Apple Microarchitecture Research by Dougall Johnson

M1/A14 P-core (Firestorm): Overview | Base Instructions | SIMD and FP Instructions
M1/A14 E-core (Icestorm):  Overview | Base Instructions | SIMD and FP Instructions

SUB (register, lsl, 64-bit)

Test 1: uops

Code:

  sub x0, x0, x1, lsl #17
  mov x0, 1
  mov x1, 2

(no loop instructions)

1000 unrolls and 1 iteration

Retires: 1.000

Issues: 2.000

Integer unit issues: 2.001

Load/store unit issues: 0.000

SIMD/FP unit issues: 0.000

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
100420302001200110005226510001000200020011000
100420302001200110005226510001000200020011000
100420302001200110005226510001000200020011000
100420302001200110005226510001000200020011000
100420302001200110005226510001000200020011000
100420302001200110005226510001000200020011000
100420302001200110005226510001000200020011000
100420302001200110005226510001000200020011000
100420302001200110005226510001000200020011000
100420302001200110005226510001000200020011000

Test 2: Latency 1->2

Code:

  sub x0, x0, x1, lsl #17
  mov x0, 1
  mov x1, 2

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 2.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule ldst uop (55)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
102042003020101201010101045291431010410210202242000110100
102042003020101201010101045291431010410210202242000110100
102042003020101201010101045291861010410212202242000110100
102042003020101201010101045291861010410212202242000110100
102042003020101201010101045291861010410212202242000110100
102042003020101201010101045291861010410212202242000110100
102042003020101201010101045291861010410212202242000110100
102042003020101201010101045291861010410212202242000110100
102042003020101201010101045291861010410212202242000110100
102042003020101201010101045291861010410212202242000110100

1000 unrolls and 10 iterations

Result (median cycles for code): 2.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
10024200302002120021100255291971002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205296141005810072200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010

Test 3: Latency 1->3

Code:

  sub x0, x1, x0, lsl #17
  mov x0, 1
  mov x1, 2

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 2.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)map ldst uop inputs (80)map simd uop inputs (81)? int output thing (e9)? ldst retires (ed)? simd retires (ee)? int retires (ef)
102042003020101201011010452910010104102102022000200010010100
102042003020101201011010452918610104102122022400200010010100
102042003020101201011010452918610104102122022400200010010100
102042003020101201011010452918610104102122022400200010010100
102042003020101201011010452918610104102122022400200010010100
102042003020101201011010452918610104102122022400200010010100
102042003020101201011010452918610104102122022400200010010100
102042003020101201011010452918610104102122022400200010010100
102042003020101201011010452918610104102122022400200010010100
102042003020101201011010452918610104102122022400200010010100

1000 unrolls and 10 iterations

Result (median cycles for code): 2.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
10024200302002120021100255292761002510032200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010
10024200302002120021100205292531002010020200202001110010

Test 4: throughput

Count: 8

Code:

  sub x0, x8, x9, lsl #17
  sub x1, x8, x9, lsl #17
  sub x2, x8, x9, lsl #17
  sub x3, x8, x9, lsl #17
  sub x4, x8, x9, lsl #17
  sub x5, x8, x9, lsl #17
  sub x6, x8, x9, lsl #17
  sub x7, x8, x9, lsl #17
  mov x8, 9
  mov x9, 10

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.6675

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)dispatch ldst uop (58)int uops in schedulers (59)simd uops in schedulers (5a)dispatch uop (78)map int uop (7c)map ldst uop (7d)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
8020453404160117160117801300136029008012980234016026816001680100
8020453415160116160116801290136083808013080236016027216001780100
8020453404160117160117801300136083808013080236016027216001780100
8020453404160117160117801300136083808013080236016027216001780100
8020453404160117160117801300136083808013080236016027216001780100
8020453404160117160117801300136083808013080236016027216001780100
8020453404160117160117801300136083808013080236016027216001780100
8020453404160117160117801300136083808013080236016027216001780100
8020453404160117160117801300136083808013080236016027216001780100
8020453404160117160117801300136079108013080237016027216001780100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.6671

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)schedule ldst uop (55)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
8002453390160039160039008005101359975800518005816009616002980010
8002453371160021160021008002001359903800208002016002016001180010
8002453371160021160021008002001359903800208002016002016001180010
8002453371160021160021008002001359903800208002016002016001180010
8002453371160021160021008002001359903800208002016002016001180010
8002453371160021160021008002001359903800208002016002016001180010
8002453371160021160021008002001359903800208002016002016001180010
8002453371160021160021008002001359903800208002016002016001180010
8002453371160021160021008002001359903800208002016019616007480010
848217355216381616234554141782132411359903800208002016002016001180010