Apple M1 Microarchitecture Research by Dougall Johnson

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

FNMSUB (scalar, H)

Test 1: uops

Code:

  fnmsub h0, h0, h1, h2
  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)031e1f3f4e51inst issue (52)~issue fp/simd (54)~dispatch fp/simd (57)huge thing fp/simd (5b)60696d6edispatch 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
10044037300052734072510001000100053190804018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730006134072510001000100053190804018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730006134072510001000100053190814018403740373258338951000100030004037403711100110002073116113473100040384038403840384038
1004403730006134072510001000100053190814018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730006134072510001000100053190804018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730006134072510001000100053190814018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730006134072510001000100053190814018403740373258338951000100030004037403711100110002073116113473100040384038403840384038
1004403730006134072510001000100053190804018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403730006134072510001000100053190804018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
1004403731006134072510001000100053190814018403740373258338951000100030004037403711100110000073116113473100040384038403840384038

Test 2: Latency 1->2

Code:

  fnmsub h0, h0, h1, h2
  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)03091e3a3f4e51inst 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)72scheduler 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
102044003730020061394072510100100100001001000050057069080400184003740037381080338745101002001000020030000400374003711102011009910010010000100046071012162239479100001004008540038400384003840038
102044003730000061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038
102044003729900061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000071012163239479100001004003840038400384003840038
102044003730000061394072510100100100001001000050057069080400184003740037381080338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038
102044003730000061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038
1020440037300000613940725101001001000010010000500570690804001840037400373810803387451010020010000200300004003740037111020110099100100100001000250071012162239479100001004003840038400384003840038
102044003730000061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100020071012162239479100001004003840038400384003840038
1020440037300012061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038
102044003730000061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038
102044003729900061394072510100100100001001000050057069080400184003740037381087338745101002001000020030000400374003711102011009910010010000100000071012162239479100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)03080b18191e3f4e51inst 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)9fa0a6a8a9acbranch mispredict (cb)cfd5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
10024400373000000025239407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000006402162239473010000104003840038400384003840038
1002440037300000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000006402162239512010000104003840038400384003840038
10024400373000000066639407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000006402162239473010000104003840038400384003840038
1002440037300000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000006402162239473010000104003840038400384003840038
100244003730000001229439407251001010100001010000505706908040018400374003738130338767100102210000203000040037400371110021109101010000100000006402162239473010000104003840038400384003840038
10024400372990000216139407251001010100001010000505706908040018400374003738130338767100102210000203000040037400371110021109101010000100000006402162239473010000104003840038400384003840038
10024400373000000016639407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000006402162239473010000104003840038400384003840038
10024400373000000040639407251001010100001010000505706908040018400374003738135338767100102010000203000040037400371110021109101010000100000006402162239473010000104003840038400384003840038
1002440037299000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100010606402162239473010000104003840038400384003840038
1002440037300000006139407251001010100001010000505706908040018400374003738130338767100102010160203000040037400371110021109101010000100000006402162239600010000104003840038400384003840038

Test 3: Latency 1->3

Code:

  fnmsub h0, h1, h0, h2
  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)72scheduler 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)9fa8a9acc5branch mispredict (cb)cdcfd0d5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
102044003730057061394072510100100100001001000050057069081400184003740037381150638740101002001000820030024400374003711102011009910010010000100000111717001600394900100001004003840038400384003840038
1020440037300120161394072510100100100001001000050057069081400184003740037381150338745101002001000020030000400374003711102011009910010010000100000000710121622394790100001004003840038400384003840038
1020440037299468061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000000710121622394790100001004003840038400384003840038
10204400372995100726394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000000712121622394790100001004003840038400384003840038
1020440037300546061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000000710121622394790100001004003840038400384003840038
1020440037299609061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000000710121622394790100001004003840038400384003840038
1020440037300513061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000000710121622394790100001004003840038400384003840038
10204400372990061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000000710121622394790100001004003840038400384003840038
102044003730000726394072510100100100001001000050057069081400184003740037381080338745101002001016620030000400374003711102011009910010010000100000000710131622394790100001004003840038400384003840038
10204400373000061394072510100100100001001000050057069081400184003740037381080338745101002001000020030000400374003711102011009910010010000100000000710121622394790100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)03071e3a3f4e51inst 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)60696b6d6edispatch 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
10024400372990002513940725100101010000101000050570690804001804003740037381303387671001020100002030000400374003711100211091010100001000640416333947310000104003840038400384003840038
1002440037299000613940725100101010000101000050570690814001804003740037381303387671001020100002030000400374003711100211091010100001000640316333947310000104003840038400384003840038
100244003729911201723940725100181010000101000050570690814001804003740037381303387671001020100002030000400374003711100211091010100001000640316333947310000104003840038400384003840038
1002440037300000613940725100101010000101000050570690814001804003740037381303387671001020100002030000400374003711100211091010100001000640316333947310000104003840038400384003840038
1002440037300000613940725100101010000101000050570690804001804003740037381303387671001020100002030000400374003711100211091010100001000640316333947310000104003840038400384003840038
1002440037300072020453940725100101010000101000050570690804001804003740037381303387671001020100002030000400374003711100211091010100001000640316333947310000104003840038400384003840038
1002440037300000613940725100101010000101000050570690814001804003740037381303387671001020100002030000400374003711100211091010100001000640316333947310000104003840038400384003840038
1002440037300000613940725100101010000101000050570690814001804003740037381303387671001020100002030000400374003711100211091010100001000640316343947310000104003840038400384003840038
1002440037299000613940725100101010000101000050570690814001804003740037381303387671001020100002030000400374003711100211091010100001000640316333947310000104003840038400384003840038
1002440037300000613940725100101010000101000050570690804001804003740037381303387671001020100002030000400374003711100211091010100001000640316333947310000104003840038400384003840038

Test 4: Latency 1->4

Code:

  fnmsub h0, h1, h2, h0
  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)9fa8acc5branch mispredict (cb)cdcfd0d5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1020440037300006139407251010010010000100100005005706908400184003740037381157387411010020010008200300244003740037111020110099100100100001000011171700160139489100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381156387401010020010008200300244003740037111020110099100100100001000011171700160039489100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162339479100001004003840038400384003840038
1020440037300906139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162339479100001004003840038400384003840038
10204400373000015639407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162239479100001004003840038400384003840038
102044003730000613940725101001001000010010000500570690840018400374003738108338745101002001000020030000400374003711102011009910010010000100016500071012162339479100001004003840038400384003840038
1020440037299006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162239479100001004003840038400384003840038
1020440037300006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162239479100001004003840038400384003840038
1020440037299006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071012162339479100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)031e1f3f4e51inst 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)9faccfd5d6ddinst fetch restart (de)e0eb? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
10024400372990061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001006402162239473110000104003840038400384003840038
100244003730000726394072510010101000010100005057069080400184003740037381303387671001020100002030000400374003711100211091010100001006402162239473010000104007140038400384003840038
10024400373003061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001006402162239473010000104003840038400384003840038
10024400373000061394072510010101000010100005057069080400184003740037381303387671001020100002030000400374003711100211091010100001006402162339473010000104003840038400384003840038
10024400373000061394072510010101000010100005057069080400184003740037381303387671001020100002030000400374003711100211091010100001006402162239473010000104003840038400384003840038
10024400373000061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001006403162339473010000104003840038400384003840038
10024400373000061394072510010101000010100005057069080400184003740037381303387671001020100002030000400374003711100211091010100001006403162239473010000104003840038400854003840038
10024400372990061394072510010101000010100005057069080400184003740037381303387671001020100002030000400374003711100211091010100001006402162239473010000104003840038400384003840038
10024400372990061394072510010101000010100005057069080400184003740037381303387671001020100002030000400374003711100211091010100001006402162239473010000104003840038400384003840038
10024400372990061394072510010101000010100005057069080400184003740037381303387671001020100002030000400374003711100211091010100001006402162339473010000104003840038400384003840038

Test 5: throughput

Count: 8

Code:

  fnmsub h0, h8, h9, h10
  fnmsub h1, h8, h9, h10
  fnmsub h2, h8, h9, h10
  fnmsub h3, h8, h9, h10
  fnmsub h4, h8, h9, h10
  fnmsub h5, h8, h9, h10
  fnmsub h6, h8, h9, h10
  fnmsub h7, h8, h9, h10
  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)03080b18191e1f3f51inst 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)9fa0a1a6a7a8a9acc2cfd5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
8020420059150000048042258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010000000000511011611200370800001002004120041200412004120041
8020420040150000030517258010010080000100800005006400001200212004020040997339998801002008000020024000020040200401180201100991001008000010000000000511011611200370800001002004120041200412004120041
802042004015000000042258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010000000000511011611200370800001002010220041200412004120041
802042004015000000042258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010000000000511011611200370800001002004120041200412004120041
802042004015000000042258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010000000000511011611200370800001002004120041200412004120041
802042004015000000042258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010000000000511011611200370800001002004120041200412004120041
802042004015000000042258010010080000100800005006400001200212004020040997339998801002008000020024000020040200401180201100991001008000010000000000511011611200370800001002004120041200412004120041
802042004015000000042258010010080000100800005006400000200632004020040997339998801002008000020024000020040200401180201100991001008000010000000030511011611200370800001002004120041200412004120090
802042004015000006042258010010080000100800005006400001200212004020040997339998801002008000020024000020040200401180201100991001008000010000000000511011611200370800001002004120041200412004120041
802042004015000000042258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010000000000511011611200370800001002004120041200412004120041

1000 unrolls and 10 iterations

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

retire (01)cycle (02)03040b181e1f3f51inst 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)9fa1a6a7a8acbranch mispredict (cb)cfd5d6ddinst fetch restart (de)e0eb? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
80024200501500000041258001010800001080000506400001200212004020040999631002080010208000020240000200402004011800211091010800001000000050202516252520037080000102004120041200412004120041
80024200401500000041258001010800001080000506400001200212004020040999631002080010208000020240000200402004011800211091010800001000000050202516202520037080000102004120041200412004120041
8002420040150000285041258001010800001080000506400001200212004020040999631002080010208000020240000200402004011800211091010800001000000050202516122720037080000102004120041200412004120041
800242004015000000412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010000030502024162525200993880000102004120041200412004120041
800242004015000000412580010108000010800005064000002002120040200409996310020800102080000202400002004020040118002110910108000010000000502020172213200372280000102004120041200412004120041
800242004015000000412580010108000010800005064000002002120040200409996310020800102080000202400002004020040118002110910108000010000000502025162525200373080000102004120041200412004120041
800242004015000000412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010000100502025162424200373080000102004120041200412004120041
800242004015000000412580010108000010800005064000002002120040200409996310020800102080000202400002004020040118002110910108000010000090502024162424200373080000102004120041200412004120041
800242004015000000412580010108000010800005064000012002120040200409996310020800102080000202400002004020040118002110910108000010000000502015162512200373580000102004120041200412004120041
800242004015000000412580010108000010800005064000002002120040200409996310020800102080000202400002004020040118002110910108000010000000502025162525200373880000102004120041200412004120041