Firestorm: SADALP (vector, 4H) measurements

Code:

  sadalp v0.4h, v1.8b

Initialization

  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: 0.000

Load/store unit issues: 0.000

SIMD/FP unit issues: 1.000

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

retire uop (01)	cycle (02)	03	19	1e	3f	4e	51	schedule uop (52)	schedule simd uop (54)	dispatch simd uop (57)	ldst uops in schedulers (5b)	60	69	6d	6e	map stall dispatch (70)	map rewind (75)	map stall (76)	dispatch uop (78)	map simd uop (7e)	map simd uop inputs (81)	82	83	flush restart other nonspec (84)	85	inst all (8c)	inst simd alu (9a)	l1d cache writeback (a8)	ac	cf	d5	map dispatch bubble (d6)	dd	fetch restart (de)	e0	? simd retires (ee)	f5	f6	f7	f8	fd
1004	3037	23	0	0	121	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	2	16	1	1	2630	1000	3038	3038	3038	3038	3038
1004	3037	23	0	0	61	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	1	16	1	1	2630	1000	3038	3038	3038	3038	3038
1004	3037	22	0	0	61	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	1	16	1	1	2630	1000	3038	3038	3038	3038	3038
1004	3037	22	0	0	156	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	1	16	1	1	2630	1000	3038	3038	3038	3038	3038
1004	3037	22	0	0	61	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	1	16	1	1	2630	1000	3038	3038	3038	3038	3038
1004	3084	23	0	69	82	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	1	16	1	1	2630	1000	3038	3038	3038	3038	3038
1004	3037	23	0	3	61	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	1	16	1	1	2630	1000	3038	3038	3038	3038	3038
1004	3037	23	0	0	61	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	1	16	1	1	2630	1000	3038	3038	3038	3038	3038
1004	3037	23	0	0	61	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	1	16	1	1	2630	1000	3038	3038	3038	3038	3038
1004	3037	22	0	156	61	2548	25	1000	1000	1000	398313	1	3018	3037	3037	2415	3	2895	1000	1000	2000	3037	3037	1	1	1001	1000	0	0	73	1	16	1	1	2630	1000	3038	3038	3038	3038	3038

Test 2: Latency 1->1

Code:

  sadalp v0.4h, v1.8b

Initialization

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

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 3.0037

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

retire uop (01)	cycle (02)	03	1e	3f	4e	51	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)	69	6d	6e	map stall dispatch (70)	map rewind (75)	map stall (76)	dispatch uop (78)	map int uop (7c)	map simd uop (7e)	map int uop inputs (7f)	map simd uop inputs (81)	82	83	flush restart other nonspec (84)	85	inst all (8c)	inst branch (8d)	inst branch taken (90)	inst branch cond (94)	inst int alu (97)	inst simd alu (9a)	9f	l1d cache writeback (a8)	cd	cf	d5	map dispatch bubble (d6)	dd	fetch restart (de)	e0	? simd retires (ee)	? int retires (ef)	f5	f6	f7	f8	fd
10204	30037	225	6	61	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10100	200	10000	200	20000	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	61	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10100	200	10000	200	20000	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038
10204	30037	233	24	61	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10100	200	10000	200	20000	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	726	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10100	200	10000	200	20000	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038
10204	30037	225	27	61	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10410	212	10000	200	20000	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	441	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10100	200	10000	200	20000	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	156	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10100	200	10000	200	20000	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	61	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10100	200	10000	200	20000	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	61	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10100	200	10000	200	20000	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	61	29548	25	10100	100	10000	100	10000	500	4277313	30018	30037	30037	28265	3	28745	10100	200	10000	200	20000	30037	30037	1	1	10202	100	99	100	100	10000	100	0	0	710	1	16	1	1	29634	10000	100	30038	30038	30038	30038	30038

1000 unrolls and 10 iterations

Result (median cycles for code): 3.0037

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

retire uop (01)	cycle (02)	03	18	19	1e	1f	3f	4e	51	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)	60	69	6d	6e	map stall dispatch (70)	map rewind (75)	map stall (76)	dispatch uop (78)	map int uop (7c)	map simd uop (7e)	map int uop inputs (7f)	map simd uop inputs (81)	82	83	flush restart other nonspec (84)	85	inst all (8c)	inst branch (8d)	inst branch taken (90)	inst branch cond (94)	inst int alu (97)	inst simd alu (9a)	9f	l1d tlb miss (a1)	l1d cache writeback (a8)	ac	cd	cf	d5	map dispatch bubble (d6)	dd	fetch restart (de)	e0	? simd retires (ee)	? int retires (ef)	f5	f6	f7	f8	fd
10024	30037	225	0	3	12	0	71	29548	25	10010	10	10000	10	10000	50	4277313	0	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	0	0	0	640	3	16	2	2	29630	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	0	18	0	61	29548	25	10010	10	10000	10	10000	50	4277313	1	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	0	0	0	640	2	16	2	2	29630	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	0	0	0	61	29548	25	10010	10	10000	10	10000	50	4277313	0	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	0	0	0	640	2	16	2	2	29630	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	0	0	0	61	29548	25	10010	10	10000	10	10000	50	4277313	0	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30084	1	1	10021	10	9	10	10	10000	10	0	0	0	0	640	2	16	2	2	29630	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	0	0	0	61	29548	25	10010	10	10000	10	10000	50	4277313	1	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	0	0	0	640	2	16	2	2	29630	10000	10	30038	30038	30038	30038	30038
10024	30037	224	0	2	0	0	251	29548	25	10010	10	10000	10	10000	50	4277313	0	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	0	0	0	640	2	16	2	2	29630	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	2	396	176	61	29548	25	10020	10	10000	10	10000	50	4277313	1	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	0	0	0	640	2	16	2	2	29630	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	0	0	0	61	29548	25	10010	10	10000	10	10000	50	4277313	0	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	0	15	0	640	2	16	2	2	29630	10000	10	30038	30038	30038	30038	30038
10024	30037	233	0	0	0	0	61	29548	25	10010	10	10000	10	10000	50	4277313	0	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	0	0	0	640	2	16	2	2	29630	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	0	0	0	631	29548	25	10010	10	10000	10	10000	50	4277313	0	30018	30037	30037	28287	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	2	0	0	1	640	2	16	2	2	29630	10000	10	30038	30038	30038	30038	30038

Test 3: Latency 1->2

Code:

  sadalp v0.4h, v0.8b

Initialization

  movi v0.16b, 1

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 3.0037

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

retire uop (01)	cycle (02)	03	18	19	1e	3f	4e	51	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)	60	69	6d	6e	map stall dispatch (70)	map rewind (75)	map stall (76)	dispatch uop (78)	map int uop (7c)	map simd uop (7e)	map int uop inputs (7f)	map simd uop inputs (81)	82	83	flush restart other nonspec (84)	85	inst all (8c)	inst branch (8d)	inst branch taken (90)	inst branch cond (94)	inst int alu (97)	inst simd alu (9a)	9f	l1d cache writeback (a8)	ac	c2	branch cond mispred nonspec (c5)	branch mispred nonspec (cb)	cd	cf	map dispatch bubble (d6)	dd	fetch restart (de)	e0	? simd retires (ee)	? int retires (ef)	f5	f6	f7	f8	fd
10204	30037	225	0	0	0	61	29547	25	10100	100	10000	100	10000	500	4277160	0	30018	30037	30037	28274	6	28740	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	0	1	1	1	717	16	0	0	29645	10000	100	30038	30038	30038	30038	30038
10204	30037	224	0	0	0	726	29547	25	10100	100	10000	100	10000	500	4277160	0	30018	30037	30037	28271	6	28740	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	0	1	1	1	718	16	0	0	29645	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	0	0	61	29547	25	10100	100	10000	100	10000	500	4277160	0	30018	30037	30037	28271	7	28741	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	0	1	1	1	718	16	0	0	29646	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	0	0	61	29547	25	10100	100	10000	100	10000	500	4277160	0	30018	30037	30037	28271	6	28741	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	0	1	1	1	717	16	0	0	29645	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	0	0	61	29547	25	10100	100	10000	100	10000	500	4277160	0	30018	30037	30037	28271	7	28740	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	0	1	1	1	718	16	0	0	29646	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	0	0	61	29547	25	10100	100	10000	100	10000	500	4277160	1	30018	30037	30037	28271	7	28740	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	0	1	1	1	717	16	0	0	29645	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	0	0	61	29547	25	10100	100	10000	100	10000	500	4277160	0	30018	30037	30037	28271	6	28740	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	0	1	1	1	718	16	0	0	29645	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	0	0	61	29547	25	10100	100	10000	100	10000	500	4277160	0	30018	30084	30037	28271	7	28741	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	0	1	1	1	717	16	0	0	29645	10000	100	30038	30038	30038	30038	30038
10204	30037	225	0	0	0	61	29547	25	10100	100	10000	100	10000	500	4277160	0	30018	30037	30037	28271	7	28741	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	1	3	0	1	1	1	718	16	0	0	29645	10000	100	30038	30038	30038	30038	30038
10204	30037	224	0	0	0	61	29547	25	10100	100	10000	100	10000	500	4277160	0	30018	30037	30037	28271	6	28740	10100	200	10008	200	20016	30037	30037	1	1	10201	100	99	100	100	10000	100	0	0	0	1	1	1	718	16	0	0	29645	10000	100	30038	30038	30038	30038	30038

1000 unrolls and 10 iterations

Result (median cycles for code): 3.0037

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

retire uop (01)	cycle (02)	03	1e	3f	4e	51	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)	60	69	6d	6e	map stall dispatch (70)	map rewind (75)	map stall (76)	dispatch uop (78)	map int uop (7c)	map simd uop (7e)	map int uop inputs (7f)	map simd uop inputs (81)	82	83	flush restart other nonspec (84)	85	inst all (8c)	inst branch (8d)	inst branch taken (90)	inst branch cond (94)	inst int alu (97)	inst simd alu (9a)	9f	l1d cache writeback (a8)	ac	cd	cf	d5	map dispatch bubble (d6)	dd	fetch restart (de)	e0	? simd retires (ee)	? int retires (ef)	f5	f6	f7	f8	fd
10024	30037	225	12	61	29547	25	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	29	6	0	640	3	16	2	2	29629	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	103	29547	25	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	45	0	662	2	16	2	2	29629	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	61	29547	25	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	51	9	0	640	2	16	2	2	29629	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	61	29547	25	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	43	9	0	640	2	16	2	2	29629	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	61	29547	25	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	51	6	0	640	2	16	2	2	29629	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	61	29547	25	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	54	9	0	640	2	16	2	2	29629	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	61	29547	25	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	36	0	0	640	2	16	2	2	29629	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	61	29547	44	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	1	21	0	640	2	16	2	2	29629	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	61	29547	25	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	0	84	0	640	2	16	2	2	29629	10000	10	30038	30038	30038	30038	30038
10024	30037	225	0	61	29547	25	10010	10	10000	10	10000	50	4277160	0	30018	30037	30037	28286	3	28767	10010	20	10000	20	20000	30037	30037	1	1	10021	10	9	10	10	10000	10	51	0	0	640	2	16	2	2	29629	10000	10	30038	30038	30038	30038	30038

Test 4: throughput

Count: 8

Code:

  movi v0.16b, 0
  sadalp v0.4h, v8.8b
  movi v1.16b, 0
  sadalp v1.4h, v8.8b
  movi v2.16b, 0
  sadalp v2.4h, v8.8b
  movi v3.16b, 0
  sadalp v3.4h, v8.8b
  movi v4.16b, 0
  sadalp v4.4h, v8.8b
  movi v5.16b, 0
  sadalp v5.4h, v8.8b
  movi v6.16b, 0
  sadalp v6.4h, v8.8b
  movi v7.16b, 0
  sadalp v7.4h, v8.8b

Initialization

  movi v8.16b, 9

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.2508

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

retire uop (01)	cycle (02)	03	mmu table walk data (08)	09	18	19	1e	1f	3f	51	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)	60	69	6d	6e	map rewind (75)	map stall (76)	dispatch uop (78)	map int uop (7c)	map simd uop (7e)	map int uop inputs (7f)	map simd uop inputs (81)	82	83	flush restart other nonspec (84)	85	inst all (8c)	inst branch (8d)	inst branch taken (90)	inst branch cond (94)	inst int alu (97)	inst simd alu (9a)	9f	l1d tlb miss (a1)	l1d cache writeback (a8)	a9	ac	c2	branch cond mispred nonspec (c5)	branch mispred nonspec (cb)	cd	cf	d5	map dispatch bubble (d6)	dd	e0	? simd retires (ee)	? int retires (ef)	f5	f6	f7	f8	fd
160204	20090	150	0	0	0	0	0	0	29	25	80116	100	80016	100	80028	500	640196	1	20045	20065	20065	6	12	80128	200	80028	200	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	0	0	3	0	1	1	1	10119	1	49	0	20062	160000	100	20066	20066	20066	20066	20066
160204	20065	151	0	0	0	0	0	0	29	25	80116	100	80016	100	80028	500	641060	1	20045	20065	20168	6	12	80128	200	80028	200	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	0	0	0	0	1	1	1	10119	0	16	0	20062	160000	100	20066	20066	20148	20066	20066
160204	20065	150	0	0	0	0	0	264	29	25	80116	100	80016	100	80028	500	640196	1	20045	20065	20065	6	12	80128	200	80028	200	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	0	0	6	0	1	1	1	10119	0	48	0	20062	160000	100	20066	20066	20066	20066	20066
160204	20146	150	0	0	0	0	0	0	29	25	80116	100	80016	100	80028	500	640196	1	20045	20065	20065	6	33	80128	200	80028	200	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	28	0	0	0	1	1	1	10119	0	16	0	20062	160000	100	20066	20066	20066	20066	20066
160204	20065	150	0	0	0	0	0	0	29	25	80116	100	80016	100	80028	500	640196	0	20045	20065	20065	6	12	80128	200	80028	200	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	0	0	0	4	1	1	1	10119	0	16	0	20062	160000	100	20066	20066	20066	20066	20066
160204	20065	151	0	0	0	0	0	0	29	25	80116	100	80016	100	80028	500	641036	0	20045	20065	20065	6	12	80128	200	80028	200	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	0	0	0	0	1	1	1	10147	0	16	0	20062	160000	100	20066	20066	20066	20148	20066
160204	20065	151	0	1	0	0	0	0	29	25	80116	100	80016	100	80028	500	640196	0	20045	20065	20065	6	12	80128	200	80028	200	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	0	0	0	0	1	1	1	10119	0	16	0	20062	160000	100	20066	20147	20066	20066	20066
160204	20065	151	0	0	0	0	0	0	29	25	80116	100	80016	100	80028	500	640196	0	20045	20065	20065	14	12	80128	200	80028	200	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	0	0	0	0	1	1	1	10119	0	16	0	20062	160000	100	20066	20066	20066	20066	20066
160204	20065	150	0	0	0	0	0	0	29	25	80116	100	80016	100	80028	500	640196	0	20045	20065	20065	13	12	80128	200	80028	202	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	0	0	0	0	1	1	1	10119	0	16	0	20062	160000	100	20066	20066	20066	20066	20066
160204	20065	150	0	0	0	0	0	0	29	25	80116	100	80016	100	80028	500	640196	0	20045	20065	20065	6	12	80128	200	80028	200	160056	20065	20065	1	1	160201	100	99	100	100	160000	100	0	0	0	0	0	1	1	1	10119	0	16	0	20062	160000	100	20066	20066	20066	20066	20066

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.2506

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

retire uop (01)	cycle (02)	03	3f	51	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)	5f	60	69	6d	6e	map rewind (75)	map stall (76)	dispatch uop (78)	map int uop (7c)	map simd uop (7e)	map int uop inputs (7f)	map simd uop inputs (81)	82	83	flush restart other nonspec (84)	85	inst all (8c)	inst branch (8d)	inst branch taken (90)	inst branch cond (94)	inst int alu (97)	inst simd alu (9a)	9f	l1d cache writeback (a8)	ac	branch cond mispred nonspec (c5)	cf	d0	d2	l1i cache miss demand (d3)	d5	map dispatch bubble (d6)	d9	da	db	dd	fetch restart (de)	e0	ea	? simd retires (ee)	? int retires (ef)	f5	f6	f7	f8	fd
160024	20073	150	51	25	80010	10	80000	10	80000	50	640000	0	1	20031	20046	20050	3	22	80010	20	80000	20	160000	20050	20050	1	1	160021	10	9	10	10	160000	10	0	0	0	10027	6	1	1	4	20	2	1	2	3	4	20043	15	160000	10	20055	20047	20047	20047	20047
160024	20046	150	55	25	80010	10	80000	10	80000	50	640000	1	0	20027	20046	20046	3	22	80010	20	80000	20	160000	20046	20046	1	1	160021	10	9	10	10	160000	10	0	0	0	10028	3	1	1	4	20	2	1	1	4	3	20043	15	160000	10	20055	20047	20047	20047	20047
160024	20046	150	45	25	80010	10	80000	10	80000	50	640000	1	0	20027	20046	20046	3	22	80010	20	80000	20	160000	20046	20046	1	1	160021	10	9	10	10	160000	10	1	0	0	10028	3	1	1	3	20	2	1	1	4	5	20043	15	160000	10	20055	20047	20047	20047	20047
160024	20046	150	45	25	80010	10	80000	10	80000	50	640000	1	0	20027	20046	20046	3	22	80010	20	80000	20	160000	20046	20046	1	1	160021	10	9	10	10	160000	10	0	0	0	10027	3	1	1	4	20	2	2	1	4	4	20043	15	160000	10	20067	20047	20047	20047	20047
160024	20046	150	45	25	80010	10	80000	10	80000	50	640000	1	1	20027	20046	20046	3	22	80010	20	80000	20	160000	20050	20046	1	1	160021	10	9	10	10	160000	10	1	3	0	10027	3	1	1	4	20	2	1	1	3	4	20047	30	160000	10	20055	20047	20047	20051	20047
160024	20046	150	45	25	80010	10	80000	10	80000	50	640000	1	0	20027	20046	20046	3	22	80010	20	80000	20	160000	20046	20050	1	1	160021	10	9	10	10	160000	10	0	21	0	10030	3	1	1	3	20	2	1	1	2	6	20043	15	160000	10	20059	20047	20047	20047	20047
160024	20050	150	45	25	80010	10	80000	10	80000	50	640000	1	0	20027	20046	20046	3	22	80010	20	80000	20	160000	20046	20046	1	1	160021	10	9	10	10	160000	10	0	0	0	10026	3	1	1	3	20	2	1	1	4	3	20043	15	160000	10	20055	20047	20047	20047	20047
160024	20046	150	45	25	80010	10	80000	10	80000	50	640000	1	1	20027	20046	20046	3	22	80010	20	80000	20	160000	20046	20046	1	1	160021	10	9	10	10	160000	10	0	0	0	10026	3	1	1	4	20	2	1	1	4	4	20043	15	160000	10	20067	20047	20047	20047	20047
160024	20046	150	45	75	80010	10	80000	10	80000	50	640000	1	1	20027	20046	20046	3	22	80010	20	80000	20	160000	20046	20046	1	1	160021	10	9	10	10	160000	10	0	0	0	10028	6	1	1	4	20	2	2	1	5	3	20043	15	160000	10	20067	20047	20047	20047	20047
160024	20046	150	45	25	80010	10	80000	10	80000	50	640000	0	1	20031	20046	20046	3	22	80010	20	80000	20	160000	20046	20046	1	1	160021	10	9	10	10	160000	10	0	0	0	10027	3	1	2	3	20	2	1	1	3	5	20043	15	160000	10	20059	20047	20047	20047	20047

Test 5: throughput

Count: 16

Code:

  sadalp v0.4h, v16.8b
  sadalp v1.4h, v16.8b
  sadalp v2.4h, v16.8b
  sadalp v3.4h, v16.8b
  sadalp v4.4h, v16.8b
  sadalp v5.4h, v16.8b
  sadalp v6.4h, v16.8b
  sadalp v7.4h, v16.8b
  sadalp v8.4h, v16.8b
  sadalp v9.4h, v16.8b
  sadalp v10.4h, v16.8b
  sadalp v11.4h, v16.8b
  sadalp v12.4h, v16.8b
  sadalp v13.4h, v16.8b
  sadalp v14.4h, v16.8b
  sadalp v15.4h, v16.8b

Initialization

  movi v16.16b, 17

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.2502

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

retire uop (01)	cycle (02)	03	09	18	1e	37	3a	3f	51	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)	60	69	6b	6d	6e	map stall dispatch (70)	map rewind (75)	map stall (76)	dispatch uop (78)	map int uop (7c)	map simd uop (7e)	map int uop inputs (7f)	map simd uop inputs (81)	82	83	flush restart other nonspec (84)	85	inst all (8c)	inst branch (8d)	inst branch taken (90)	inst branch cond (94)	inst int alu (97)	inst simd alu (9a)	9f	ld unit uop (a6)	l1d cache writeback (a8)	a9	ac	branch cond mispred nonspec (c5)	branch mispred nonspec (cb)	cd	cf	d5	map dispatch bubble (d6)	dd	fetch restart (de)	e0	? simd retires (ee)	? int retires (ef)	f5	f6	f7	f8	fd
160204	40059	300	0	0	9	0	0	30	25	160117	100	160008	100	160020	500	2399214	1	40020	0	40039	40048	19977	6	19999	160120	200	160032	200	320064	40039	40048	1	1	160201	100	99	100	100	160000	100	0	0	0	0	1	1	1	10118	2	16	1	1	40036	160000	100	40040	40049	40049	40040	40040
160204	40039	299	0	0	0	1	0	39	25	160117	100	160017	100	160020	500	1280132	1	40021	0	40040	40049	19977	6	19999	160120	200	160032	200	320064	40039	40039	1	1	160201	100	99	100	100	160000	100	0	0	0	0	1	1	1	10118	1	16	1	1	40037	160000	100	40041	40049	40049	40040	40041
160204	40040	301	0	2	0	0	0	39	25	160108	100	160105	100	160020	500	1280132	1	40021	0	40049	40049	19977	6	19990	160120	200	160032	200	320064	40040	40049	1	1	160201	100	99	100	100	160000	100	0	0	0	0	1	1	1	10118	1	16	1	1	40046	160000	100	40040	40040	40040	40050	40041
160204	40040	300	0	0	0	0	0	40	25	160108	100	160017	100	160020	500	1320131	1	40021	0	40049	40039	19977	6	19990	160120	200	160032	200	320064	40039	40040	1	1	160201	100	99	100	100	160000	100	0	0	0	0	1	1	1	10118	1	16	1	1	40036	160000	100	40050	40049	40255	40040	40049
160204	40048	300	1	0	0	1	0	30	25	160109	100	160009	100	160020	500	1280132	1	40030	0	40040	40040	19977	6	19990	160120	200	160032	200	320064	40039	40040	1	1	160201	100	99	100	100	160000	100	0	0	0	0	1	1	1	10118	1	16	1	1	40036	160000	100	40041	40040	40050	40040	40040
160204	40049	299	0	0	0	17	0	30	25	160117	100	160017	100	160020	500	2399159	1	40020	0	40049	40039	19977	6	19999	160120	200	160032	200	320064	40049	40039	1	1	160201	100	99	100	100	160000	100	0	0	0	0	1	1	1	10118	1	16	1	1	40046	160000	100	40049	40049	40040	40040	40050
160204	40039	300	0	0	0	0	0	40	25	160109	100	160009	100	160020	500	2399131	0	40020	0	40040	40040	19977	6	19999	160120	200	160032	200	320064	40039	40040	1	1	160201	100	99	100	100	160000	100	0	0	0	0	1	1	1	10118	2	16	1	1	40037	160000	100	40049	40049	40040	40040	40050
160204	40049	300	0	0	0	1	0	40	25	160108	100	160017	100	160020	500	1320132	1	40021	0	40049	40039	19977	6	19999	160120	200	160032	200	320064	40049	40039	1	1	160201	100	99	100	100	160000	100	0	0	0	0	1	1	1	10118	1	16	1	1	40036	160000	100	40040	40040	40041	40041	40040
160204	40039	300	0	0	0	1	0	230	25	160109	100	160008	100	160020	500	1320129	1	40029	0	40040	40048	19977	6	19991	160120	200	160032	200	320064	40039	40040	1	1	160201	100	99	100	100	160000	100	0	56	0	3	1	1	1	10118	1	16	1	1	40045	160000	100	40049	40040	40040	40041	40049
160204	40048	300	0	0	0	1	0	31	25	160118	100	160018	100	160020	500	1280132	1	40020	0	40049	40039	19977	6	19990	160120	200	160032	200	320064	40049	40039	1	1	160201	100	99	100	100	160000	100	0	0	0	0	1	1	1	10118	1	16	1	1	40045	160000	100	40040	40040	40041	40050	40041

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.2502

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

retire uop (01)	cycle (02)	03	mmu table walk data (08)	l2 tlb miss data (0b)	18	19	1e	1f	37	3f	4e	51	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)	5f	60	61	69	6d	6e	map stall dispatch (70)	map rewind (75)	map stall (76)	dispatch uop (78)	map int uop (7c)	map simd uop (7e)	map int uop inputs (7f)	map simd uop inputs (81)	82	83	flush restart other nonspec (84)	85	inst all (8c)	inst branch (8d)	inst branch taken (90)	inst branch cond (94)	inst int alu (97)	inst simd alu (9a)	9f	l1d tlb miss (a1)	l1d cache writeback (a8)	ac	c2	cf	d0	d2	l1i cache miss demand (d3)	d5	map dispatch bubble (d6)	d9	da	db	dd	fetch restart (de)	e0	ea	eb	? simd retires (ee)	? int retires (ef)	f5	f6	f7	f8	fd
160024	40050	300	0	0	0	0	615	0	0	55	0	25	160028	10	160000	10	160000	50	2398999	1	1	5	40029	40039	40039	19996	3	20020	160010	20	160000	20	320000	40048	40040	1	1	160021	10	9	10	10	160000	10	0	0	0	0	10022	8	2	1	6	16	2	1	1	4	5	40036	15	5	160000	10	40050	40049	40049	40040	40050
160024	40049	300	0	0	0	0	0	0	0	52	0	25	160010	10	160017	10	160000	50	2399027	1	1	0	40021	40039	40039	19996	3	20020	160010	20	160000	20	320000	40049	40039	1	1	160021	10	9	10	10	160000	10	0	0	0	0	10024	11	5	2	5	16	4	2	2	5	5	40036	30	10	160000	10	40049	40040	40049	40040	40041
160024	40049	299	0	0	0	0	0	0	0	55	0	25	160011	10	160001	10	160000	50	2399027	1	1	5	40021	40039	40039	19996	3	20028	160010	20	160000	20	320000	40039	40048	1	1	160021	10	9	10	10	160000	10	0	0	0	0	10022	8	2	1	5	16	2	1	1	5	4	40046	15	5	160000	10	40050	40040	40049	40040	40040
160024	40039	299	0	0	0	0	0	0	0	55	0	25	160027	10	160001	10	160000	50	2398999	1	0	5	40020	40048	40102	19996	3	20019	160010	20	160000	20	320000	40039	40049	1	1	160021	10	9	10	10	160000	10	0	0	237	0	10024	6	5	2	4	16	4	2	2	5	4	40037	30	5	160000	10	40040	40040	40049	40050	40040
160024	40040	300	0	0	0	0	0	0	0	47	0	25	160010	10	160000	10	160000	50	2399027	1	1	5	40029	40039	40039	19996	3	20019	160010	20	160000	20	320000	40039	40039	1	1	160021	10	9	10	10	160000	10	0	0	0	0	10022	8	4	1	4	16	2	1	1	5	4	40045	15	5	160000	10	40040	40049	40040	40049	40040
160024	40039	299	0	0	0	0	0	0	0	56	0	25	160011	10	160000	10	160000	50	1280000	1	1	5	40020	40039	40049	19996	3	20028	160010	20	160000	20	320000	40048	40039	1	1	160021	10	9	10	10	160000	10	0	0	935	0	10022	8	4	1	5	16	2	1	1	6	5	40045	15	6	160000	10	40040	40040	40049	40040	40172
160024	40060	300	0	0	0	0	0	0	0	433	0	25	160010	10	160017	10	160000	50	2399027	1	1	0	40114	40040	40039	19996	3	20019	160010	20	160000	20	320000	40051	40049	1	1	160021	10	9	10	10	160000	10	0	0	0	4	10044	8	4	1	4	16	2	1	1	4	5	40045	15	5	160000	10	40040	40437	40041	40050	40040
160024	40040	300	0	0	2	0	0	0	0	56	0	25	160010	10	160001	10	160000	50	2399027	1	1	5	40030	40040	40048	19996	3	20019	160010	20	160000	20	320000	40039	40049	1	1	160021	10	9	10	10	160000	10	0	0	0	0	10022	3	4	1	5	41	2	1	1	5	4	40036	15	5	160000	10	40041	40040	40050	40040	40050
160024	40049	300	0	0	0	0	0	0	17	56	0	25	160027	10	160017	10	160000	50	2398999	1	0	5	40029	40040	40048	19996	3	20019	160010	20	160000	20	320000	40039	40048	1	1	160021	10	9	10	10	160000	10	0	0	129	0	10046	8	4	1	6	16	2	1	1	5	5	40036	15	5	160000	10	40052	40050	40041	40040	40050
160024	40040	300	0	0	0	0	0	0	17	56	0	25	160010	10	160000	10	160000	50	2399027	1	0	5	40020	40040	40049	19996	3	20029	160010	20	160000	20	320000	40049	40040	1	1	160021	10	9	10	10	160000	10	0	0	0	0	10022	8	4	1	5	16	2	1	1	5	4	40046	15	5	160000	10	40049	40049	40040	40049	40040

SADALP (vector, 4H)

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: throughput

100 unrolls and 100 iterations

1000 unrolls and 10 iterations

Test 5: throughput

100 unrolls and 100 iterations

1000 unrolls and 10 iterations