes- copy lenght DWORDS+MOVZX- uses ESP
241 cycles, COPYAtoB_XZZE-  503 bytes- copy lenght DWORDS+MOVZX
241 cycles, COPYAtoB_XZE-   503 bytes- copy lenght DWORDS+MOVZX
242 cycles, COPYAtoB_XZZC-  503 bytes- copy lenght DWORDS+MOVZX
242 cycles, COPYAtoB_XZZFS- 503 bytes- copy lenght DWORDS+MOVZX- uses ESP
243 cycles, COPYAtoB_XZZF-  503 bytes- copy lenght DWORDS+MOVZX
268 cycles, COPYAtoB_XZES-  503 bytes- copy lenght DWORDS+MOVZX- uses ESP
269 cycles, COPYAtoB_XZZCS- 503 bytes- copy lenght DWORDS+MOVZX- uses ESP
275 cycles, COPYAtoB_SSEE- 2062 bytes- copy 16 BYTES+MOVZX

278 cycles, COPYAtoB_YZZE- 1027 bytes- copy length 16 BYTES+MOVZX
283 cycles, COPYAtoB_YZZH- 1027 bytes- copy length 16 BYTES+MOVZX
284 cycles, COPYAtoB_YZZG- 1027 bytes- copy length 16 BYTES+MOVZX
286 cycles, memcpy_1-      1027 bytes- copy regcopy

292 cycles, COPYAtoB_SSED- 2062 bytes- copy 16 BYTES+MOVZX
293 cycles, COPYAtoB_SSEF- 2062 bytes- copy 16 BYTES+MOVZX
294 cycles, COPYAtoB_SSEES-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
294 cycles, COPYAtoB_SSEDS-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
294 cycles, COPYAtoB_SSEFS-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
296 cycles, memcpy_2-      2062 bytes- copy memcpy SSE
310 cycles, COPYAtoB_SSEG- 2062 bytes- copy 16 BYTES+MOVZX
319 cycles, COPYAtoB_SSEGS-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
329 cycles, COPYAtoB_WZZE- 1027 bytes- copy 16 BYTES+MOVZX
331 cycles, COPYAtoB_SSEHS-2062 bytes- copy 16 BYTES+MOVZX- uses ESP
331 cycles, COPYAtoB_SSEH- 2062 bytes- copy 16 BYTES+MOVZX
468 cycles, COPYAtoB_XZZFS-1027 bytes- copy lenght DWORDS+MOVZX- uses ESP
469 cycles, COPYAtoB_XZZES-1027 bytes- copy lenght DWORDS+MOVZX- uses ESP
470 cycles, COPYAtoB_XZZF- 1027 bytes- copy lenght DWORDS+MOVZX
471 cycles, COPYAtoB_XZZE- 1027 bytes- copy lenght DWORDS+MOVZX
471 cycles, COPYAtoB_XZZC- 1027 bytes- copy lenght DWORDS+MOVZX
473 cycles, COPYAtoB_XZE-  1027 bytes- copy lenght DWORDS+MOVZX
497 cycles, COPYAtoB_XZES- 1027 bytes- copy lenght DWORDS+MOVZX- uses ESP
499 cycles, COPYAtoB_XZZCS-1027 bytes- copy lenght DWORDS+MOVZX- uses ESP
509 cycles, COPYAtoB_YZZE- 2062 bytes- copy length 16 BYTES+MOVZX
568 cycles, COPYAtoB_YZZH- 2062 bytes- copy length 16 BYTES+MOVZX
570 cycles, COPYAtoB_YZZG- 2062 bytes- copy length 16 BYTES+MOVZX
583 cycles, memcpy_1-      2062 bytes- copy regcopy
677 cycles, COPYAtoB_WZZE- 2062 bytes- copy 16 BYTES+MOVZX
919 cycles, COPYAtoB_XZZFS-2062 bytes- copy lenght DWORDS+MOVZX- uses ESP
919 cycles, COPYAtoB_XZZES-2062 bytes- copy lenght DWORDS+MOVZX- uses ESP
920 cycles, COPYAtoB_XZZE- 2062 bytes- copy lenght DWORDS+MOVZX
921 cycles, COPYAtoB_XZE-  2062 bytes- copy lenght DWORDS+MOVZX
922 cycles, COPYAtoB_XZZC- 2062 bytes- copy lenght DWORDS+MOVZX
922 cycles, COPYAtoB_XZZF- 2062 bytes- copy lenght DWORDS+MOVZX
948 cycles, COPYAtoB_XZZCS-2062 bytes- copy lenght DWORDS+MOVZX- uses ESP
948 cycles, COPYAtoB_XZES- 2062 bytes- copy lenght DWORDS+MOVZX- uses ESP
********** END III **********

RuiLoureiro

#126
July 10, 2014, 03:03:09 AM
Hi,
        Now, i used crt_memcpy, memcpy_1, memcpy_2, memcpy_3, memcpy_4
        procedures and some SSE procedures.
       
        Gunther, could you post the results of your i7,
        if you dont mind ?
        Thank you.
       
        These are my results.
Quote
NOT ALIGNED
----------------------------------------------------
Intel(R) Pentium(R) 4 CPU 3.40GHz (SSE3)
-----------------------------------------------------
***** Time table *****

  34 cycles, COPYAtoB_SSEE-  15 bytes- copy 16 BYTES+MOVZX
  35 cycles, COPYAtoB_YZZI-  15 bytes- copy length 16 BYTES+MOVZX
  35 cycles, crt_memcpy-     15 bytes- copy crt_memcpy
  35 cycles, COPYAtoB_XZZF-  15 bytes- copy lenght DWORDS+MOVZX
  37 cycles, COPYAtoB_WZZE-  15 bytes- copy 16 BYTES+MOVZX
  37 cycles, COPYAtoB_YZZH-  15 bytes- copy length 16 BYTES+MOVZX
  37 cycles, COPYAtoB_XZZE-  15 bytes- copy lenght DWORDS+MOVZX
  37 cycles, COPYAtoB_XZZC-  15 bytes- copy lenght DWORDS+MOVZX
  39 cycles, COPYAtoB_YZZG-  15 bytes- copy length 16 BYTES+MOVZX
  39 cycles, COPYAtoB_YZZE-  15 bytes- copy length 16 BYTES+MOVZX
  47 cycles, COPYAtoB_XZE-   15 bytes- copy lenght DWORDS+MOVZX
  50 cycles, COPYAtoB_SSEI-  15 bytes- copy 16 BYTES+MOVZX
  51 cycles, COPYAtoB_SSEJ-  15 bytes- copy 16 BYTES+MOVZX
  61 cycles, COPYAtoB_SSEH-  15 bytes- copy 16 BYTES+MOVZX
  68 cycles, memcpy_1-       53 bytes- copy regcopy
  72 cycles, memcpy_1-       15 bytes- copy regcopy
 
  72 cycles, COPYAtoB_YZZI-  53 bytes- copy length 16 BYTES+MOVZX
  73 cycles, memcpy_4-       15 bytes- copy memcpy * 8
  73 cycles, COPYAtoB_YZZH-  53 bytes- copy length 16 BYTES+MOVZX
  75 cycles, COPYAtoB_YZZG-  53 bytes- copy length 16 BYTES+MOVZX
  75 cycles, COPYAtoB_XZZC-  53 bytes- copy lenght DWORDS+MOVZX
  76 cycles, COPYAtoB_XZE-   53 bytes- copy lenght DWORDS+MOVZX
  76 cycles, COPYAtoB_XZZF-  53 bytes- copy lenght DWORDS+MOVZX
  78 cycles, COPYAtoB_XZZE-  53 bytes- copy lenght DWORDS+MOVZX
  78 cycles, COPYAtoB_YZZE-  53 bytes- copy length 16 BYTES+MOVZX
  89 cycles, COPYAtoB_WZZE-  53 bytes- copy 16 BYTES+MOVZX
  90 cycles, COPYAtoB_SSEJ-  53 bytes- copy 16 BYTES+MOVZX
  94 cycles, COPYAtoB_SSEI-  53 bytes- copy 16 BYTES+MOVZX
  98 cycles, COPYAtoB_SSEH-  53 bytes- copy 16 BYTES+MOVZX
100 cycles, COPYAtoB_SSEE-  53 bytes- copy 16 BYTES+MOVZX

103 cycles, memcpy_4-       53 bytes- copy memcpy * 8
106 cycles, memcpy_1-      103 bytes- copy regcopy
108 cycles, COPYAtoB_YZZG- 103 bytes- copy length 16 BYTES+MOVZX
108 cycles, COPYAtoB_YZZI- 103 bytes- copy length 16 BYTES+MOVZX
108 cycles, crt_memcpy-     53 bytes- copy crt_memcpy
109 cycles, COPYAtoB_YZZH- 103 bytes- copy length 16 BYTES+MOVZX
111 cycles, COPYAtoB_YZZE- 103 bytes- copy length 16 BYTES+MOVZX
112 cycles, COPYAtoB_WZZE- 103 bytes- copy 16 BYTES+MOVZX
121 cycles, COPYAtoB_XZZF- 103 bytes- copy lenght DWORDS+MOVZX
122 cycles, COPYAtoB_XZZC- 103 bytes- copy lenght DWORDS+MOVZX
125 cycles, COPYAtoB_XZE-  103 bytes- copy lenght DWORDS+MOVZX
130 cycles, COPYAtoB_XZZE- 103 bytes- copy lenght DWORDS+MOVZX
132 cycles, memcpy_4-      103 bytes- copy memcpy * 8
139 cycles, crt_memcpy-    103 bytes- copy crt_memcpy

172 cycles, COPYAtoB_YZZG- 203 bytes- copy length 16 BYTES+MOVZX
173 cycles, COPYAtoB_YZZI- 203 bytes- copy length 16 BYTES+MOVZX
173 cycles, COPYAtoB_YZZE- 203 bytes- copy length 16 BYTES+MOVZX
174 cycles, COPYAtoB_YZZH- 203 bytes- copy length 16 BYTES+MOVZX
177 cycles, memcpy_1-      203 bytes- copy regcopy
206 cycles, COPYAtoB_SSEJ- 103 bytes- copy 16 BYTES+MOVZX
209 cycles, COPYAtoB_SSEI- 103 bytes- copy 16 BYTES+MOVZX
211 cycles, COPYAtoB_SSEE- 103 bytes- copy 16 BYTES+MOVZX
212 cycles, COPYAtoB_WZZE- 203 bytes- copy 16 BYTES+MOVZX
212 cycles, memcpy_4-      203 bytes- copy memcpy * 8
213 cycles, COPYAtoB_XZZE- 203 bytes- copy lenght DWORDS+MOVZX
213 cycles, COPYAtoB_XZZC- 203 bytes- copy lenght DWORDS+MOVZX
214 cycles, COPYAtoB_SSEH- 103 bytes- copy 16 BYTES+MOVZX
214 cycles, COPYAtoB_XZZF- 203 bytes- copy lenght DWORDS+MOVZX
225 cycles, crt_memcpy-    203 bytes- copy crt_memcpy
231 cycles, COPYAtoB_XZE-  203 bytes- copy lenght DWORDS+MOVZX
250 cycles, memcpy_2-       15 bytes- copy memcpy SSE
256 cycles, memcpy_3-       15 bytes- copy memcpyxmmU SSE
277 cycles, memcpy_3-       53 bytes- copy memcpyxmmU SSE
278 cycles, memcpy_2-       53 bytes- copy memcpy SSE
379 cycles, memcpy_3-      103 bytes- copy memcpyxmmU SSE
393 cycles, memcpy_2-      103 bytes- copy memcpy SSE
394 cycles, COPYAtoB_YZZI- 503 bytes- copy length 16 BYTES+MOVZX
395 cycles, COPYAtoB_YZZG- 503 bytes- copy length 16 BYTES+MOVZX
398 cycles, COPYAtoB_YZZH- 503 bytes- copy length 16 BYTES+MOVZX

412 cycles, COPYAtoB_SSEE- 203 bytes- copy 16 BYTES+MOVZX
414 cycles, memcpy_1-      503 bytes- copy regcopy
418 cycles, COPYAtoB_YZZE- 503 bytes- copy length 16 BYTES+MOVZX
427 cycles, COPYAtoB_XZZC- 503 bytes- copy lenght DWORDS+MOVZX
428 cycles, COPYAtoB_WZZE- 503 bytes- copy 16 BYTES+MOVZX
429 cycles, COPYAtoB_SSEH- 203 bytes- copy 16 BYTES+MOVZX
430 cycles, COPYAtoB_SSEI- 203 bytes- copy 16 BYTES+MOVZX
431 cycles, COPYAtoB_SSEJ- 203 bytes- copy 16 BYTES+MOVZX

436 cycles, memcpy_4-      503 bytes- copy memcpy * 8
442 cycles, COPYAtoB_XZZF- 503 bytes- copy lenght DWORDS+MOVZX
444 cycles, COPYAtoB_XZZE- 503 bytes- copy lenght DWORDS+MOVZX
454 cycles, crt_memcpy-    503 bytes- copy crt_memcpy
503 cycles, COPYAtoB_XZE-  503 bytes- copy lenght DWORDS+MOVZX
620 cycles, memcpy_2-      203 bytes- copy memcpy SSE
631 cycles, memcpy_3-      203 bytes- copy memcpyxmmU SSE

726 cycles, memcpy_1-     1027 bytes- copy regcopy
734 cycles, COPYAtoB_YZZG-1027 bytes- copy length 16 BYTES+MOVZX
769 cycles, COPYAtoB_YZZI-1027 bytes- copy length 16 BYTES+MOVZX
787 cycles, COPYAtoB_YZZH-1027 bytes- copy length 16 BYTES+MOVZX
788 cycles, COPYAtoB_YZZE-1027 bytes- copy length 16 BYTES+MOVZX
808 cycles, COPYAtoB_XZZE-1027 bytes- copy lenght DWORDS+MOVZX
815 cycles, COPYAtoB_XZE- 1027 bytes- copy lenght DWORDS+MOVZX
815 cycles, COPYAtoB_XZZC-1027 bytes- copy lenght DWORDS+MOVZX
817 cycles, COPYAtoB_WZZE-1027 bytes- copy 16 BYTES+MOVZX
825 cycles, memcpy_4-     1027 bytes- copy memcpy * 8
835 cycles, crt_memcpy-   1027 bytes- copy crt_memcpy
851 cycles, COPYAtoB_XZZF-1027 bytes- copy lenght DWORDS+MOVZX

965 cycles, COPYAtoB_SSEH- 503 bytes- copy 16 BYTES+MOVZX
968 cycles, COPYAtoB_SSEJ- 503 bytes- copy 16 BYTES+MOVZX
970 cycles, COPYAtoB_SSEI- 503 bytes- copy 16 BYTES+MOVZX
989 cycles, COPYAtoB_SSEE- 503 bytes- copy 16 BYTES+MOVZX
1142 cycles, memcpy_2-      503 bytes- copy memcpy SSE
1163 cycles, memcpy_3-      503 bytes- copy memcpyxmmU SSE
1529 cycles, COPYAtoB_YZZI-2062 bytes- copy length 16 BYTES+MOVZX
1538 cycles, COPYAtoB_YZZG-2062 bytes- copy length 16 BYTES+MOVZX
1563 cycles, COPYAtoB_XZZC-2062 bytes- copy lenght DWORDS+MOVZX
1581 cycles, COPYAtoB_YZZE-2062 bytes- copy length 16 BYTES+MOVZX
1585 cycles, COPYAtoB_WZZE-2062 bytes- copy 16 BYTES+MOVZX
1603 cycles, memcpy_1-     2062 bytes- copy regcopy
1622 cycles, COPYAtoB_XZZF-2062 bytes- copy lenght DWORDS+MOVZX
1622 cycles, memcpy_4-     2062 bytes- copy memcpy * 8
1626 cycles, crt_memcpy-   2062 bytes- copy crt_memcpy
1626 cycles, COPYAtoB_XZE- 2062 bytes- copy lenght DWORDS+MOVZX
1630 cycles, COPYAtoB_XZZE-2062 bytes- copy lenght DWORDS+MOVZX
1734 cycles, COPYAtoB_YZZH-2062 bytes- copy length 16 BYTES+MOVZX
2029 cycles, COPYAtoB_SSEE-1027 bytes- copy 16 BYTES+MOVZX
2060 cycles, COPYAtoB_SSEH-1027 bytes- copy 16 BYTES+MOVZX
2074 cycles, COPYAtoB_SSEJ-1027 bytes- copy 16 BYTES+MOVZX
2097 cycles, memcpy_3-     1027 bytes- copy memcpyxmmU SSE
2119 cycles, COPYAtoB_SSEI-1027 bytes- copy 16 BYTES+MOVZX
2123 cycles, memcpy_2-     1027 bytes- copy memcpy SSE
4043 cycles, COPYAtoB_SSEE-2062 bytes- copy 16 BYTES+MOVZX
4065 cycles, COPYAtoB_SSEI-2062 bytes- copy 16 BYTES+MOVZX
4070 cycles, COPYAtoB_SSEH-2062 bytes- copy 16 BYTES+MOVZX
4086 cycles, COPYAtoB_SSEJ-2062 bytes- copy 16 BYTES+MOVZX
4088 cycles, memcpy_2-     2062 bytes- copy memcpy SSE
4391 cycles, memcpy_3-     2062 bytes- copy memcpyxmmU SSE
********** END III **********

Gunther

#127
July 10, 2014, 04:06:42 AM
Hi Rui,

the results for CopyString54 are attached as CopyStrin54.zip. I hope that helps:

Gunther
You have to know the facts before you can distort them.

RuiLoureiro

#128
July 10, 2014, 04:44:14 AM
Hi Gunther,
            It helps, thank you.  :t
            Now i want to test another procedure.         
            Could you run CopyString55, please ?
            Thanks.
       
These are my results.
Quote
-----------------------------------------------------
Intel(R) Pentium(R) 4 CPU 3.40GHz (SSE3)
----------------------------------------------------
***** Time table *****

34 cycles, COPYAtoB_YZZI-   15 bytes- copy length 16 BYTES+MOVZX
35 cycles, COPYAtoB_SSEE-   15 bytes- copy 16 BYTES+MOVZX
35 cycles, crt_memcpy-      15 bytes- copy crt_memcpy
35 cycles, COPYAtoB_XZZF-   15 bytes- copy lenght DWORDS+MOVZX
36 cycles, COPYAtoB_YZZH-   15 bytes- copy length 16 BYTES+MOVZX
36 cycles, COPYAtoB_XZZC-   15 bytes- copy lenght DWORDS+MOVZX
37 cycles, COPYAtoB_XZZE-   15 bytes- copy lenght DWORDS+MOVZX
37 cycles, COPYAtoB_WZZE-   15 bytes- copy 16 BYTES+MOVZX
39 cycles, COPYAtoB_YZZG-   15 bytes- copy length 16 BYTES+MOVZX
39 cycles, COPYAtoB_YZZE-   15 bytes- copy length 16 BYTES+MOVZX
44 cycles, COPYAtoB_XZE-    15 bytes- copy lenght DWORDS+MOVZX
48 cycles, COPYAtoB_SSEK-   15 bytes- copy 16 BYTES+MOVZX
50 cycles, COPYAtoB_SSEJ-   15 bytes- copy 16 BYTES+MOVZX
51 cycles, COPYAtoB_SSEI-   15 bytes- copy 16 BYTES+MOVZX
59 cycles, COPYAtoB_SSEH-   15 bytes- copy 16 BYTES+MOVZX
68 cycles, memcpy_1-        53 bytes- copy regcopy
71 cycles, COPYAtoB_YZZI-   53 bytes- copy length 16 BYTES+MOVZX
71 cycles, COPYAtoB_YZZG-   53 bytes- copy length 16 BYTES+MOVZX
71 cycles, memcpy_4-        15 bytes- copy memcpy * 8
74 cycles, COPYAtoB_WZZE-   53 bytes- copy 16 BYTES+MOVZX
75 cycles, COPYAtoB_XZZE-   53 bytes- copy lenght DWORDS+MOVZX
75 cycles, COPYAtoB_XZZF-   53 bytes- copy lenght DWORDS+MOVZX
77 cycles, COPYAtoB_YZZE-   53 bytes- copy length 16 BYTES+MOVZX
79 cycles, COPYAtoB_XZZC-   53 bytes- copy lenght DWORDS+MOVZX
81 cycles, memcpy_1-        15 bytes- copy regcopy
83 cycles, COPYAtoB_XZE-    53 bytes- copy lenght DWORDS+MOVZX
84 cycles, COPYAtoB_SSEK-   53 bytes- copy 16 BYTES+MOVZX
89 cycles, COPYAtoB_SSEI-   53 bytes- copy 16 BYTES+MOVZX
91 cycles, COPYAtoB_SSEE-   53 bytes- copy 16 BYTES+MOVZX
94 cycles, COPYAtoB_SSEJ-   53 bytes- copy 16 BYTES+MOVZX
100 cycles, COPYAtoB_YZZH-   53 bytes- copy length 16 BYTES+MOVZX
100 cycles, COPYAtoB_SSEH-   53 bytes- copy 16 BYTES+MOVZX
102 cycles, memcpy_4-        53 bytes- copy memcpy * 8
107 cycles, crt_memcpy-      53 bytes- copy crt_memcpy
107 cycles, memcpy_1-       103 bytes- copy regcopy
108 cycles, COPYAtoB_YZZG-  103 bytes- copy length 16 BYTES+MOVZX
108 cycles, COPYAtoB_YZZI-  103 bytes- copy length 16 BYTES+MOVZX
110 cycles, COPYAtoB_YZZE-  103 bytes- copy length 16 BYTES+MOVZX
110 cycles, COPYAtoB_YZZH-  103 bytes- copy length 16 BYTES+MOVZX
112 cycles, COPYAtoB_WZZE-  103 bytes- copy 16 BYTES+MOVZX
121 cycles, COPYAtoB_XZZF-  103 bytes- copy lenght DWORDS+MOVZX
122 cycles, COPYAtoB_XZZC-  103 bytes- copy lenght DWORDS+MOVZX
123 cycles, COPYAtoB_XZZE-  103 bytes- copy lenght DWORDS+MOVZX
133 cycles, memcpy_4-       103 bytes- copy memcpy * 8
134 cycles, COPYAtoB_XZE-   103 bytes- copy lenght DWORDS+MOVZX
139 cycles, crt_memcpy-     103 bytes- copy crt_memcpy
170 cycles, COPYAtoB_YZZG-  203 bytes- copy length 16 BYTES+MOVZX
175 cycles, COPYAtoB_YZZH-  203 bytes- copy length 16 BYTES+MOVZX
175 cycles, memcpy_1-       203 bytes- copy regcopy
179 cycles, COPYAtoB_YZZE-  203 bytes- copy length 16 BYTES+MOVZX
188 cycles, COPYAtoB_YZZI-  203 bytes- copy length 16 BYTES+MOVZX
198 cycles, COPYAtoB_SSEK-  103 bytes- copy 16 BYTES+MOVZX
207 cycles, COPYAtoB_SSEI-  103 bytes- copy 16 BYTES+MOVZX
210 cycles, COPYAtoB_WZZE-  203 bytes- copy 16 BYTES+MOVZX
211 cycles, COPYAtoB_SSEE-  103 bytes- copy 16 BYTES+MOVZX
211 cycles, COPYAtoB_XZZC-  203 bytes- copy lenght DWORDS+MOVZX
212 cycles, COPYAtoB_SSEH-  103 bytes- copy 16 BYTES+MOVZX
212 cycles, COPYAtoB_XZZE-  203 bytes- copy lenght DWORDS+MOVZX
213 cycles, memcpy_4-       203 bytes- copy memcpy * 8
214 cycles, COPYAtoB_XZZF-  203 bytes- copy lenght DWORDS+MOVZX
216 cycles, COPYAtoB_SSEJ-  103 bytes- copy 16 BYTES+MOVZX
224 cycles, crt_memcpy-     203 bytes- copy crt_memcpy
233 cycles, COPYAtoB_XZE-   203 bytes- copy lenght DWORDS+MOVZX
250 cycles, memcpy_2-        15 bytes- copy memcpy SSE
254 cycles, memcpy_3-        15 bytes- copy memcpyxmmU SSE
276 cycles, memcpy_3-        53 bytes- copy memcpyxmmU SSE
299 cycles, memcpy_2-        53 bytes- copy memcpy SSE
381 cycles, memcpy_3-       103 bytes- copy memcpyxmmU SSE
392 cycles, COPYAtoB_YZZG-  503 bytes- copy length 16 BYTES+MOVZX
392 cycles, COPYAtoB_YZZI-  503 bytes- copy length 16 BYTES+MOVZX
393 cycles, memcpy_1-       503 bytes- copy regcopy
397 cycles, COPYAtoB_YZZH-  503 bytes- copy length 16 BYTES+MOVZX
403 cycles, memcpy_2-       103 bytes- copy memcpy SSE
408 cycles, COPYAtoB_YZZE-  503 bytes- copy length 16 BYTES+MOVZX
408 cycles, COPYAtoB_SSEE-  203 bytes- copy 16 BYTES+MOVZX
421 cycles, COPYAtoB_XZZC-  503 bytes- copy lenght DWORDS+MOVZX
424 cycles, COPYAtoB_XZZE-  503 bytes- copy lenght DWORDS+MOVZX
428 cycles, COPYAtoB_SSEH-  203 bytes- copy 16 BYTES+MOVZX
429 cycles, COPYAtoB_SSEI-  203 bytes- copy 16 BYTES+MOVZX
435 cycles, COPYAtoB_SSEJ-  203 bytes- copy 16 BYTES+MOVZX
437 cycles, COPYAtoB_WZZE-  503 bytes- copy 16 BYTES+MOVZX
440 cycles, crt_memcpy-     503 bytes- copy crt_memcpy
447 cycles, COPYAtoB_XZZF-  503 bytes- copy lenght DWORDS+MOVZX
448 cycles, COPYAtoB_XZE-   503 bytes- copy lenght DWORDS+MOVZX
452 cycles, COPYAtoB_SSEK-  203 bytes- copy 16 BYTES+MOVZX
464 cycles, memcpy_4-       503 bytes- copy memcpy * 8
623 cycles, memcpy_2-       203 bytes- copy memcpy SSE
631 cycles, memcpy_3-       203 bytes- copy memcpyxmmU SSE
728 cycles, COPYAtoB_YZZI- 1027 bytes- copy length 16 BYTES+MOVZX
732 cycles, COPYAtoB_YZZG- 1027 bytes- copy length 16 BYTES+MOVZX
732 cycles, memcpy_1-      1027 bytes- copy regcopy
733 cycles, COPYAtoB_YZZH- 1027 bytes- copy length 16 BYTES+MOVZX
787 cycles, COPYAtoB_YZZE- 1027 bytes- copy length 16 BYTES+MOVZX
806 cycles, COPYAtoB_XZZF- 1027 bytes- copy lenght DWORDS+MOVZX
807 cycles, COPYAtoB_XZE-  1027 bytes- copy lenght DWORDS+MOVZX
830 cycles, memcpy_4-      1027 bytes- copy memcpy * 8
830 cycles, COPYAtoB_XZZE- 1027 bytes- copy lenght DWORDS+MOVZX
850 cycles, crt_memcpy-    1027 bytes- copy crt_memcpy
864 cycles, COPYAtoB_WZZE- 1027 bytes- copy 16 BYTES+MOVZX
874 cycles, COPYAtoB_XZZC- 1027 bytes- copy lenght DWORDS+MOVZX
960 cycles, COPYAtoB_SSEK-  503 bytes- copy 16 BYTES+MOVZX
964 cycles, COPYAtoB_SSEI-  503 bytes- copy 16 BYTES+MOVZX
965 cycles, COPYAtoB_SSEJ-  503 bytes- copy 16 BYTES+MOVZX
965 cycles, COPYAtoB_SSEH-  503 bytes- copy 16 BYTES+MOVZX
986 cycles, COPYAtoB_SSEE-  503 bytes- copy 16 BYTES+MOVZX
1166 cycles, memcpy_2-      503 bytes- copy memcpy SSE
1241 cycles, memcpy_3-      503 bytes- copy memcpyxmmU SSE
1526 cycles, COPYAtoB_YZZH-2062 bytes- copy length 16 BYTES+MOVZX
1527 cycles, COPYAtoB_YZZG-2062 bytes- copy length 16 BYTES+MOVZX
1536 cycles, COPYAtoB_YZZI-2062 bytes- copy length 16 BYTES+MOVZX
1556 cycles, memcpy_1-     2062 bytes- copy regcopy
1559 cycles, COPYAtoB_XZZE-2062 bytes- copy lenght DWORDS+MOVZX
1561 cycles, COPYAtoB_XZE- 2062 bytes- copy lenght DWORDS+MOVZX
1571 cycles, COPYAtoB_XZZF-2062 bytes- copy lenght DWORDS+MOVZX
1574 cycles, COPYAtoB_YZZE-2062 bytes- copy length 16 BYTES+MOVZX
1577 cycles, COPYAtoB_WZZE-2062 bytes- copy 16 BYTES+MOVZX
1585 cycles, COPYAtoB_XZZC-2062 bytes- copy lenght DWORDS+MOVZX
1606 cycles, memcpy_4-     2062 bytes- copy memcpy * 8
1620 cycles, crt_memcpy-   2062 bytes- copy crt_memcpy
2019 cycles, COPYAtoB_SSEE-1027 bytes- copy 16 BYTES+MOVZX
2038 cycles, COPYAtoB_SSEK-1027 bytes- copy 16 BYTES+MOVZX
2049 cycles, COPYAtoB_SSEH-1027 bytes- copy 16 BYTES+MOVZX
2051 cycles, COPYAtoB_SSEJ-1027 bytes- copy 16 BYTES+MOVZX
2061 cycles, COPYAtoB_SSEI-1027 bytes- copy 16 BYTES+MOVZX
2110 cycles, memcpy_3-     1027 bytes- copy memcpyxmmU SSE
2116 cycles, memcpy_2-     1027 bytes- copy memcpy SSE
4029 cycles, COPYAtoB_SSEI-2062 bytes- copy 16 BYTES+MOVZX
4031 cycles, COPYAtoB_SSEJ-2062 bytes- copy 16 BYTES+MOVZX
4042 cycles, COPYAtoB_SSEE-2062 bytes- copy 16 BYTES+MOVZX
4046 cycles, COPYAtoB_SSEK-2062 bytes- copy 16 BYTES+MOVZX
4057 cycles, COPYAtoB_SSEH-2062 bytes- copy 16 BYTES+MOVZX
4085 cycles, memcpy_2-     2062 bytes- copy memcpy SSE
4389 cycles, memcpy_3-     2062 bytes- copy memcpyxmmU SSE
********** END III **********

Gunther

#129
July 10, 2014, 04:55:12 AM
Rui,

results are attached.

Gunther
You have to know the facts before you can distort them.

RuiLoureiro

#130
July 10, 2014, 08:19:48 PM
Hi Gunther,
            Thank you for your work  :t
            I am trying to understand the behaviour of
            each procedure.
            COPYAtoB_SSEJ are EQUAL to COPYAtoB_SSEK
            but 'J' has 1push+1pop and 'K' not (30 cycles more).
            I don't believe that 'J' is best than 'K'.
            Something seems to be wrong here.
---------------------------
Results from Gunther
---------------------------
Quote
CopyString54.txt
-------------------------------------------------------------
Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz (SSE4)
--------------------------------------------------------------
***** Time table *****

14 cycles, crt_memcpy-     15 bytes- copy crt_memcpy
17 cycles, COPYAtoB_SSEJ-  53 bytes- copy 16 BYTES+MOVZX           <<<<---J

23 cycles, COPYAtoB_SSEJ- 103 bytes- copy 16 BYTES+MOVZX           <<<<---J
25 cycles, memcpy_1-       53 bytes- copy regcopy
27 cycles, memcpy_1-       15 bytes- copy regcopy
27 cycles, COPYAtoB_SSEJ-  15 bytes- copy 16 BYTES+MOVZX
30 cycles, memcpy_4-       15 bytes- copy memcpy * 8
32 cycles, memcpy_2-       15 bytes- copy memcpy SSE
34 cycles, memcpy_3-       15 bytes- copy memcpyxmmU SSE
35 cycles, memcpy_4-       53 bytes- copy memcpy * 8
36 cycles, crt_memcpy-     53 bytes- copy crt_memcpy
38 cycles, memcpy_2-       53 bytes- copy memcpy SSE
39 cycles, memcpy_3-       53 bytes- copy memcpyxmmU SSE
39 cycles, memcpy_1-      103 bytes- copy regcopy

41 cycles, COPYAtoB_SSEJ- 203 bytes- copy 16 BYTES+MOVZX           <<<<---J
44 cycles, memcpy_2-      103 bytes- copy memcpy SSE
46 cycles, memcpy_4-      103 bytes- copy memcpy * 8
47 cycles, crt_memcpy-    103 bytes- copy crt_memcpy
48 cycles, memcpy_3-      103 bytes- copy memcpyxmmU SSE
50 cycles, memcpy_3-      203 bytes- copy memcpyxmmU SSE
60 cycles, memcpy_2-      203 bytes- copy memcpy SSE
67 cycles, memcpy_4-      203 bytes- copy memcpy * 8
68 cycles, memcpy_1-      203 bytes- copy regcopy
68 cycles, crt_memcpy-    203 bytes- copy crt_memcpy

69 cycles, memcpy_3-      503 bytes- copy memcpyxmmU SSE           >>>>>---
82 cycles, COPYAtoB_SSEJ- 503 bytes- copy 16 BYTES+MOVZX           <<<<---J
86 cycles, memcpy_4-      503 bytes- copy memcpy * 8
93 cycles, crt_memcpy-    503 bytes- copy crt_memcpy

102 cycles, memcpy_3-     1027 bytes- copy memcpyxmmU SSE           >>>>>---
122 cycles, memcpy_4-     1027 bytes- copy memcpy * 8
129 cycles, crt_memcpy-   1027 bytes- copy crt_memcpy
132 cycles, memcpy_2-      503 bytes- copy memcpy SSE
132 cycles, COPYAtoB_SSEJ-1027 bytes- copy 16 BYTES+MOVZX           <<<<---J
146 cycles, memcpy_1-      503 bytes- copy regcopy

184 cycles, memcpy_3-     2062 bytes- copy memcpyxmmU SSE           >>>>>---
187 cycles, memcpy_2-     1027 bytes- copy memcpy SSE
210 cycles, memcpy_4-     2062 bytes- copy memcpy * 8
215 cycles, crt_memcpy-   2062 bytes- copy crt_memcpy
262 cycles, COPYAtoB_SSEJ-2062 bytes- copy 16 BYTES+MOVZX           <<<<---J

286 cycles, memcpy_1-     1027 bytes- copy regcopy
297 cycles, memcpy_2-     2062 bytes- copy memcpy SSE
582 cycles, memcpy_1-     2062 bytes- copy regcopy
********** END III **********
Quote
CopyString55.txt
--------------------------------------------------------------
Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz (SSE4)
--------------------------------------------------------------
***** Time table *****

14 cycles, crt_memcpy-     15 bytes- copy crt_memcpy
17 cycles, COPYAtoB_SSEK-  53 bytes- copy 16 BYTES+MOVZX       <<<<<---- K
17 cycles, COPYAtoB_SSEJ-  53 bytes- copy 16 BYTES+MOVZX       <<<<<---- J

25 cycles, memcpy_1-       53 bytes- copy regcopy
26 cycles, COPYAtoB_SSEJ-  15 bytes- copy 16 BYTES+MOVZX       <<<<<---- J
29 cycles, memcpy_1-       15 bytes- copy regcopy
30 cycles, COPYAtoB_SSEK-  15 bytes- copy 16 BYTES+MOVZX       <<<<<---- K
31 cycles, memcpy_4-       15 bytes- copy memcpy * 8
33 cycles, memcpy_2-       15 bytes- copy memcpy SSE
35 cycles, memcpy_3-       15 bytes- copy memcpyxmmU SSE
36 cycles, memcpy_4-       53 bytes- copy memcpy * 8
37 cycles, COPYAtoB_SSEJ- 103 bytes- copy 16 BYTES+MOVZX       <<<<<---- J
37 cycles, COPYAtoB_SSEK- 103 bytes- copy 16 BYTES+MOVZX       <<<<<---- K

37 cycles, crt_memcpy-     53 bytes- copy crt_memcpy
38 cycles, memcpy_2-       53 bytes- copy memcpy SSE
39 cycles, memcpy_3-       53 bytes- copy memcpyxmmU SSE
41 cycles, memcpy_1-      103 bytes- copy regcopy
46 cycles, memcpy_4-      103 bytes- copy memcpy * 8
47 cycles, crt_memcpy-    103 bytes- copy crt_memcpy
48 cycles, COPYAtoB_SSEJ- 203 bytes- copy 16 BYTES+MOVZX       <<<<<---- J
51 cycles, COPYAtoB_SSEK- 203 bytes- copy 16 BYTES+MOVZX       <<<<<---- K

52 cycles, memcpy_2-      103 bytes- copy memcpy SSE
53 cycles, memcpy_3-      103 bytes- copy memcpyxmmU SSE
55 cycles, memcpy_3-      203 bytes- copy memcpyxmmU SSE
62 cycles, memcpy_2-      203 bytes- copy memcpy SSE
67 cycles, COPYAtoB_SSEJ- 503 bytes- copy 16 BYTES+MOVZX       <<<<<---- J

68 cycles, memcpy_4-      203 bytes- copy memcpy * 8
69 cycles, memcpy_1-      203 bytes- copy regcopy
70 cycles, crt_memcpy-    203 bytes- copy crt_memcpy
72 cycles, memcpy_3-      503 bytes- copy memcpyxmmU SSE       >>>>>>>--
88 cycles, memcpy_4-      503 bytes- copy memcpy * 8
94 cycles, COPYAtoB_SSEK- 503 bytes- copy 16 BYTES+MOVZX       <<<<<---- K
97 cycles, crt_memcpy-    503 bytes- copy crt_memcpy

119 cycles, memcpy_3-     1027 bytes- copy memcpyxmmU SSE       >>>>>>--
126 cycles, memcpy_2-      503 bytes- copy memcpy SSE
133 cycles, memcpy_4-     1027 bytes- copy memcpy * 8
133 cycles, COPYAtoB_SSEJ-1027 bytes- copy 16 BYTES+MOVZX       <<<<<---- J
138 cycles, crt_memcpy-   1027 bytes- copy crt_memcpy
149 cycles, memcpy_1-      503 bytes- copy regcopy
162 cycles, COPYAtoB_SSEK-1027 bytes- copy 16 BYTES+MOVZX       <<<<<---- K
194 cycles, memcpy_2-     1027 bytes- copy memcpy SSE

196 cycles, memcpy_3-     2062 bytes- copy memcpyxmmU SSE       >>>>>>--
219 cycles, crt_memcpy-   2062 bytes- copy crt_memcpy
220 cycles, memcpy_4-     2062 bytes- copy memcpy * 8
262 cycles, COPYAtoB_SSEJ-2062 bytes- copy 16 BYTES+MOVZX       <<<<<---- J
289 cycles, memcpy_1-     1027 bytes- copy regcopy
296 cycles, COPYAtoB_SSEK-2062 bytes- copy 16 BYTES+MOVZX       <<<<<---- K

306 cycles, memcpy_2-     2062 bytes- copy memcpy SSE
597 cycles, memcpy_1-     2062 bytes- copy regcopy
********** END III **********

RuiLoureiro

#131
July 10, 2014, 09:30:03 PM
Hi Gunther,

            Now, i want to test SSEL in your i7.         
            Could you run CopyString56, please ?
            Thanks.
       
These are my results.
Quote
-----------------------------------------------------
Intel(R) Pentium(R) 4 CPU 3.40GHz (SSE3)
-----------------------------------------------------
...   
  59 cycles, COPYAtoB_SSEL,   15 bytes - copy 16 BYTES+MOVZX
  64 cycles, COPYAtoB_SSEL,   53 bytes - copy 16 BYTES+MOVZX
199 cycles, COPYAtoB_SSEL,  103 bytes - copy 16 BYTES+MOVZX
445 cycles, COPYAtoB_SSEL,  203 bytes - copy 16 BYTES+MOVZX
959 cycles, COPYAtoB_SSEL,  503 bytes - copy 16 BYTES+MOVZX
2026 cycles, COPYAtoB_SSEL, 1027 bytes - copy 16 BYTES+MOVZX
4025 cycles, COPYAtoB_SSEL, 2062 bytes - copy 16 BYTES+MOVZX

  58 cycles, COPYAtoB_SSEJ,   15 bytes - copy 16 BYTES+MOVZX
  91 cycles, COPYAtoB_SSEJ,   53 bytes - copy 16 BYTES+MOVZX
204 cycles, COPYAtoB_SSEJ,  103 bytes - copy 16 BYTES+MOVZX
430 cycles, COPYAtoB_SSEJ,  203 bytes - copy 16 BYTES+MOVZX
974 cycles, COPYAtoB_SSEJ,  503 bytes - copy 16 BYTES+MOVZX
2046 cycles, COPYAtoB_SSEJ, 1027 bytes - copy 16 BYTES+MOVZX
4025 cycles, COPYAtoB_SSEJ, 2062 bytes - copy 16 BYTES+MOVZX

  50 cycles, COPYAtoB_SSEK,   15 bytes - copy 16 BYTES+MOVZX
  88 cycles, COPYAtoB_SSEK,   53 bytes - copy 16 BYTES+MOVZX
189 cycles, COPYAtoB_SSEK,  103 bytes - copy 16 BYTES+MOVZX
429 cycles, COPYAtoB_SSEK,  203 bytes - copy 16 BYTES+MOVZX
962 cycles, COPYAtoB_SSEK,  503 bytes - copy 16 BYTES+MOVZX
2035 cycles, COPYAtoB_SSEK, 1027 bytes - copy 16 BYTES+MOVZX
4044 cycles, COPYAtoB_SSEK, 2062 bytes - copy 16 BYTES+MOVZX

Gunther

#132
July 11, 2014, 12:06:29 AM
Rui,

results are attached in c56.zip. I think it's time now to sum up and provide the source code for the best procedures.

Gunther
You have to know the facts before you can distort them.

RuiLoureiro

#133
July 11, 2014, 01:28:19 AM
Quote from: Gunther on July 11, 2014, 12:06:29 AM
Rui,
...
I think it's time now to sum up and provide the source code for the best procedures.
Gunther
Hi Gunther,
            You are doing a very good work.
            Thank you so much for this work.  :t

I am trying to get the best procedure that uses SSE (in your i7)
without alignement. I think i still need to adjust some code.
But if someone wants to know some proc i will give it to him.
If you want i will send to you.

Only now, we may compare COPYAtoB_SSEJ with COPYAtoB_SSEK
and we may see K is better than J, as i said before.
But it is not good for 15 bytes.

---------------------------------
Some results from Gunther
----------------------------------
Quote
--------------------------------------------------------------
Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz (SSE4)
--------------------------------------------------------------
13 cycles, crt_memcpy,   15 bytes - copy crt_memcpy
35 cycles, crt_memcpy,   53 bytes - copy crt_memcpy


46 cycles, crt_memcpy,  103 bytes - copy crt_memcpy
72 cycles, crt_memcpy,  203 bytes - copy crt_memcpy
95 cycles, crt_memcpy,  503 bytes - copy crt_memcpy
138 cycles, crt_memcpy, 1027 bytes - copy crt_memcpy
219 cycles, crt_memcpy, 2062 bytes - copy crt_memcpy

40 cycles, memcpy_1,   15 bytes - copy memcpy_1
24 cycles, memcpy_1,   53 bytes - copy memcpy_1
39 cycles, memcpy_1,  103 bytes - copy memcpy_1
69 cycles, memcpy_1,  203 bytes - copy memcpy_1
147 cycles, memcpy_1,  503 bytes - copy memcpy_1
294 cycles, memcpy_1, 1027 bytes - copy memcpy_1
606 cycles, memcpy_1, 2062 bytes - copy memcpy_1

33 cycles, memcpy_2,   15 bytes - copy memcpy_2
38 cycles, memcpy_2,   53 bytes - copy memcpy_2
54 cycles, memcpy_2,  103 bytes - copy memcpy_2
62 cycles, memcpy_2,  203 bytes - copy memcpy_2
125 cycles, memcpy_2,  503 bytes - copy memcpy_2
189 cycles, memcpy_2, 1027 bytes - copy memcpy_2
304 cycles, memcpy_2, 2062 bytes - copy memcpy_2
-----------------------------------------------------------
43 cycles, COPYAtoB_SSEL,   15 bytes - copy 16 BYTES+MOVZX
16 cycles, COPYAtoB_SSEL,   53 bytes - copy 16 BYTES+MOVZX

36 cycles, COPYAtoB_SSEL,  103 bytes - copy 16 BYTES+MOVZX
51 cycles, COPYAtoB_SSEL,  203 bytes - copy 16 BYTES+MOVZX

109 cycles, COPYAtoB_SSEL,  503 bytes - copy 16 BYTES+MOVZX
165 cycles, COPYAtoB_SSEL, 1027 bytes - copy 16 BYTES+MOVZX
301 cycles, COPYAtoB_SSEL, 2062 bytes - copy 16 BYTES+MOVZX
-----------------------------------------------------------
34 cycles, memcpy_3,   15 bytes - copy memcpy_3
39 cycles, memcpy_3,   53 bytes - copy memcpy_3

52 cycles, memcpy_3,  103 bytes - copy memcpy_3
54 cycles, memcpy_3,  203 bytes - copy memcpy_3

71 cycles, memcpy_3,  503 bytes - copy memcpy_3
117 cycles, memcpy_3, 1027 bytes - copy memcpy_3
199 cycles, memcpy_3, 2062 bytes - copy memcpy_3
----------------------------------------------------------
32 cycles, memcpy_4,   15 bytes - copy memcpy_4
35 cycles, memcpy_4,   53 bytes - copy memcpy_4
46 cycles, memcpy_4,  103 bytes - copy memcpy_4
69 cycles, memcpy_4,  203 bytes - copy memcpy_4
87 cycles, memcpy_4,  503 bytes - copy memcpy_4
130 cycles, memcpy_4, 1027 bytes - copy memcpy_4
214 cycles, memcpy_4, 2062 bytes - copy memcpy_4

16 cycles, COPYAtoB_SSEE,   15 bytes - copy 16 BYTES+MOVZX
18 cycles, COPYAtoB_SSEE,   53 bytes - copy 16 BYTES+MOVZX
40 cycles, COPYAtoB_SSEE,  103 bytes - copy 16 BYTES+MOVZX
53 cycles, COPYAtoB_SSEE,  203 bytes - copy 16 BYTES+MOVZX
110 cycles, COPYAtoB_SSEE,  503 bytes - copy 16 BYTES+MOVZX
164 cycles, COPYAtoB_SSEE, 1027 bytes - copy 16 BYTES+MOVZX
300 cycles, COPYAtoB_SSEE, 2062 bytes - copy 16 BYTES+MOVZX

48 cycles, COPYAtoB_SSEH,   15 bytes - copy 16 BYTES+MOVZX
18 cycles, COPYAtoB_SSEH,   53 bytes - copy 16 BYTES+MOVZX
41 cycles, COPYAtoB_SSEH,  103 bytes - copy 16 BYTES+MOVZX
55 cycles, COPYAtoB_SSEH,  203 bytes - copy 16 BYTES+MOVZX
72 cycles, COPYAtoB_SSEH,  503 bytes - copy 16 BYTES+MOVZX
135 cycles, COPYAtoB_SSEH, 1027 bytes - copy 16 BYTES+MOVZX
282 cycles, COPYAtoB_SSEH, 2062 bytes - copy 16 BYTES+MOVZX

27 cycles, COPYAtoB_SSEI,   15 bytes - copy 16 BYTES+MOVZX
15 cycles, COPYAtoB_SSEI,   53 bytes - copy 16 BYTES+MOVZX
39 cycles, COPYAtoB_SSEI,  103 bytes - copy 16 BYTES+MOVZX
49 cycles, COPYAtoB_SSEI,  203 bytes - copy 16 BYTES+MOVZX
68 cycles, COPYAtoB_SSEI,  503 bytes - copy 16 BYTES+MOVZX
134 cycles, COPYAtoB_SSEI, 1027 bytes - copy 16 BYTES+MOVZX
268 cycles, COPYAtoB_SSEI, 2062 bytes - copy 16 BYTES+MOVZX

41 cycles, COPYAtoB_SSEJ,   15 bytes - copy 16 BYTES+MOVZX
20 cycles, COPYAtoB_SSEJ,   53 bytes - copy 16 BYTES+MOVZX
39 cycles, COPYAtoB_SSEJ,  103 bytes - copy 16 BYTES+MOVZX
59 cycles, COPYAtoB_SSEJ,  203 bytes - copy 16 BYTES+MOVZX
115 cycles, COPYAtoB_SSEJ,  503 bytes - copy 16 BYTES+MOVZX
163 cycles, COPYAtoB_SSEJ, 1027 bytes - copy 16 BYTES+MOVZX
309 cycles, COPYAtoB_SSEJ, 2062 bytes - copy 16 BYTES+MOVZX

41 cycles, COPYAtoB_SSEK,   15 bytes - copy 16 BYTES+MOVZX
19 cycles, COPYAtoB_SSEK,   53 bytes - copy 16 BYTES+MOVZX
38 cycles, COPYAtoB_SSEK,  103 bytes - copy 16 BYTES+MOVZX
52 cycles, COPYAtoB_SSEK,  203 bytes - copy 16 BYTES+MOVZX
87 cycles, COPYAtoB_SSEK,  503 bytes - copy 16 BYTES+MOVZX
139 cycles, COPYAtoB_SSEK, 1027 bytes - copy 16 BYTES+MOVZX
277 cycles, COPYAtoB_SSEK, 2062 bytes - copy 16 BYTES+MOVZX

RuiLoureiro

#134
July 11, 2014, 02:28:57 AM Last Edit: July 11, 2014, 04:09:21 AM by RuiLoureiro
Hi Gunther,

        I would like to see the results from CopyString57
        Could you run it, please ?
        Thanks
(unfortunately i have not an i7 like yours, till now, to do it)       

Here are some results.
Quote
memcpy_3 doesn't preserve ESI, EDI

-----------------------------------------------------
Intel(R) Pentium(R) 4 CPU 3.40GHz (SSE3)
-----------------------------------------------------
  35 cycles, crt_memcpy,   15 bytes - copy crt_memcpy
108 cycles, crt_memcpy,   53 bytes - copy crt_memcpy
139 cycles, crt_memcpy,  103 bytes - copy crt_memcpy
226 cycles, crt_memcpy,  203 bytes - copy crt_memcpy
446 cycles, crt_memcpy,  503 bytes - copy crt_memcpy
836 cycles, crt_memcpy, 1027 bytes - copy crt_memcpy
1670 cycles, crt_memcpy, 2062 bytes - copy crt_memcpy

  82 cycles, memcpy_1,   15 bytes - copy memcpy_1
  68 cycles, memcpy_1,   53 bytes - copy memcpy_1
106 cycles, memcpy_1,  103 bytes - copy memcpy_1
179 cycles, memcpy_1,  203 bytes - copy memcpy_1
400 cycles, memcpy_1,  503 bytes - copy memcpy_1
726 cycles, memcpy_1, 1027 bytes - copy memcpy_1
1596 cycles, memcpy_1, 2062 bytes - copy memcpy_1

251 cycles, memcpy_2,   15 bytes - copy memcpy_2
278 cycles, memcpy_2,   53 bytes - copy memcpy_2
383 cycles, memcpy_2,  103 bytes - copy memcpy_2
623 cycles, memcpy_2,  203 bytes - copy memcpy_2
1141 cycles, memcpy_2,  503 bytes - copy memcpy_2
2115 cycles, memcpy_2, 1027 bytes - copy memcpy_2
4113 cycles, memcpy_2, 2062 bytes - copy memcpy_2

  61 cycles, COPYAtoB_SSEL, 15 bytes - copy 16 BYTES+MOVZX
  73 cycles, COPYAtoB_SSEL, 53 bytes - copy 16 BYTES+MOVZX
198 cycles, COPYAtoB_SSEL, 103 bytes - copy 16 BYTES+MOVZX
441 cycles, COPYAtoB_SSEL, 203 bytes - copy 16 BYTES+MOVZX
957 cycles, COPYAtoB_SSEL, 503 bytes - copy 16 BYTES+MOVZX
2028 cycles, COPYAtoB_SSEL, 1027 bytes - copy 16 BYTES+MOVZX
4037 cycles, COPYAtoB_SSEL, 2062 bytes - copy 16 BYTES+MOVZX

256 cycles, memcpy_3, 15 bytes - copy memcpy_3
279 cycles, memcpy_3, 53 bytes - copy memcpy_3
380 cycles, memcpy_3, 103 bytes - copy memcpy_3
634 cycles, memcpy_3, 203 bytes - copy memcpy_3
1154 cycles, memcpy_3, 503 bytes - copy memcpy_3
2088 cycles, memcpy_3, 1027 bytes - copy memcpy_3
4364 cycles, memcpy_3, 2062 bytes - copy memcpy_3

  72 cycles, memcpy_4,   15 bytes - copy memcpy_4
102 cycles, memcpy_4,   53 bytes - copy memcpy_4
132 cycles, memcpy_4,  103 bytes - copy memcpy_4
213 cycles, memcpy_4,  203 bytes - copy memcpy_4
431 cycles, memcpy_4,  503 bytes - copy memcpy_4
828 cycles, memcpy_4, 1027 bytes - copy memcpy_4
1628 cycles, memcpy_4, 2062 bytes - copy memcpy_4

  50 cycles, COPYAtoB_SSEJ,   15 bytes - copy 16 BYTES+MOVZX
  90 cycles, COPYAtoB_SSEJ,   53 bytes - copy 16 BYTES+MOVZX
209 cycles, COPYAtoB_SSEJ,  103 bytes - copy 16 BYTES+MOVZX
432 cycles, COPYAtoB_SSEJ,  203 bytes - copy 16 BYTES+MOVZX
963 cycles, COPYAtoB_SSEJ,  503 bytes - copy 16 BYTES+MOVZX
2048 cycles, COPYAtoB_SSEJ, 1027 bytes - copy 16 BYTES+MOVZX
4021 cycles, COPYAtoB_SSEJ, 2062 bytes - copy 16 BYTES+MOVZX

  47 cycles, COPYAtoB_SSEK,   15 bytes - copy 16 BYTES+MOVZX
  85 cycles, COPYAtoB_SSEK,   53 bytes - copy 16 BYTES+MOVZX
186 cycles, COPYAtoB_SSEK,  103 bytes - copy 16 BYTES+MOVZX
431 cycles, COPYAtoB_SSEK,  203 bytes - copy 16 BYTES+MOVZX
961 cycles, COPYAtoB_SSEK,  503 bytes - copy 16 BYTES+MOVZX
2038 cycles, COPYAtoB_SSEK, 1027 bytes - copy 16 BYTES+MOVZX
4067 cycles, COPYAtoB_SSEK, 2062 bytes - copy 16 BYTES+MOVZX

  28 cycles, COPYAtoB_SSEM,   15 bytes - copy 16 BYTES+MOVZX
  61 cycles, COPYAtoB_SSEM,   53 bytes - copy 16 BYTES+MOVZX
177 cycles, COPYAtoB_SSEM,  103 bytes - copy 16 BYTES+MOVZX
416 cycles, COPYAtoB_SSEM,  203 bytes - copy 16 BYTES+MOVZX
962 cycles, COPYAtoB_SSEM,  503 bytes - copy 16 BYTES+MOVZX
2056 cycles, COPYAtoB_SSEM, 1027 bytes - copy 16 BYTES+MOVZX
4019 cycles, COPYAtoB_SSEM, 2062 bytes - copy 16 BYTES+MOVZX

  36 cycles, COPYAtoB_SSEN,   15 bytes - copy 16 BYTES+MOVZX
  66 cycles, COPYAtoB_SSEN,   53 bytes - copy 16 BYTES+MOVZX
158 cycles, COPYAtoB_SSEN,  103 bytes - copy 16 BYTES+MOVZX
421 cycles, COPYAtoB_SSEN,  203 bytes - copy 16 BYTES+MOVZX
950 cycles, COPYAtoB_SSEN,  503 bytes - copy 16 BYTES+MOVZX
2043 cycles, COPYAtoB_SSEN, 1027 bytes - copy 16 BYTES+MOVZX
4032 cycles, COPYAtoB_SSEN, 2062 bytes - copy 16 BYTES+MOVZX

  34 cycles, COPYAtoB_SSEA,   15 bytes - copy 16 BYTES+MOVZX
  97 cycles, COPYAtoB_SSEA,   53 bytes - copy 16 BYTES+MOVZX
212 cycles, COPYAtoB_SSEA,  103 bytes - copy 16 BYTES+MOVZX
446 cycles, COPYAtoB_SSEA,  203 bytes - copy 16 BYTES+MOVZX
988 cycles, COPYAtoB_SSEA,  503 bytes - copy 16 BYTES+MOVZX
2065 cycles, COPYAtoB_SSEA, 1027 bytes - copy 16 BYTES+MOVZX
4093 cycles, COPYAtoB_SSEA, 2062 bytes - copy 16 BYTES+MOVZX
Print
Go Up Pages 1 ... 7 8 9 10 11
User actions