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

FMAXV (4H)

Test 1: uops

Code:

  fmaxv h0, v0.4h
  movi v0.16b, 1
  movi v1.16b, 2

(no loop instructions)

1000 unrolls and 1 iteration

Retires: 1.000

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)
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000

Test 2: Latency 1->2

Code:

  fmaxv h0, v0.4h
  movi v0.16b, 1
  movi v1.16b, 2

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 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)
1020430033101011011000010010000300768905101002001000420020012110000100
1020430033101011011000010010000300768905101002001000620020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100

1000 unrolls and 10 iterations

Result (median cycles for code): 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)
10024300331002121100002010000707689051002020100002020076111000010
10025300661002721100062010031707689051002020100002020008111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010

Test 3: throughput

Count: 8

Code:

  fmaxv h0, v8.4h
  fmaxv h1, v8.4h
  fmaxv h2, v8.4h
  fmaxv h3, v8.4h
  fmaxv h4, v8.4h
  fmaxv h5, v8.4h
  fmaxv h6, v8.4h
  fmaxv h7, v8.4h
  movi v8.16b, 9

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.5004

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)
80204400568010510180004100800083003200368010820080012200160024180000100
80204400358010710180006100800103003200368010820080012200160024180000100
80204400358010510180004100800083003200368010820080012200160024180000100
80204400358010510180004100800083003200368010820080012200160024180000100
80204400358010510180004100800083003200368010820080012200160024180000100
80204400358010510180004100800083003200368010820080012200160024180000100
80204400358010510180004100800083003200368010820080012200160024180000100
80204400358010510180004100800083003200368010820080012200160024180000100
80204400358010510180004100800083003200368010820080012200160024180000100
80204400358010510180004100800083003200368010820080012200160128180000100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.5004

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)
800244024080027218000620800107032004480030208001620160000118000010
800244003580021218000020800007032000080020208000020160128118000010
800244004580021218000020800007032000080020208000020160000118000010
800244003580021218000020800007032000080020208000020160000118000010
800244003580021218000020800007032000080020208000020160000118000010
800244003580021218000020800007032000080020208000020160000118000010
800244003580021218000020800007032000080020208000020160000118000010
800244003580021218000020800007032000080020208000020160000118000010
800244003580021218000020800007032000080020208000020160036118000010
800244003580021218000020800007032000080020208000020160000118000010