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

PMULL + EOR (8H)

Test 1: uops

Code:

  pmull v0.8h, v1.8b, v2.8b
  eor v0.16b, v0.16b, v3.16b
  movi v0.16b, 1
  movi v1.16b, 2
  movi v2.16b, 3

(no loop instructions)

1000 unrolls and 1 iteration

Retires: 2.000

Issues: 1.000

Integer unit issues: 0.001

Load/store unit issues: 0.000

SIMD/FP unit issues: 1.000

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch simd uop (57)ldst uops in schedulers (5b)dispatch uop (78)map simd uop (7e)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)
200410201001110001000400010002000400012000
20045531001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045621001110001000400010002000400012000
20045351001110001000400010002000400012000

Test 2: Latency 1->2

Chain cycles: 2

Code:

  pmull v0.8h, v1.8b, v2.8b
  eor v0.16b, v0.16b, v3.16b
  eor v1.16b, v1.16b, v0.16b
  nop
  movi v0.16b, 1
  movi v1.16b, 2
  movi v2.16b, 3

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code, minus 2 chain cycles): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
40204500882011110120010100200343001279085201002003000620060012130000100
40204500882011110120010100200343001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279625201342003004720060012130000100
40204500332010110120000100200003001279597201342003005020060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620260100230000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204501992013110120030100201023001279085201002003000620060012130000100

1000 unrolls and 10 iterations

Result (median cycles for code, minus 2 chain cycles): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
40024500332001111200001020000301279085200102030008206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010

Test 3: Latency 1->3

Chain cycles: 2

Code:

  pmull v0.8h, v1.8b, v2.8b
  eor v0.16b, v0.16b, v3.16b
  eor v2.16b, v2.16b, v0.16b
  nop
  movi v0.16b, 1
  movi v1.16b, 2
  movi v2.16b, 3

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code, minus 2 chain cycles): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)dispatch ldst uop (58)int uops in schedulers (59)simd uops in schedulers (5a)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map ldst uop (7d)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
402055006620108101200071002003258724552974812798922176815406813010020060012130000100
40204500332010110120000100200000300012790852010020003000620060012130000100
40204500332010110120000100200000300012790852010020003000620060012130000100
40204500332010110120000100200000300012790852010020003000620060012130000100
40204500332010110120000100200000300012790852010020003000620060012130000100
40204500332010110120000100200000300012790852010020003000620060012130000100
40204500332010110120000100200000300012790852010020003000620060012130000100
40204500332010110120000100200000300012790852010020003000620060012130000100
40204500332010110120000100200000300012790852010020003000620060098130000100
40204500332010110120000100200000300012790852010020003000620060012130000100

1000 unrolls and 10 iterations

Result (median cycles for code, minus 2 chain cycles): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)dispatch ldst uop (58)int uops in schedulers (59)simd uops in schedulers (5a)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map ldst uop (7d)map simd uop (7e)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)
4002450033200111120000102000003001279085200102003000020060000103000010
4002450033200111120000102000003001279085200102003000020060000103000010
4002450033200111120000102000003001279085200102003000020060000103000010
4002450033200111120000102000003001279085200102003000020060000103000010
4002450033200111120000102000003001279085200102003000020060000103000010
4002450033200111120000102000003001279085200102003000020060000103000010
4002450033200111120000102000003001279085200102003000020060000103000010
4002450033200111120000102000003001279085200102003000020060000103000010
4002450033200111120000102000003001279085200102003000020060000103000010
4002450033200111120000102000003001279085200102003000020060000103000010

Test 4: Latency 1->4

Chain cycles: 2

Code:

  pmull v0.8h, v1.8b, v2.8b
  eor v0.16b, v0.16b, v3.16b
  eor v3.16b, v3.16b, v0.16b
  nop
  movi v0.16b, 1
  movi v1.16b, 2
  movi v2.16b, 3

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code, minus 2 chain cycles): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001281092202882003020120060012130000100
40204500332010110120000100200003001281451203372003025120060012130000100
40204500332010110120000100200003001279085201002003000620060404130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204502642020510120104100201913001279085201002003000620060012130000100
40204500332010110120000100200003071281128202932023020620060012130000100
40204500332010110120000100200003001281110202892003020320060012130000100
40204500332010110120000100200003001279085201002003000620260392230000100

1000 unrolls and 10 iterations

Result (median cycles for code, minus 2 chain cycles): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
40025505282024111202301020427301282121202952030302206038213000010
40024504392019513201821220336301282111202942030291206028813000010
40024501492006311200521020098301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301280623201542030149206040213000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010

Test 5: throughput

Count: 8

Code:

  pmull v0.8h, v8.8b, v9.8b
  eor v0.16b, v0.16b, v10.16b
  pmull v1.8h, v8.8b, v9.8b
  eor v1.16b, v1.16b, v10.16b
  pmull v2.8h, v8.8b, v9.8b
  eor v2.16b, v2.16b, v10.16b
  pmull v3.8h, v8.8b, v9.8b
  eor v3.16b, v3.16b, v10.16b
  pmull v4.8h, v8.8b, v9.8b
  eor v4.16b, v4.16b, v10.16b
  pmull v5.8h, v8.8b, v9.8b
  eor v5.16b, v5.16b, v10.16b
  pmull v6.8h, v8.8b, v9.8b
  eor v6.16b, v6.16b, v10.16b
  pmull v7.8h, v8.8b, v9.8b
  eor v7.16b, v7.16b, v10.16b
  movi v8.16b, 9
  movi v9.16b, 10
  movi v10.16b, 11

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.5011

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
16020440479801091018000810080012300320052801122001600242003200481160000100
16020440113801101018000910080013300320373801742001600922003200481160000100
16020440100801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200521160000100
16020440100801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.5052

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
16002443845800201180009108001330320052800222016002420320000116000010
16002441164800111180000108000030320000800102016000020320000116000010
16002440426800111180000108000030320656801322016013220320000116000010
16002440403800111180000108000030321312802542016026420320000116000010
16002440401800111180000108000030320000800102016000020320000116000010
16002440424800111180000108000030320000800102016000020320000116000010
16002440401800111180000108000030320000800102016000020320000116000010
16002440417800111180000108000030320000800102016000020320000116000010
16002440417800111180000108000030320000800102016000020320000116000010
16002440417800111180000108000030320000800102016000020320000116000010