Icestorm: FMLAL2 (by element, 2S) measurements

Code:

  fmlal2 v0.2s, v1.2h, v2.h[1]

Initialization

  movi v0.16b, 1
  movi v1.16b, 2
  movi v2.16b, 3

(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

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

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)
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000
1004	4033	1001	1	1000	1000	101557	1000	1000	3000	1	1000

Test 2: Latency 1->1

Code:

  fmlal2 v0.2s, v1.2h, v2.h[1]

Initialization

  movi v0.16b, 1
  movi v1.16b, 2
  movi v2.16b, 3

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 4.0033

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

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)
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	307	1028904	10132	202	10046	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0033

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

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)
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10004	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028890	10050	20	10041	20	30012	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10025	40066	10025	21	10004	20	10030	70	1028557	10020	20	10000	20	30012	11	10000	10

Test 3: Latency 1->2

Code:

  fmlal2 v0.2s, v0.2h, v1.h[1]

Initialization

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

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 4.0033

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

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)
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10006	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30132	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0033

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

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)
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10006	20	30000	11	10000	10
10025	40066	10025	21	10004	20	10030	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10

Test 4: Latency 1->3

Code:

  fmlal2 v0.2s, v1.2h, v0.h[1]

Initialization

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

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 4.0033

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

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)
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10006	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30012	1	10000	100
10204	40033	10101	101	10000	100	10000	300	1028557	10100	200	10004	200	30132	1	10000	100

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0033

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

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)
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10006	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30132	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	70	1028557	10020	20	10000	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	69	1028904	10050	20	10046	20	30000	11	10000	10
10024	40033	10021	21	10000	20	10000	65	1028557	10020	20	10005	20	30000	11	10000	10

Test 5: throughput

Count: 8

Code:

  movi v0.16b, 0
  fmlal2 v0.2s, v8.2h, v9.h[1]
  movi v1.16b, 0
  fmlal2 v1.2s, v8.2h, v9.h[1]
  movi v2.16b, 0
  fmlal2 v2.2s, v8.2h, v9.h[1]
  movi v3.16b, 0
  fmlal2 v3.2s, v8.2h, v9.h[1]
  movi v4.16b, 0
  fmlal2 v4.2s, v8.2h, v9.h[1]
  movi v5.16b, 0
  fmlal2 v5.2s, v8.2h, v9.h[1]
  movi v6.16b, 0
  fmlal2 v6.2s, v8.2h, v9.h[1]
  movi v7.16b, 0
  fmlal2 v7.2s, v8.2h, v9.h[1]

Initialization

  movi v8.16b, 9
  movi v9.16b, 10

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.5011

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

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)
160204	40500	80110	101	80009	100	80013	300	320056	80113	200	80013	200	240039	1	160000	100
160204	40121	80110	101	80009	100	80013	300	320052	80112	200	80012	200	240036	1	160000	100
160204	40101	80110	101	80009	100	80013	300	320196	80148	200	80048	200	240036	1	160000	100
160204	40086	80109	101	80008	100	80012	300	320052	80112	200	80012	200	240036	1	160000	100
160204	40086	80109	101	80008	100	80012	300	320052	80112	200	80012	200	240036	1	160000	100
160204	40086	80109	101	80008	100	80012	300	320052	80112	200	80012	200	240036	1	160000	100
160204	40086	80109	101	80008	100	80012	300	320052	80112	200	80012	200	240036	1	160000	100
160204	40086	80109	101	80008	100	80012	300	320052	80112	200	80012	200	240036	1	160000	100
160204	40086	80109	101	80008	100	80012	300	320052	80112	200	80012	200	240036	1	160000	100
160204	40086	80109	101	80008	100	80012	300	320052	80112	200	80012	200	240036	1	160000	100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.5054

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

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)
160024	43757	80020	11	80009	10	80013	30	320056	80023	20	80013	20	240000	1	160000	10
160024	41416	80011	11	80000	10	80000	30	320000	80010	20	80000	20	240000	1	160000	10
160025	40501	80055	11	80044	10	80048	30	320000	80010	20	80000	20	240000	1	160000	10
160024	40422	80011	11	80000	10	80000	30	320000	80010	20	80000	20	240000	1	160000	10
160024	40437	80011	11	80000	10	80000	30	320000	80010	20	80000	20	240000	1	160000	10
160024	40400	80011	11	80000	10	80000	30	320000	80010	20	80000	20	240000	1	160000	10
160024	40403	80011	11	80000	10	80000	30	320000	80010	20	80000	20	240000	1	160000	10
160024	40401	80011	11	80000	10	80000	30	320000	80010	20	80000	20	240000	1	160000	10
160024	40411	80011	11	80000	10	80000	30	320000	80010	20	80000	20	240000	1	160000	10
160024	40421	80011	11	80000	10	80000	30	320000	80010	20	80000	20	240000	1	160000	10

Test 6: throughput

Count: 12

Code:

  fmlal2 v0.2s, v12.2h, v13.h[1]
  fmlal2 v1.2s, v12.2h, v13.h[1]
  fmlal2 v2.2s, v12.2h, v13.h[1]
  fmlal2 v3.2s, v12.2h, v13.h[1]
  fmlal2 v4.2s, v12.2h, v13.h[1]
  fmlal2 v5.2s, v12.2h, v13.h[1]
  fmlal2 v6.2s, v12.2h, v13.h[1]
  fmlal2 v7.2s, v12.2h, v13.h[1]
  fmlal2 v8.2s, v12.2h, v13.h[1]
  fmlal2 v9.2s, v12.2h, v13.h[1]
  fmlal2 v10.2s, v12.2h, v13.h[1]
  fmlal2 v11.2s, v12.2h, v13.h[1]

Initialization

  movi v12.16b, 13
  movi v13.16b, 14

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.5003

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

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 ldst uop inputs (80)	map simd uop inputs (81)	? int output thing (e9)	? ldst retires (ed)	? simd retires (ee)	? int retires (ef)
120204	60085	120105	101	120004	100	120008	300	480044	120110	200	120014	200	0	360042	1	0	120000	100
120204	60036	120107	101	120006	100	120010	300	480036	120108	200	120012	200	0	360192	1	0	120000	100
120204	60036	120107	101	120006	100	120010	300	480036	120108	200	120012	200	0	360036	1	0	120000	100
120204	60036	120105	101	120004	100	120008	300	480036	120108	200	120012	200	0	360036	1	0	120000	100
120204	60036	120105	101	120004	100	120008	300	480036	120108	200	120012	200	0	360039	1	0	120000	100
120204	60036	120105	101	120004	100	120008	300	480036	120108	200	120012	200	0	360036	1	0	120000	100
120204	60036	120105	101	120004	100	120008	300	480036	120108	200	120012	200	0	360036	1	0	120000	100
120204	60036	120105	101	120004	100	120008	300	480036	120108	200	120012	200	0	360036	1	0	120000	100
120205	60072	120147	101	120046	100	120058	300	480036	120108	200	120012	200	0	360036	1	0	120000	100
120204	60036	120105	101	120004	100	120008	300	480036	120108	200	120012	200	0	360036	1	0	120000	100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.5003

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

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)
120024	60220	120017	11	120006	0	10	120010	0	30	0	481032	120270	20	0	120270	20	0	360000	1	0	120000	10
120024	60036	120011	11	120000	0	10	120000	0	30	0	480000	120010	20	0	120000	20	0	360000	1	0	120000	10
120024	60036	120011	11	120000	0	10	120000	0	30	0	480000	120010	20	0	120000	20	0	360000	1	0	120000	10
120024	60036	120011	11	120000	0	10	120000	0	30	0	480000	120010	20	0	120000	20	0	360000	1	0	120000	10
120024	60036	120011	11	120000	0	10	120000	0	30	0	480000	120010	20	0	120000	20	0	360000	1	0	120000	10
120024	60036	120011	11	120000	0	10	120000	0	30	0	480000	120010	20	0	120000	20	0	360000	1	0	120000	10
120024	60036	120011	11	120000	0	10	120000	0	30	0	480000	120010	20	0	120000	20	0	360000	1	0	120000	10
120024	60036	120011	11	120000	0	10	120000	0	30	0	480000	120010	20	0	120000	20	0	360000	1	0	120000	10
120024	60036	120011	11	120000	0	10	120000	0	30	0	480000	120010	20	0	120000	20	0	360000	1	0	120000	10
120024	60036	120011	11	120000	0	10	120000	0	30	0	480212	120065	20	0	120062	20	0	360000	1	0	120000	10

FMLAL2 (by element, 2S)

Test 1: uops

1000 unrolls and 1 iteration

Test 2: Latency 1->1

100 unrolls and 100 iterations

1000 unrolls and 10 iterations

Test 3: Latency 1->2

100 unrolls and 100 iterations

1000 unrolls and 10 iterations

Test 4: Latency 1->3

100 unrolls and 100 iterations

1000 unrolls and 10 iterations

Test 5: throughput

100 unrolls and 100 iterations

1000 unrolls and 10 iterations

Test 6: throughput

100 unrolls and 100 iterations

1000 unrolls and 10 iterations