SiFive U74

SiFive Freedom U740 (U74 core), 1200 MHz, (28 nm). 4 cores. 16 GB (DDR4). SiFive HiFive Unmatched.

L1 Data cache = 32 KB, 64 B/line, 8-WAY, VIPT. write-back. read/write-allocate with a random replacement.
L1 Instruction cache = 32 KB, 64 B/line, 4-WAY, VIPT. read/write-allocate with a random replacement. policy.
L2 Cache = 2 MB, 64 B/line, 16-WAY, 4 banks, shared by all cores.

16-entry branch target buffer (BTB)
3.6 KiB branch history table (BHT)
8-entry indirect-jump target predictor (IJTP)
6-entry return-address stack (RAS)
Pipeline: F1, F2, D1, D2, AG, M1, M2, WB
Dual Issue

39-bit virtual address space using 38-bits of physical address space. Supported page sizes include 4 KiB, 2 MiB, and 1 GiB megapages. The default Linux page size (PAGESIZE) is 4 KiB.

L1 Data Cache Latency = 3 cycles for simple access via pointer
L1 Data Cache Latency = 5 cycles for access with complex address calculation (size_t n, *p; n = p[n]).
L2 Cache Latency = 26 cycles
RAM Latency = 26 cycles + 145 ns

2 MB pages (64-bit Linux)

Data TLB L1: 5 ? items (cover 10 MiB). full-assoc. Miss penalty = 11 cycles. Parallel miss: 12 cycles per L1 data cache read access.
TLB L2 (2 MB pages): 512 items. direct-map. Miss penalty = ? cycles.

  Size        Latency        Increase   Description

  32 K     3                            
  64 K    15                       12   + 23 (L2)
 128 K    21                        6   
 256 K    24                        3
 512 K    25                        1   
   1 M    25                        0    
   2 M    26 +   2 ns        1 + 2 ns
   4 M    26 +  77 ns          +75 ns   + 145 ns (RAM)
   8 M    26 + 112 ns          +35 ns 
  16 M    32 + 129 ns        6 +17 ns   + 11 (L1 TLB miss)
  32 M    35 + 137 ns        3 + 8 ns   
  64 M    36 + 141 ns        1 + 4 ns
 128 M    37 + 143 ns        1 + 2 ns
 256 M    37 + 144 ns            1 ns    
 512 M    38 + 145 ns        1 + 1 ns    
1024 M    40 + 145 ns        2    
2048 M    44 + 145 ns        4          + ? (L2 TLB miss)
4096 M    48 + 145 ns        4           
8192 M    53 + 145 ns        7

4 KB pages (64-bit Linux)

Data TLB L1 (4 KB pages): 40 items. full-assoc. Miss penalty = 11 cycles. Parallel miss: 12 cycles per L1 data cache read access.
TLB L2: 512 items. direct-map. Miss penalty = 24 cycles (translation entry from L1/L2 cache). Parallel miss: 22 cycles per L2 cache read access.

  Size        Latency        Increase   Description

  32 K     3                            
  64 K    15                       12   + 23 (L2)
 128 K    21                        6   
 256 K    30                        9   + 11 (L1 TLB miss)
 512 K    35                        5   
   1 M    38                        3

MISC

Branch misprediction penalty = 4 cycles

8-bytes range cross penalty = 708 cycles

L1 B/W (Parallel Random Read) = 1 cycle per one access

L2->L1 B/W (Parallel Random Read) = 24 cycles per access
L2->L1 B/W (Read, 64 bytes step) = 24 cycles per cache line (no HW prefetch L2->L1)
L2->L1 B/W (Read, 64 bytes step - pointer chasing) = 26 cycles per cache line (no HW prefetch L2->L1)
L2 Write B/W (skip, 64 Bytes step) = 24 cycles per write

RAM Read B/W (Parallel Random Read) = 161 ns per access
RAM Read B/W (Read, 8-64 Bytes step) = 430-464 MB/s (no HW prefetch)
RAM Read B/W (Read, 64 Bytes step - pointer chasing) = 461 MB/s (no HW prefetch)
RAM Write B/W (skip, 8 - 64 Bytes step) = 290 MB/s

7-Zip Benchmark

### Linux 5.15
## THP on
# gcc-11 -mtune=sifive-7-series -O2

7-Zip (z) 21.07 (LE) : Copyright (c) 1999-2021 Igor Pavlov : 2021-12-26
 64-bit locale=C.UTF-8 Threads:4

 mt*
Compiler: 11.2.0 GCC 11.2.0
Linux : 5.15.0-1014-generic : #16-Ubuntu SMP Thu Jun 2 13:45:22 UTC 2022 : riscv64
PageSize:4KB THP:madvise hwcap:112D
LE 

1T CPU Freq (MHz):  1180  1183  1183  1183  1183  1183  1183
2T CPU Freq (MHz): 198% 1164   199% 1178  

RAM size:   15990 MB,  # CPU hardware threads:   4
RAM usage:    437 MB,  # Benchmark threads:      1

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:        806   100    785    784  |      13237   100   1131   1130
23:        757   100    772    772  |      13009   100   1126   1126
24:        735   100    792    791  |      12731   100   1118   1118
25:        739   100    845    844  |      12452   100   1109   1108
----------------------------------  | ------------------------------
Avr:       759   100    798    798  |      12857   100   1121   1121
Tot:             100    960    959


RAM usage:    444 MB,  # Benchmark threads:      2

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:       1323   175    738   1288  |      25916   200   1107   2213
23:       1321   184    733   1346  |      25490   200   1104   2206
24:       1313   187    757   1413  |      24961   200   1096   2191
25:       1311   187    799   1498  |      24373   200   1087   2169
----------------------------------  | ------------------------------
Avr:      1317   183    757   1386  |      25185   200   1099   2195
Tot:             191    928   1790


RAM usage:    889 MB,  # Benchmark threads:      4

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:       2338   327    695   2275  |      49533   392   1079   4226
23:       2316   343    688   2361  |      48328   390   1073   4182
24:       2111   318    713   2270  |      46532   383   1066   4084
25:       2212   343    736   2526  |      45735   385   1058   4070
----------------------------------  | ------------------------------
Avr:      2244   333    708   2358  |      47532   387   1069   4140
Tot:             360    888   3249



7-Zip (z) 21.07 (LE) : Copyright (c) 1999-2021 Igor Pavlov : 2021-12-26
 64-bit locale=C.UTF-8 Threads:4

 mt=* m=* freq=1196
freq= 1196
Compiler: 11.2.0 GCC 11.2.0
Linux : 5.15.0-1014-generic : #16-Ubuntu SMP Thu Jun 2 13:45:22 UTC 2022 : riscv64
PageSize:4KB THP:madvise hwcap:112D
LE

1T CPU Freq (MHz):  1182  1181  1183  1170  1175  1183  1177
2T CPU Freq (MHz): 199% 1179   198% 1171

RAM size:   15990 MB,  # CPU hardware threads:   4
RAM usage:    227 MB,  # Benchmark threads:      1


Method           Speed Usage    R/U Rating   E/U Effec
                 KiB/s     %   MIPS   MIPS     %     %

CPU                      100   1183   1182    99    99
CPU                      100   1183   1183    99    99
CPU                      100   1183   1183    99    99

LZMA:x1           2281   100    842    841    70    70
                 14389   100   1152   1151    96    96
LZMA:x3           1331   100    818    818    68    68
                 12902   100   1083   1083    91    91
LZMA:x5:mt1        747   100    935    934    78    78
                 12635   100   1066   1066    89    89
LZMA:x5:mt2       1297   183    886   1620    74   135
                 12640   100   1066   1066    89    89
Deflate:x1        6583   100    836    836    70    70
                 28677   100    892    891    75    75
Deflate:x5        2452   100    944    944    79    79
                 28530   100    889    886    74    74
Deflate:x7         827   100    918    917    77    77
                 29103   100    903    903    76    76
Deflate64:x5      2261   100    978    977    82    82
                 29214   100    914    914    76    76
BZip2:x1          1188   100    718    718    60    60
                  8220   100    892    891    75    75
BZip2:x5           795   100    664    664    56    56
                  5373   100   1055   1055    88    88
BZip2:x5:mt2      1362   194    586   1137    49    95
                  8618   165   1024   1692    86   141
BZip2:x7           274   100    713    712    60    60
                  5398   100   1059   1059    89    89
PPMD:x1            843   100    873    873    73    73
                   743   100    876    875    73    73
PPMD:x5            531   100    902    901    75    75
                   473   100    887    887    74    74
Delta:4         197155   100    606    606    51    51
                218808   100    897    896    75    75
BCJ             255671   100   1048   1047    88    88
                340917   100   1396   1396   117   117
AES256CBC:1      19328   100    475    475    40    40
                 19752   100    485    485    41    41
AES256CBC:2

AES256CBC:3

CRC32:1         111956   100    815    815    68    68
CRC32:4         262796   100    587    587    49    49
CRC32:8         260876   100    354    354    30    30
CRC32:32
CRC32:64
CRC64           234931   100    481    481    40    40
SHA256:1         33054   100    674    674    56    56
SHA256:2
SHA1:1           64595   100    605    605    51    51
SHA1:2
BLAKE2sp         32577   100    717    717    60    60

CPU                      100   1183   1183    99    99
------------------------------------------------------
Tot:                     111    904   1002    75    84




RAM usage:    911 MB,  # Benchmark threads:      4


Method           Speed Usage    R/U Rating   E/U Effec
                 KiB/s     %   MIPS   MIPS     %     %

CPU                      390   1183   4611    99   386
CPU                      391   1183   4622    99   386
CPU                      392   1183   4643    99   388

LZMA:x1           5867   383    565   2163    47   181
                 49349   359   1099   3946    92   330
LZMA:x3           4216   383    676   2591    57   217
                 48157   390   1036   4040    87   338
LZMA:x5:mt1       2534   389    815   3167    68   265
                 47462   392   1022   4001    85   335
LZMA:x5:mt2       2445   378    807   3055    68   255
                 47318   390   1022   3989    85   334
Deflate:x1       19476   386    640   2473    54   207
                111522   377    918   3465    77   290
Deflate:x5        8126   389    805   3129    67   262
                108511   364    926   3369    77   282
Deflate:x7        2932   389    835   3249    70   272
                111035   367    939   3446    79   288
Deflate64:x5      6870   387    767   2969    64   248
                111914   377    928   3499    78   293
BZip2:x1          3461   351    596   2091    50   175
                 27306   390    759   2960    63   247
BZip2:x5          2484   388    535   2074    45   173
                 17004   389    858   3337    72   279
BZip2:x5:mt2      2378   390    509   1985    43   166
                 16754   382    861   3288    72   275
BZip2:x7           909   390    605   2357    51   197
                 17288   388    874   3390    73   283
PPMD:x1           3081   384    831   3187    69   266
                  2701   388    820   3182    69   266
PPMD:x5           1632   375    739   2767    62   231
                  1532   388    739   2871    62   240
Delta:4         780471   393    610   2398    51   200
                861065   393    897   3527    75   295
BCJ            1003222   394   1044   4109    87   344
               1341875   393   1397   5496   117   460
AES256CBC:1      74899   390    473   1841    40   154
                 77542   393    485   1906    41   159
AES256CBC:2

AES256CBC:3

CRC32:1         438927   392    815   3195    68   267
CRC32:4        1025481   390    586   2289    49   191
CRC32:8        1010119   387    354   1370    30   115
CRC32:32
CRC32:64
CRC64           902610   385    481   1849    40   155
SHA256:1        130948   392    681   2671    57   223
SHA256:2
SHA1:1          252876   391    605   2367    51   198
SHA1:2
BLAKE2sp        127514   391    718   2805    60   235

CPU                      393   1183   4651    99   389
------------------------------------------------------
Tot:                     385    823   3169    69   265

Links

SiFive U74 at sifive.com

SiFive U74 (7 Series) at WikiChip