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

USRA (vector, 4S)

Test 1: uops

Code:

  usra v0.4s, v1.4s, #3
  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->1

Code:

  usra v0.4s, v1.4s, #3
  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)
1020430033101011011000010010000300768905101002001000620020008110000100
1020430033101011011000010010000300768905101002001000420020010110000100
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)
10024300331002121100002010000707689051002020100062020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020008111000010
10025300661002721100062010031707689051002020100002020084111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100062020000111000010

Test 3: Latency 1->2

Code:

  usra v0.4s, v0.4s, #3
  movi v0.16b, 1

(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)
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000307769162101332021004220020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020096110000100
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)
10024300331002121100002010000707689051002020100042020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020318111000010

Test 4: throughput

Count: 8

Code:

  movi v0.16b, 0
  usra v0.4s, v8.4s, #3
  movi v1.16b, 0
  usra v1.4s, v8.4s, #3
  movi v2.16b, 0
  usra v2.4s, v8.4s, #3
  movi v3.16b, 0
  usra v3.4s, v8.4s, #3
  movi v4.16b, 0
  usra v4.4s, v8.4s, #3
  movi v5.16b, 0
  usra v5.4s, v8.4s, #3
  movi v6.16b, 0
  usra v6.4s, v8.4s, #3
  movi v7.16b, 0
  usra v7.4s, v8.4s, #3
  movi v8.16b, 9

(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)
1602044044980110101800091008001330032005280112200800122001600261160000100
1602044012180109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602054013180147101800461008005030032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100

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)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)
16002443941800191180008010800123032000080010208000020160000116000010
16002441260800111180000010800003032000080010208000020160000116000010
16002440424800111180000010800003032019680058208004820160100116000010
174825848089216777688004243577285800463032020080059208004920160024116000010
16002440425800111180000010800003032000080010208000020160000116000010
16002440405800111180000010800003032000080010208000020160000116000010
16002440427800111180000010800003032000080010208000020160000116000010
16002440397800111180000010800003032000080010208000020160000116000010
16002440397800111180000010800003032000080010208000020160000116000010
16002440396800111180000010800003032000080010208000020160000116000010

Test 5: throughput

Count: 16

Code:

  usra v0.4s, v16.4s, #3
  usra v1.4s, v16.4s, #3
  usra v2.4s, v16.4s, #3
  usra v3.4s, v16.4s, #3
  usra v4.4s, v16.4s, #3
  usra v5.4s, v16.4s, #3
  usra v6.4s, v16.4s, #3
  usra v7.4s, v16.4s, #3
  usra v8.4s, v16.4s, #3
  usra v9.4s, v16.4s, #3
  usra v10.4s, v16.4s, #3
  usra v11.4s, v16.4s, #3
  usra v12.4s, v16.4s, #3
  usra v13.4s, v16.4s, #3
  usra v14.4s, v16.4s, #3
  usra v15.4s, v16.4s, #3
  movi v16.16b, 17

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.5002

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)
160204801021601051011600041001600083006400361601082001600122003200281160000100
160204800351601071011600061001600103006400361601082001600122003200241160000100
160204800351601051011600041001600083006400361601082001600122003200241160000100
160204800351601051011600041001600083006400361601082001600122003201281160000100
160204800351601051011600041001600083006402081601542001600632003200241160000100
160205800731601421011600411001600533006400441601102001600142003200241160000100
160204800351601051011600041001600083006400361601082001600122003200281160000100
160204800351601051011600041001600083006400361601082001600122003200241160000100
160204800351601051011600041001600083006400361601082001600122003200241160000100
160204800351601051011600041001600083006400361601082001600122003200241160000100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.5002

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)
1600248010016001511160004010160008306400001600102016000020320000116000010
1600248012316001111160000010160000306400001600102016000020320000116000010
1600248004016001111160000010160000306400001600102016000020320000116000010
1600248003516001111160000010160000306400001600102016000020320000116000010
1600248003516001111160000010160000306400001600102016000020320000116000010
1600248003516001111160000010160000306400001600102016000020320000116000010
1600248003516001111160000010160000306400001600102016000020320000116000010
1600248003516001111160000010160000306400001600102016000020320000116000010
1600248003516001111160000010160000306400001600102016000020320000116000010
1600248003516001111160000010160000306401681600552016004920320000116000010