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

LDRSW

Test 1: uops

Code:

  ldrsw x0, [x6]

Initialization

  mov x0, 1
  mov x1, 2
  mov x8, 0

(no loop instructions)

1000 unrolls and 1 iteration

Retires: 1.000

Issues: 1.000

Integer unit issues: 0.001

Load/store unit issues: 1.000

SIMD/FP unit issues: 0.000

Recorded non-zero counters (first 10 of 64 runs)

retire uop (01)	cycle (02)	schedule uop (52)	schedule int uop (53)	schedule ldst uop (55)	dispatch ldst uop (58)	simd uops in schedulers (5a)	dispatch uop (78)	map ldst uop (7d)	map ldst uop inputs (80)	? int output thing (e9)	? ldst retires (ed)
1005	705	1025	1	1024	1000	8238	1000	1000	1000	1	1000
1004	554	1001	1	1000	1000	8238	1000	1000	1000	1	1000
1004	554	1001	1	1000	1000	8292	1000	1000	1000	1	1000
1004	554	1001	1	1000	1000	8238	1000	1000	1000	1	1000
1004	554	1001	1	1000	1000	8238	1000	1000	1000	1	1000
1004	554	1001	1	1000	1000	8238	1000	1000	1000	1	1000
1004	554	1001	1	1000	1000	8238	1000	1000	1000	1	1000
1004	554	1001	1	1000	1000	8238	1000	1000	1000	1	1000
1004	554	1001	1	1000	1000	8238	1000	1000	1000	1	1000
1004	554	1001	1	1000	1000	8238	1000	1000	1000	1	1000

Test 2: Latency 1->2 (with chain penalty)

Chain cycles: 3

Code:

  ldrsw x0, [x6]
  eor x8, x8, x0
  eor x8, x8, x0
  add x6, x6, x8

Initialization

  mov x0, 1
  mov x1, 2
  mov x8, 0

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code, minus 3 chain cycles): 4.0042

Recorded non-zero counters (first 10 of 64 runs)

retire uop (01)	cycle (02)	schedule uop (52)	schedule int uop (53)	schedule ldst uop (55)	dispatch int uop (56)	dispatch ldst uop (58)	int uops in schedulers (59)	simd uops in schedulers (5a)	dispatch uop (78)	map int uop (7c)	map ldst uop (7d)	map int uop inputs (7f)	map ldst uop inputs (80)	? int output thing (e9)	? ldst retires (ed)	? int retires (ef)
40205	70158	40108	30107	10001	30130	10003	1859824	694153	40106	30212	10004	60224	10004	30003	10000	30100
40204	70042	40102	30102	10000	30103	10003	1859446	694013	40106	30212	10004	60224	10004	30002	10000	30100
40204	70042	40102	30102	10000	30103	10003	1859446	694013	40106	30212	10004	60224	10004	30002	10000	30100
40204	70042	40102	30102	10000	30103	10003	1859446	694013	40106	30212	10004	60224	10004	30002	10000	30100
40204	70042	40102	30102	10000	30103	10003	1859446	694013	40106	30212	10004	60224	10004	30002	10000	30100
40204	70042	40102	30102	10000	30103	10003	1859446	694013	40106	30212	10004	60224	10004	30002	10000	30100
40204	70042	40102	30102	10000	30103	10003	1859446	694013	40106	30212	10004	60224	10004	30002	10000	30100
40204	70042	40102	30102	10000	30103	10003	1860175	694303	40106	30212	10004	60224	10004	30002	10000	30100
40204	70042	40102	30102	10000	30103	10003	1859446	694013	40106	30212	10004	60224	10004	30002	10000	30100
40204	70042	40102	30102	10000	30103	10003	1859446	694013	40106	30212	10004	60224	10004	30002	10000	30100

1000 unrolls and 10 iterations

Result (median cycles for code, minus 3 chain cycles): 4.0040

Recorded non-zero counters (first 10 of 64 runs)

retire uop (01)	cycle (02)	schedule uop (52)	schedule int uop (53)	schedule ldst uop (55)	dispatch int uop (56)	dispatch ldst uop (58)	int uops in schedulers (59)	simd uops in schedulers (5a)	dispatch uop (78)	map int uop (7c)	map ldst uop (7d)	map int uop inputs (7f)	map ldst uop inputs (80)	? int output thing (e9)	? ldst retires (ed)	? int retires (ef)
40025	70447	40018	30017	10001	30040	10003	1859623	694693	40016	30032	10004	60020	10000	30003	10000	30010
40024	70047	40013	30013	10000	30010	10000	1859652	694714	40010	30020	10000	60020	10000	30003	10000	30010
40024	70040	40012	30012	10000	30010	10000	1859652	694714	40010	30020	10000	60020	10000	30003	10000	30010
40024	70040	40012	30012	10000	30010	10000	1859652	694714	40010	30020	10000	60122	10017	30009	10000	30010
40024	70040	40012	30012	10000	30010	10000	1859652	694714	40010	30020	10000	60020	10000	30003	10000	30010
40024	70040	40012	30012	10000	30010	10000	1859652	694714	40010	30020	10000	60020	10000	30003	10000	30010
40024	70040	40012	30012	10000	30010	10000	1859652	694714	40010	30020	10000	60020	10000	30003	10000	30010
40024	70040	40012	30012	10000	30010	10000	1859652	694714	40010	30020	10000	60020	10000	30003	10000	30010
40024	70040	40012	30012	10000	30010	10000	1861731	695561	40010	30020	10000	60020	10000	30003	10000	30010
40024	70043	40012	30012	10000	30010	10000	1859706	694736	40010	30020	10000	60020	10000	30003	10000	30010

Test 3: throughput

Count: 8

Code:

  ldrsw x0, [x6]
  ldrsw x0, [x6]
  ldrsw x0, [x6]
  ldrsw x0, [x6]
  ldrsw x0, [x6]
  ldrsw x0, [x6]
  ldrsw x0, [x6]
  ldrsw x0, [x6]

Initialization

  mov x7, x6
  mov x8, x6
  mov x9, x6
  mov x10, x6
  mov x11, x6
  mov x12, x6
  mov x13, x6

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.5006

Recorded non-zero counters (first 10 of 64 runs)

retire uop (01)	cycle (02)	schedule uop (52)	schedule int uop (53)	schedule ldst uop (55)	dispatch int uop (56)	dispatch ldst uop (58)	int uops in schedulers (59)	simd uops in schedulers (5a)	dispatch uop (78)	map int uop (7c)	map ldst uop (7d)	map int uop inputs (7f)	map ldst uop inputs (80)	? int output thing (e9)	? ldst retires (ed)	? int retires (ef)
80205	40523	80137	101	80036	100	80008	300	248262	80108	200	80012	200	80012	1	80000	100
80204	40049	80101	101	80000	100	80008	300	640142	80108	200	80012	200	80012	1	80000	100
80204	40049	80101	101	80000	100	80008	300	640142	80108	200	80012	200	80012	1	80000	100
80204	40049	80101	101	80000	100	80008	300	640142	80108	200	80012	200	80012	1	80000	100
80204	40049	80101	101	80000	100	80008	300	640142	80108	200	80012	200	80012	1	80000	100
80204	40049	80101	101	80000	100	80008	300	640142	80108	200	80012	200	80012	1	80000	100
80204	40049	80101	101	80000	100	80008	300	640142	80108	200	80012	200	80012	1	80000	100
80204	40049	80101	101	80000	100	80008	300	640142	80108	200	80012	200	80012	1	80000	100
80204	40049	80101	101	80000	100	80008	300	640142	80108	200	80012	200	80012	1	80000	100
80204	40049	80101	101	80000	100	80008	300	640142	80108	200	80012	200	80072	1	80000	100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.5006

Recorded non-zero counters (first 10 of 64 runs)

retire uop (01)	cycle (02)	schedule uop (52)	schedule int uop (53)	schedule ldst uop (55)	dispatch int uop (56)	dispatch ldst uop (58)	int uops in schedulers (59)	simd uops in schedulers (5a)	dispatch uop (78)	map int uop (7c)	map ldst uop (7d)	map int uop inputs (7f)	map ldst uop inputs (80)	? int output thing (e9)	? ldst retires (ed)	? int retires (ef)
80025	40285	80039	11	80028	10	80008	30	320262	80018	20	80012	20	80000	1	80000	10
80024	40054	80011	11	80000	10	80000	30	640112	80010	20	80000	20	80000	1	80000	10
80025	40101	80046	11	80035	10	80000	30	642344	80010	20	80000	20	80000	1	80000	10
80024	40056	80011	11	80000	10	80000	30	640112	80010	20	80000	20	80000	1	80000	10
80024	40047	80011	11	80000	10	80000	30	640112	80010	20	80000	20	80000	1	80000	10
80024	40047	80011	11	80000	10	80000	30	640112	80010	20	80000	20	80000	1	80000	10
80024	40047	80011	11	80000	10	80000	30	640112	80010	20	80000	20	80000	1	80000	10
80024	40047	80011	11	80000	10	80000	30	640166	80010	20	80000	20	80000	1	80000	10
80024	40047	80011	11	80000	10	80000	30	640112	80010	20	80000	20	80000	1	80000	10
80025	40100	80046	11	80035	10	80000	30	640112	80010	20	80000	20	80000	1	80000	10