Apple M1 Microarchitecture Research by Dougall Johnson

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

FNMADD (scalar, S)

Test 1: uops

Code:

  fnmadd 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)696d6edispatch 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
100440373002513407251000100010005319084018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
10044037300613407251000100010005319084018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
100440373001053407251000100010005319084018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
100440373006134072510001000100053190840184037403732583389510001000300040374037111001100005773116113473100040384038403840384038
10044037300613407251000100010005319084018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
10044037300613407251000100010005319084018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
10044037300613407251000100010005319084018403740373258338951000100030004037403711100110001073116113473100040384038403840384038
10044037300613407251000100010005319084018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
10044037300613407251000100010005319084018403740373258338951000100030004037403711100110000073116113473100040384038403840384038
10044037300613407251000100010005319084018403740373258338951000100030004037403711100110000073116113473100040384038403840384038

Test 2: Latency 1->2

Code:

  fnmadd 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)03080b1e3a3f4e51inst 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)9fa8acbranch mispredict (cb)cfd0d5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1020440037300004506139407251010010010000100100005005706908040018400374003738108033874510100200100002003000040037400371110201100991001001000010000071003162239479100001004003840038400384003840038
1020440037300003306139407251010010010000100100005005706908140018400374003738108033874510100200100002003000040037400371110202100991001001000010000071002162239479100001004003840038400384003840038
1020440037300003006139407251010010010000100100005005706908140018400374003738108033874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
1020440037300002106139407251010010010000100100005005706908140018400374003738108033874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
1020440037300002706139407251010010010000100100005005706908040018400374003738108033874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
1020440037300003606139407251010010010000100100005005706908040018400374003738108033874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
1020440037300003306139407251010010010000100100005005706908040018400374003738108033874510100200100002003000040037400371110201100991001001000010000071013163239479100001004003840038400384003840038
1020440037300003306139407251010010010000100100005005706908140018400374003738108033874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
10204400373000023106139407251010010010000100100005005706908140018400374003738108033874510100200100002003000040037400371110201100991001001000010000071012162339479100001004003840038400384003840038
102044003730000006139407251010010010000100100005005706908040018400854003738108033874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038

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)9fa1a8acc5cfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1002440037300061394072510010101000010100005057069080400184003740037381303387671001020100002030000400374003711100211091010100001005400640516323947310000104003840038400384003840038
100244003730006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100030640216223947310000104003840038400384003840038
100244003730006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
1002440037299972639407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373003661394072510010101002410100005057069080400184003740037381303387671001020100002030000400374003711100211091010100001026300640216223947310000104003840038400384003840038
100244003730006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
100244003730006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216333947310000104003840038400384003840038
100244003729906139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100190640316233947310000104003840038400384003840038
100244003730006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216233947310000104003840038400384003840038
100244003730006139407251003410100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100100640316333947310000104003840038400384003840038

Test 3: Latency 1->3

Code:

  fnmadd 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)0318191e3a3f4e51inst 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)9fa8a9accdcfd0d5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1020440037300000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000060710121622394790100001004003840038400384003840038
1020440037299000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037300006061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037300000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037299000082394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000030710121622394790100001004003840038400384003840038
1020440037300000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037300006061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037300000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037300000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037300000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)0308181e3f4e51inst 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)9fa6a8accfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1002440037300000613940725100101010000101000050570690814001840037400373813033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
1002440037300000613940725100101010000101000050570690814001840037400373813033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
1002440037300000613940725100101010000101000050570690814001840037400373813033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
1002440037299000613940725100101010000101000050570690814001840037400373813033876710010201016220300004003740037111002110910101000010000640316343947310000104003840038400384003840038
1002440037300000613940725100101010000101000050570690814001840037400373813033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
1002540037299000613940725100101010000101000050570690814001840037400373813033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
1002440037300000613940725100101010000101000050570690814001840037400373814733876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
1002440037300000943940725100101010000101000050570690814001840037400373813033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
1002440037299000613940725100101010000101000050570690814001840037400373813033876710010201000020300004003740037111002110910101000010000640316333947310000104003840085400384003840038
1002440037299000613940725100101010000101000050570690814001840037400373813033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038

Test 4: Latency 1->4

Code:

  fnmadd 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)030b181e1f3f4e51inst 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)9fa0a1a8accfd0d5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
102044003730000006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710031622394790100001004003840038400384003840038
102044003730000006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710021622394790100001004003840038400384003840038
102044003730000006139407251010011410000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710021622394790100001004003840038400384003840038
102044003729900006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710021622394790100001004003840038400384003840038
102044003729900606139407251010010010000100100005005706908400184003740037381083387451010020010000200319984008640037111020110099100100100001000000710021622394790100001004003840038400384003840038
1020440037299000053639407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710021622394790100001004003840038400384003840038
102044003730000006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710021622394790100001004003840038400384003840038
1020440037299004806139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000400710021622394790100001004003840038400384003840038
1020440037300001506139371251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710001622394790100001004003840038400384003840038
102044003730000006139407251010010010000100100005005706908400184003740037381083387451010020010000200300004003740037111020110099100100100001000000710021622394790100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)03080b18191e1f3f4e51inst 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)9fa8a9acc5cfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
100244003729900002406139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216213947310000104008640038401804003840038
10024400373000100006139407251001010100001010000505706908140018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000006139407251001010100001010000505706908040018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038

Test 5: throughput

Count: 8

Code:

  fnmadd s0, s8, s9, s10
  fnmadd s1, s8, s9, s10
  fnmadd s2, s8, s9, s10
  fnmadd s3, s8, s9, s10
  fnmadd s4, s8, s9, s10
  fnmadd s5, s8, s9, s10
  fnmadd s6, s8, s9, s10
  fnmadd 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)0308090b18191e1f3a3f51inst 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)9fa0a6a7a8a9acc2cfd5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
8020420060150000000004225801001008000010080000500640000200210200402004099733999880100200800002002400002004020040118020110099100100800001000000000511021611200370800001002004120041200412004120041
80204200401500000000042258010010080000100800005006400002006102004020040997331002680312200800002002400002004020040118020110099100100800001000000000511011611200370800001002004120041200412004120041
8020420040150000000004225801001008000010080000500640000200210200402004099733999880100200800002002404082004020040118020110099100100800001000000000511011611200370800001002004120041200412004120041
8020420040150000000004225801001008000010080000500640000200210200402004099733999880100200800002002400002004020040118020110099100100800001000000000511011611200370800001002004120041200412004120041
8020420040150010000004225801001008000010080000500640000200210200402004099733999880100200800002002400002004020040118020110099100100800001000002000511011611200370800001002004120041200412004120041
8020420040150000000004225801001008000010080000500640000200210200402004099733999880100200800002002400002004020040118020110099100100800001000000000511011611200370800001002004120041200412004120041
8020420040150000000004225801001008000010080000500640000200210200402004099733999880100200800002002400002004020040118020110099100100800001000000000511011611200370800001002004120041200412004120041
8020420040150000000004225801001008000010080000500640000200210200402004099733999880100200800002002400002004020040118020110099100100800001000000000511011611200370800001002004120041200412004120041
80204200401500000000042258010010080000100800005006400002002102004020040997339998801002008000020024000020040200401180201100991001008000010000034000511011611200370800001002004120041200412004120041
8020420040150000000004225801001008000010080000500640000200210200402004099733999880100200800002002400002004020040118020110099100100800001000000000511011611200370800001002004120041200412004120041

1000 unrolls and 10 iterations

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

retire (01)cycle (02)030708090a0b18191e1f3a3f51inst 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)9fa1a6a8a9accfd0d2d5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
8002420050150000000000014825800101080000108000050640000200212004020040999631002080010208000020240000200402004011800211091010800001000000502000716992003780000102004120041200412004120041
8002420040150000000000041258001010800001080000506400002002120040200409996310020800102080000202400002004020040118002110910108000010000005020009168122003780000102004120041200412004120041
80024200401500000000900338258001010800001080000506400002002120040200409996310020800102080000202400002004020040118002110910108000010000005020008167112003780000102004120041200412004120041
800242004015000000000004125800101080000108000050640000200212004020040999631002080010208000020240000200402004011800211091010800001000000502000816882003780000102004120041200412004120041
8002420040150000000000041258001010800001080000506400002002120040200409996310020800102080000202400002004020040118002110910108000010000005020007167102003780000102004120041200412004120041
80024200401500000000120041525800101080000108000050640000200212004020040999631002080010208000020240000200402004011800211091010800001000000502000716982003780000102004120041200412004120041
8002420040150000000000041258001010800001080000506400002002120040200409996310020800102080000202400002004020040118002110910108000010000005020009167102003780000102004120041200412004120041
8002420040150000000000041258001010800001080000506400002002120040200401000431002080010208000020240000200402004011800211091010800001000000502000816882003780000102004120041200412004120041
80024200401500000000000146258001010800001080000506400002002120040200409996310020800102080000202400002004020040118002110910108000010000005020009168102003780000102004120041200412004120041
800242004015000000000004725800101080000108000050640000200212004020040999631002080010208000020240000200402004011800211091010800001000000502000716682003780000102004120041200412004120041