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

FSQRT (scalar, H)

Test 1: uops

Code:

  fsqrt h0, h0
  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)
10048033100111000100010225010001000100011000
10048033100111000100010225010001000100011000
10048033100111000100010225010001000100011000
10048033100111000100010225010001000100011000
10048033100111000100010225010001000100011000
10048033100111000100010225010001000100011000
10048033100111000100010225010001000100011000
10048033100111000100010225010001000100011000
10048033100111000100010225010001000100011000
10048033100111000100010225010001000100011000

Test 2: Latency 1->2

Code:

  fsqrt h0, h0
  movi v0.16b, 1
  movi v1.16b, 2

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 8.0033

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)
10204800331010110110000100100003001029250101002001000620010004110000100
10204800331010110110000100100003001029250101002001000420010004110000100
10204800331010110110000100100003001029250101002001000420010004110000100
10204800331010110110000100100003001029250101002001000420010004110000100
10205800661010510310002102100153001029250101002001000420010004110000100
10204800331010110110000100100003001029250101002001000420010004110000100
10204800331010110110000100100003001029250101002001000420010004110000100
10204800331010110110000100100003001029250101002001000420010004110000100
10205800661010310110002100100153001029250101002001000420010004110000100
10204800331010110110000100100003001029250101002001000420010004110000100

1000 unrolls and 10 iterations

Result (median cycles for code): 8.0033

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)
100248003310021211000020100007010292501002020100002010000111000010
100248003310021211000020100007010292501002020100002010000111000010
100248003310021211000020100007010292501002020100002010000111000010
100248003310021211000020100007010292501002020100002010000111000010
100248003310021211000020100007010292501002020100002010000111000010
100248003310021211000020100007010292501002020100002010000111000010
100248003310021211000020100007010292501002020100002010000111000010
100248003310021211000020100007010292501002020100002010000111000010
100248003310021211000020100007010292501002020100002010000111000010
100248003310021211000020100007010292501002020100002010000111000010

Test 3: throughput

Count: 8

Code:

  fsqrt h0, h8
  fsqrt h1, h8
  fsqrt h2, h8
  fsqrt h3, h8
  fsqrt h4, h8
  fsqrt h5, h8
  fsqrt h6, h8
  fsqrt h7, h8
  movi v8.16b, 9

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 2.0005

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)
8020416003980101101800000100800003001999769801002008000420080004180000100
8020416003980101101800000100800003001999769801002008000420080004180000100
8020416003980101101800000100800003001999769801002008000420080004180000100
8020416003980101101800000100800003001999982801242008003520080004180000100
8020416003980101101800000100800003001999769801002008000420080004180000100
8020416003980101101800000100800003001999769801002008000420080004180000100
8020416003980101101800000100800003001999769801002008000420080004180000100
8020416003980101101800000100800003001999982801242008003620080004180000100
8020416003980101101800000100800003001999769801002008000420080004180000100
8020416003980101101800000100800003001999769801002008000420080004180000100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 2.0005

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)
8002416003980021218000020800007019997698002020800042080000118000010
8002416003980021218000020800007019997698002020800002080000118000010
8002416003980021218000020800007019997698002020800002080000118000010
8002516007880032218001120800247019997698002020800002080000118000010
8002416003980021218000020800007019997698002020800002080000118000010
8002416003980021218000020800007019997698002020800002080000118000010
8002516007880032218001120800247019997698002020800002080036118000010
8002416003980021218000020800007019997698002020800002080000118000010
8002416003980021218000020800007019997698002020800002080000118000010
8002416003980021218000020800007019997698002020800002080000118000010