Apple M1 Microarchitecture Research by Dougall Johnson

Firestorm: Overview | Base Instructions | SIMD and FP Instructions
Icestorm:  Overview | Base Instructions | SIMD and FP Instructions

FMSUB (scalar, S)

Test 1: uops

Code:

  fmsub s0, s0, s1, s2
  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: 0.000

Load/store unit issues: 0.000

SIMD/FP unit issues: 1.000

retire (01)cycle (02)031e3f4e51inst issue (52)~issue fp/simd (54)~dispatch fp/simd (57)huge thing fp/simd (5b)60696b6d6edispatch stall (70)scheduler rewind (75)scheduler stall (76)~dispatch op (78)~map op fp/simd (7e)~map lookup fp/simd (81)8283pipeline redirect (84)85inst all (8c)inst neon or fp (9a)a8accfd5d6ddinst fetch restart (de)e0? fp/simd (ee)f5f6f7f8fd
1004403730361340725100010001000531908140180403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730061340725100010001000531908140180403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730061340725100010001000531908140180403740373258338951000100030004037403711100110000073416123473100040384038403840384038
1004403730061340725100010001000531908140180403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730061340725100010001000531908140180403740373258338951000100030004037403711100110000073116113473100040384038403840384038
10044037300170340725100010001000531908140180403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730061340725100010001000531908140180403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730061340725100010001000531908140180403740373258338951000100030004037403711100110002073116113473100040384038403840384038
1004403730061340725100010001000531908140180403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403731061340725100010001000531908140180403740373258338951000100030004037403711100110000073116113473100040384038403840384038

Test 2: Latency 1->2

Code:

  fmsub s0, s0, s1, s2
  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): 4.0037

retire (01)cycle (02)031e3a3f4e51inst issue (52)~issue int (53)~issue fp/simd (54)~dispatch int (56)~dispatch fp/simd (57)huge thing int (59)huge thing fp/simd (5b)60696d6edispatch stall (70)scheduler rewind (75)scheduler stall (76)~dispatch op (78)~map op int (7c)~map op fp/simd (7e)~map lookup int (7f)~map lookup fp/simd (81)8283pipeline redirect (84)85inst all (8c)inst branch (8d)inst branch taken (90)inst b.cc (94)inst integer (97)inst neon or fp (9a)9fa6a8accdcfd0d5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1020440037299006139407251010010010000100100005005706908040018400374003738108338745101002001000020030000400374003711102011009910010010000100040071012163239479100001004003840038400384003840038
10204400373000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000360071013162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908040018400374003738108338745101002001000020030000400374003711102011009910010010000100020071012163239479100001004003840038400384003840038
1020440084300006139407251010010010000100100005005706908040018400374003738108338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908040018400374003738108338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908040018400374003738108338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908040018400374003738108338745101002001000020030000400374003711102011009910010010000100000171012162239479100001004003840038400384003840038
102044003730072061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000220071012162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908140018400374003738108338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038
1020440037299006139407251010010010000100100005005706908040018400374003738108338745101002001000020030000400374003721102011009910010010000100000071012162239479100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)0307080a18191e1f3f4e51inst issue (52)~issue int (53)~issue fp/simd (54)~dispatch int (56)~dispatch fp/simd (57)huge thing int (59)huge thing fp/simd (5b)696b6d6edispatch stall (70)scheduler rewind (75)scheduler stall (76)~dispatch op (78)~map op int (7c)~map op fp/simd (7e)~map lookup int (7f)~map lookup fp/simd (81)8283pipeline redirect (84)85inst all (8c)inst branch (8d)inst branch taken (90)inst b.cc (94)inst integer (97)inst neon or fp (9a)9fa0a1a6a8acc2branch mispredict (cb)cfd5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1002440037299000000061394072510010101000010100005057069084001804003740037381303387671001020100002030000400374003711100211091010100001000040006403163339473010000104003840038400384003840038
1002440037299000000061394072510010101000010100005057069084001804003740037381303387671001020100002030000400374003711100211091010100001000000006403163339473010000104003840038400384003840038
10024400373000000000613940744100161010012121014850570690840018040037400373813073876710010201000020300004003740037111002110910101000010222229740207874804439799010000104022640453404654041740454
1002440464303200910118879257513933512010064121005410111849357180764033304046440509381762338935113422211136243438640468405081011002110910101000010000432360207844885439797110000104046040277405094051140464
100244041430341199120070489133931721810070121006012114805057222644001804003740037381303387671001020100002030000400374003711100211091010100001000000006403163339473010000104003840038400384003840038
1002440037300000000061394072510010101000010100005057069084001804003740037381303387671001020100002030000400374003711100211091010100001000000006403163339473010000104003840038400384003840038
10024400373000000000232394072510010101000010100005057069084001804003740037381303387671001020100002030000400374003711100211091010100001000000016403163339473010000104003840038400384003840038
1002440037300000000061394072510010101000010100005057069084001804003740037381303387671001020100002030000400374003711100211091010100001000000006403163339473010000104003840038400384003840038
1002440037300000000084394072510010101000010100005057069084001804003740037381303387671001020100002030000400374003711100211091010100001002210006403163339473010000104003840038400384003840038
1002440037300000000061394072510010101000010100005057069084001804007840037381303387671001020100002030000400374003711100211091010100001000000006403163339473010000104003840038400384003840038

Test 3: Latency 1->3

Code:

  fmsub s0, s1, s0, s2
  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): 4.0037

retire (01)cycle (02)031e3a3f4e51inst issue (52)~issue int (53)~issue fp/simd (54)~dispatch int (56)~dispatch fp/simd (57)huge thing int (59)huge thing fp/simd (5b)696d6edispatch stall (70)scheduler rewind (75)scheduler stall (76)~dispatch op (78)~map op int (7c)~map op fp/simd (7e)~map lookup int (7f)~map lookup fp/simd (81)8283pipeline redirect (84)85inst all (8c)inst branch (8d)inst branch taken (90)inst b.cc (94)inst integer (97)inst neon or fp (9a)9fa0a1a8acc2cfd0d5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071013163239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071212162339479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000009071012162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162239479100001004003840038400384003840038
10204400373000033539407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162339479100001004003840038400384003840038
1020440037299006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071013162239479100001004003840038400384003840038
1020440037299006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000020071012162339479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000240071012162239479100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)031e3f4e51inst issue (52)~issue int (53)~issue fp/simd (54)~dispatch int (56)~dispatch fp/simd (57)huge thing int (59)huge thing fp/simd (5b)60696d6edispatch stall (70)scheduler rewind (75)scheduler stall (76)~dispatch op (78)~map op int (7c)~map op fp/simd (7e)~map lookup int (7f)~map lookup fp/simd (81)8283pipeline redirect (84)85inst all (8c)inst branch (8d)inst branch taken (90)inst b.cc (94)inst integer (97)inst neon or fp (9a)9fa0a1a6a7a8acc2cfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
100244003730006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000000640216223947310000104003840038400384003840084
1002440084300010339407251001010100001010000505706908040018400374008538130338767100102010160203000040037400371110021109101010000100000000640216223947310000104003840038400384003840038
100244003730006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000000640216223947310000104003840038400384003840038
100244003730006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000000640216223947310000104003840038400384003840038
100244003729906139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000000640216223947310000104003840038400384003840038
10024400373005496139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000000640216223947310000104003840038400384003840038
100244003729906139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000060640216223947310000104003840038400384003840038
1002440037300025139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000000640216223947310000104003840038400384003840038
100244003730006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000000640216323947310000104003840038400384003840038
100244003730006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000000640216223947310000104003840038400384003840038

Test 4: Latency 1->4

Code:

  fmsub s0, s1, s2, s0
  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): 4.0037

retire (01)cycle (02)0308181e1f3a3f4e51inst issue (52)~issue int (53)~issue fp/simd (54)~dispatch int (56)~dispatch fp/simd (57)huge thing int (59)huge thing fp/simd (5b)60696d6edispatch stall (70)scheduler rewind (75)scheduler stall (76)~dispatch op (78)~map op int (7c)~map op fp/simd (7e)~map lookup int (7f)~map lookup fp/simd (81)8283pipeline redirect (84)85inst all (8c)inst branch (8d)inst branch taken (90)inst b.cc (94)inst integer (97)inst neon or fp (9a)9fa1a8a9accfd0d5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
102044003730000000726394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710021622394790100001004003840038400384003840038
10204400373000000061394072510100100100001001000052257069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840083
10204400373000000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710122522394790100001004003840038400384003840038
10204400373000000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
10204400372990000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
102044003730000001251394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121623394790100001004003840038400384003840038
10204400373100000061394072510100100100001001000050057069081400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
10204400373000000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121632394790100001004003840038400384003840038
10204400373000000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
10204400373000000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)03081e3a3f4e51inst issue (52)~issue int (53)~issue fp/simd (54)~dispatch int (56)~dispatch fp/simd (57)huge thing int (59)huge thing fp/simd (5b)60696d6edispatch stall (70)scheduler rewind (75)scheduler stall (76)~dispatch op (78)~map op int (7c)~map op fp/simd (7e)~map lookup int (7f)~map lookup fp/simd (81)8283pipeline redirect (84)85inst all (8c)inst branch (8d)inst branch taken (90)inst b.cc (94)inst integer (97)inst neon or fp (9a)9fa8a9acbranch mispredict (cb)cfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
100244003730000072639407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908140018400374003738130338767100102010175203000040037400375110021109101010000100000640224223947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110022109101010000100000640216223947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908040018400374008438130338767100102210000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040224400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216323947310000104003840038400384003840038

Test 5: throughput

Count: 8

Code:

  fmsub s0, s8, s9, s10
  fmsub s1, s8, s9, s10
  fmsub s2, s8, s9, s10
  fmsub s3, s8, s9, s10
  fmsub s4, s8, s9, s10
  fmsub s5, s8, s9, s10
  fmsub s6, s8, s9, s10
  fmsub s7, s8, s9, s10
  movi v8.16b, 9
  movi v9.16b, 10
  movi v10.16b, 11

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.2505

retire (01)cycle (02)030a1e3f51inst issue (52)~issue int (53)~issue fp/simd (54)~dispatch int (56)~dispatch fp/simd (57)huge thing int (59)huge thing fp/simd (5b)696d6edispatch stall (70)scheduler rewind (75)scheduler stall (76)~dispatch op (78)~map op int (7c)~map op fp/simd (7e)~map lookup int (7f)~map lookup fp/simd (81)8283pipeline redirect (84)85inst all (8c)inst branch (8d)inst branch taken (90)inst b.cc (94)inst integer (97)inst neon or fp (9a)9fa8a9cfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
802042005915000422580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000051102162220037800001002004120041200412004120041
802042004015000422580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000051102162220037800001002004120041200412004120041
802042004015000422580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000051102162220037800001002004120041200412004120041
8020420040150005882580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000051102162220037800001002004120041200412004120041
8020420040150006122580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000251102162220037800001002004120041200412004120041
8020420040150006602580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000051102162220037800001002004120041200412004120041
802042004015000422580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000051102162220037800001002004120041200412004120041
802042004015000422580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000051102162220037800001002004120041200412004120041
8020420040150012422580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000051102162220037800001002004120041200412004120041
8020420040150109162580100100800001008000050064000020021200402004099733999880100200800002002400002004020040118020110099100100800001000051102162220037800001002004120041200412004120041

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.2505

retire (01)cycle (02)030b1e1f3a3f51inst issue (52)~issue int (53)~issue fp/simd (54)~dispatch int (56)~dispatch fp/simd (57)huge thing int (59)huge thing fp/simd (5b)60696d6edispatch stall (70)scheduler rewind (75)scheduler stall (76)~dispatch op (78)~map op int (7c)~map op fp/simd (7e)~map lookup int (7f)~map lookup fp/simd (81)8283pipeline redirect (84)85inst all (8c)inst branch (8d)inst branch taken (90)inst b.cc (94)inst integer (97)inst neon or fp (9a)9fa8accfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
800242004015004200412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020516442003780000102004120041200412004120041
800242004015102400412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020316332003780000102004120041200412004120041
8002420040150021600412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020316232003780000102004120041200412004120041
80024200401500000412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020316332003780000102004120041200412004120041
8002420040150013500412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020216332003780000102004120041200412004120041
80024200401500000412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020316332003780000102004120041200412004120041
80024200401500000412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020316332003780000102004120041200412004120041
80024200401500000412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020316332003780000102004120041200412004120041
80024200401500000412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020316232003780000102004120041200412004120041
80024200401500000412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010005020316332003780000102004120041200412004120041