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 (1Q)

Test 1: uops

Code:

  pmull v0.1q, v1.1d, v2.1d
  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)
200410291001110001000400010002000400012000
20045531001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000
20045351001110001000400010002000400012000

Test 2: Latency 1->2

Chain cycles: 2

Code:

  pmull v0.1q, v1.1d, v2.1d
  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 ldst uop inputs (80)map simd uop inputs (81)? int output thing (e9)? ldst retires (ed)? simd retires (ee)? int retires (ef)
4020450033201011012000010020000300127908520100200300062000600121030000100
4020450033201011012000010020000300127908520100200300062000600121030000100
4020550066201101032000710220032300127908520100200300062000600121030000100
4020450033201011012000010020000300127908520100200300062000600121030000100
4020450033201011012000010020000300127908520100200300062000600121030000100
4020450033201011012000010020000300127908520100200300062000600121030000100
4020450033201011012000010020000300127908520100200300062000600121030000100
4020450033201011012000010020000300127908520100200300062000600121030000100
4020450033201011012000010020000300127942420132200300482000600121030000100
4020450033201011012000010020000300127908520100200300062000600121030000100

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)
40024500332001111200001020000301279085200102030006206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010
40024500332001111200001020000301279085200102030000206000013000010

Test 3: Latency 1->3

Chain cycles: 2

Code:

  pmull v0.1q, v1.1d, v2.1d
  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)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
40204500332010110120000100200003001279424201322003004820060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100

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 simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40025500662001811200071020032030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010

Test 4: Latency 1->4

Chain cycles: 2

Code:

  pmull v0.1q, v1.1d, v2.1d
  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
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100
40204500332010110120000100200003001279085201002003000620060012130000100

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 simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
40024500332001111200001020000030012790852001020030006206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206000013000010
40024500332001111200001020000030012790852001020030000206010813000010
40024500332001111200001020000030012790852001020030000206000013000010

Test 5: throughput

Count: 8

Code:

  pmull v0.1q, v8.1d, v9.1d
  eor v0.16b, v0.16b, v10.16b
  pmull v1.1q, v8.1d, v9.1d
  eor v1.16b, v1.16b, v10.16b
  pmull v2.1q, v8.1d, v9.1d
  eor v2.16b, v2.16b, v10.16b
  pmull v3.1q, v8.1d, v9.1d
  eor v3.16b, v3.16b, v10.16b
  pmull v4.1q, v8.1d, v9.1d
  eor v4.16b, v4.16b, v10.16b
  pmull v5.1q, v8.1d, v9.1d
  eor v5.16b, v5.16b, v10.16b
  pmull v6.1q, v8.1d, v9.1d
  eor v6.16b, v6.16b, v10.16b
  pmull v7.1q, v8.1d, v9.1d
  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)
16020440492801091018000810080012300320052801122001600242003200481160000100
16020440113801101018000910080013300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100
16020440089801091018000810080012300320052801122001600242003200481160000100
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)
16002443778800191180008108001230320000800102016000020320000116000010
16002441150800111180000108000030320000800102016000020320000116000010
16002440436800111180000108000030320000800102016000020320000116000010
16002440423800111180000108000030320000800102016000020320000116000010
16002440435800111180000108000030320000800102016000020320000116000010
16002440408800111180000108000030320000800102016000020320000116000010
16002440411800111180000108000030320000800102016000020320000116000010
16002440408800111180000108000030320000800102016000020320000116000010
16002440426800111180000108000030320000800102016000020320000116000010
16002540505800731180062108007330320000800102016000020320000116000010