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

FCVTPU (scalar, H to X)

Test 1: uops

Code:

  fcvtpu x0, h0
  mov x0, 1
  mov x1, 2

(no loop instructions)

1000 unrolls and 1 iteration

Retires: 2.000

Issues: 3.000

Integer unit issues: 1.001

Load/store unit issues: 0.000

SIMD/FP unit issues: 2.000

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)? int retires (ef)
20041040300110012000200011000200020002000100110001000
20041037300110012000200011000200020002000100110001000
20041037300110012000200011000200020002000100110001000
20041037300110012000200011000200020002000100110001000
20041037300110012000200011000200020002000100110001000
20041037300110012000200011000200020002000100110001000
20041037300110012000200011000200020002000100110001000
20041037300110012000200011000200020002000100110001000
20041037300110012000200011000200020002000100110001000
20041037300110012000200011000200020002000100110001000

Test 2: Latency 1->2 roundtrip

Code:

  fcvtpu x0, h0
  fmov d0, x0
  mov x0, 1
  mov x1, 2

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 10.0032

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)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)
3020410003240101101012000010000100200001000030015467862578206301002001000220004200100022000410001100001000010100
3020410003240101101012000010000100200001000030015467862578206301002001000220004200100012000310001100001000010100
3020410003240101101012000010000100200001000030015467962578206301002001000120003200100012000310001100001000010100
3020410003240101101012000010000100200001000030015467962578206301002001000120003200100012000310001100001000010100
3020410003240101101012000010000100200001000030015467962578206301002001000120003200100202004010002100001000010100
3020410003240101101012000010000100200001000030015467962578206301002001000120003200100012000310001100001000010100
3020410003240101101012000010000100200001000030015467962578206301002001000120003200100012000310001100001000010100
3020410003240101101012000010000100200001000030015467962578206301002001000120003200100012000310001100001000010100
3020410003240101101012000010000100200001000030015467962578206301002001000120003200100012000310001100001000010100
3020410003240101101012000010000100200001000030015467962578206301002001000120003200100012000310001100001000010100

1000 unrolls and 10 iterations

Result (median cycles for code): 10.0032

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)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)
300241000354001110011200001000010200001000030154633625782063001020100022000420100002000010001100001000010010
300241000324001110011200001000010200001000030154634625782063001020100002000020100002000010001100001000010010
300241000324001110011200001000010200001000030154634625782063001020100002000020100002000010001100001000010010
300241000324001110011200001000010200001000030154634625782063001020100002000020100002000010001100001000010010
300241000324001110011200001000010200001001830154652525784833005620100202004020100002000010001100001000010010
300241000324001110011200001000010200001001830154656225785613005620100202004020100002000010001100001000010010
300241000324001110011200001000010200001000030154634625782063001020100002000020100002000010001100001000010010
300241000324001110011200001000010200001000030154637825782583001020100002000020100192004110002100001000010010
300241000534001110011200001000010200001000030154638025782583001020100002000020100002000010001100001000010010
300241000324001110011200001000010200001000030154634625782063001020100002000020100002000010001100001000010010

Test 3: throughput

Count: 8

Code:

  fcvtpu x0, h8
  fcvtpu x1, h8
  fcvtpu x2, h8
  fcvtpu x3, h8
  fcvtpu x4, h8
  fcvtpu x5, h8
  fcvtpu x6, h8
  fcvtpu x7, h8
  mov x8, 9

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 1.0005

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)
1602048005724010380101160002100160008300880041160108200160012200160012800018000080100
1602048003724010380101160002100160008300880041160108200160012200160012800018000080100
1602048003724010380101160002100160008300880041160108200160012200160012800018000080100
1602048003724010380101160002100160008300880041160108200160012200160012800018000080100
1602048003724010380101160002100160008300880041160108200160012200160012800018000080100
1602048003724010380101160002100160008300880041160108200160012200160012800018000080100
1602058007424015580115160040100160058300880041160108200160012200160012800018000080100
1602048003724010380101160002100160008300880041160108200160012200160012800018000080100
1602048003724010380101160002100160008300880041160108200160012200160012800018000080100
1602048003724010380101160002100160008300880041160108200160012200160012800018000080100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 1.0005

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)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)
1600248003724001380011160002010160008308800411600182016001220160000800018000080010
1600248003724001180011160000010160000308802901600682016006820160000800018000080010
1600248003724001180011160000010160000308800001600102016000020160000800018000080010
1600248003724001180011160000010160000308800001600102016000020160000800018000080010
1600248003724001180011160000010160000308800001600102016000020160000800018000080010
1600248003724001180011160000010160000308802371600582016005420160000800018000080010
136581810842005106839313035317642993130370308800411600182016001220160000800018000080010
1600248003724001180011160000010160000308800001600102016000020160000800018000080010
1600248003724001180011160000010160000308800001600102016000020160000800018000080010
1600248003724001180011160000010160000308800001600102016000020160000800018000080010