Apple M1 Microarchitecture Research by Dougall Johnson

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

FMADD (scalar, H)

Test 1: uops

Code:

  fmadd 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)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)a6a8a9accfd5d6ddinst fetch restart (de)e0? fp/simd (ee)f5f6f7f8fd
1004403730061340725100010001000531908401840374037325833895100010003000403740371110011000000073116113473100040384038403840384038
1004403731061340725100010001000531908401840374037325833895100010003000403740371110011000000073116113473100040384038403840384038
1004403730061340725100010001000531908401840374037325833895100010003000403740371110011000000073116113473100040384038403840384038
1004403730061340725100010001000531908401840374037325833895100010003000403740371110011000000073116113473100040384038403840384038
1004403730061340725100010001000531908401840374037325833895100010003000403740371110011000010073116113473100040384038403840384038
1004403730061340725100010001000531908401840374037325833895100010003000403740371110011000070073116113473100040384038403840384038
1004403730061340725100010001000531908401840374037325833895100010003000403740371110011000060073116113473100040384038403840384038
1004403730061340725100010001000531908401840374037325833895100010003000403740371110011000000073116113473100040384038403840384038
1004403730061340725100010001000531908401840374037325833895100010003000403740371110011000000673116113473100040384038403840384038
1004403731061340725100010001000531908401840374037325833895100010003000403740371110011000012073116113473100040384038403840384038

Test 2: Latency 1->2

Code:

  fmadd 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)9fa6a8acc5cfd0d5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1020440037299000613940725101001001000010010000500570690814001840037400373810803387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
10204400373000001033940725101001001000010010000500570690814001840037400373810873387451010020010000200300004003740037111020110099100100100001000000710121632394794100001004003840038400384003840038
1020440037299000613940725101001001000010010000500570690814001840037400373810803387451010020010000200300004003740037111020110099100100100001000000710121632394790100001004003840038400384003840038
1020440037300000613940725101001001000010010000500570690814001840037400373810803387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037300000613940725101001001000010010000500570690814001840037400373810803387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037300000613940725101001001000010010000500570690814001840037400373810803387451010020010000200300004003740037111020110099100100100001000000712121622394790100001004003840038400384003840038
1020440037300000613940725101001001000010010000500570690814001840037400373813903387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
1020440037300000613940725101001001000010010000500570690814001840037400373810803387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038
10204400373000006139407251010010010000100100005005706908140018400374003738108033874510100200100002003000040037400371110201100991001001000010001141760710121622394790100001004003840038400384003840038
1020440037300000613940725101001001000010010000500570690814001840037400373810803387451010020010000200300004003740037111020110099100100100001000000710121622394790100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)030a1e3f4e51inst 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)9fa0a6a8a9acc2cdcfd5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
10024400373000061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001000000006402163239473010000104003840038400384003840038
10024400373000061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001000000006402162239473110000104003840038400384003840038
10024400373000061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001000000006402163239473010000104003840038400384003840038
10024400372990061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001000000006403162239473010000104003840038400384003840038
10024400373000061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100221091010100001000100006403163339473010000104003840038400384003840038
10024400372990061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001000000006402162239473010000104003840038400384003840038
100244003730000726394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001000000006402162239473010000104003840038400384003840038
10024400372990061394072510010101000010100005057069081400184003740037381303387671001020100002030000400374003711100211091010100001000000006403162239473010000104003840038400384003840038
10024400373000061394072510010101000010100005057069081400184003740037381303387671015720100002030000400374003711100211091010100001000000006402162239473010000104003840038400384003840038
10024400373000061394072510010101000010100005057074391400184003740037381303387671001020100002030000400374008411100211091010100001000000006403163339473010000104003840038400384003840038

Test 3: Latency 1->3

Code:

  fmadd 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)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
10204400373000061394072510100100100001001000050057069081400184003740037381156387411010020010008200300244003740037111020110099100100100001000011171800160139490100001004003840038400384003840038
10204400372990061394072510100100100001001000050057069081400534003740037381157387411010020010008200300244003740037111020110099100100100001000011171700160039490100001004003840038400384003840038
10204400373000061394072510100100100001001000050057069081400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071002162239479100001004003840038400384003840038
10204400373000061394072510100100100001001000050057069081400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071202162239479100001004003840038400384003840134
10204400373000061394072510100100100001001000050057069081400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071002162239479100001004003840038400384003840038
10204400372990061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071003162239479100001004003840038400384003840038
1020440037300672061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071002162239479100001004003840038400384003840038
10204400373000061394072510100100100001001000050057069081400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071002162239479100001004003840038400384003840038
10204400373000061394072510100100100001001000050057069081400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071002162239479100001004003840038400384003840038
10204400373000061394072510100100100001001000050057069080400184003740037381083387451010020010000200300004003740037111020110099100100100001000000071002162239479100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)030b18191e3f4e51inst 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)9fa1a8acc5cfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
10024400372990000613940725100101010000101000050570690840018400374003738130338767100102010000203000040037400371110021109101010000100000640316223947310000104003840038400384003840038
10024400373000000613940725100101010000101000050570690840018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000613940725100101010000101000050570690840018400374003738130338767100102010000203000040037400371110022109101010000100000640216223947310000104003840038400384003840038
1002440037300000061394072510010101000010100005057069084001840037400373813033876710010201000020300004003740037111002110910101000010001170640216223947310000104003840038400384003840038
10024400373000000613940725100101010000101000050570690840018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000613940725100101010000101000050570690840018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000613940725100101010000101000050570690840018400374003738130338767100102010000203000040037400371110021109101010000100000640216323947310000104003840038400384003840038
100244003730000001033940725100101010000101000050570690840018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000613940725100101010000101000050570690840018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038
10024400373000000613940725100101010000101000050570690840018400374003738130338767100102010000203000040037400371110021109101010000100000640216223947310000104003840038400384003840038

Test 4: Latency 1->4

Code:

  fmadd 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)030818191e1f3a3f4e51inst 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)9fa1a8accfd0d5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
10204400373000000016139407251010010010000100100005005706908040018040037400373810833874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
10204400373000000006139407251010010010000100100005005706908040018340037400373810833874510100200100002003000040037400371110201100991001001000010000071012162239969100001004003840038400384003840038
102044003729900000025139407251010010010000100100005005706908040018040037400373810833874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
102044003730000000061394072510100100100001001000050057069080400180400374003738108338745101002001000020030000400374003711102011009910010010000100001871012162239479100001004003840038400384003840038
10204400373000000006139407251010010010000100100005005706908140018040037400373810833874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
10204400373000000006139407251010010010000100100005005706908140018040037400373810833874510100200100002003000040037400371110201100991001001000010000071012162339479100001004003840038400384003840038
10204400373000000006139407251010010010000100100005005706908140018040037400373810833874510100200100002003000040037400371110201100991001001000010000071012163239479100001004003840038400384003840038
10205400373000000018239407251010010010000100100005005706908140018040037400373810833874510100200100002003000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
102044008530000000010339398251010010010000100100005005706908040018040037400373810833874510400200101662023000040037400371110201100991001001000010000071012162239479100001004003840038400384003840038
10204400373000000006139407251010010010000100100005005706908040018040037400373810833874510100200100002003000040037400371110201100991001001000010002071012162239479100001004003840038400384003840038

1000 unrolls and 10 iterations

Result (median cycles for code): 4.0037

retire (01)cycle (02)0308191e3f4e51inst 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)9fa8acc2cfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
10024400373000006139407251001010100001010000505706908140018400374003738130033876710010201000020300004003740037111002110910101000010130640316333947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908140018400374003738130033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
100244003730000361394072510010101000010100005057069081400184003740037381300338767100102010000203000040037400371110021109101010000100180640316333947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908140018400374003738130033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
10024400372990006139407251001010100001010000505706908140018400374003738130033876710010201000020300004003740037111002110910101000010000640316433947310000104003840038400384003840038
100244003730000015639407251001010100001010000505706908140018400374003738130033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
100244003729900084839407251001010100001010000505706908140018400374003738130033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
10024400373000006139407251001010100001010000505706908140018400374003738130033876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
100244003730000061394072510010101000010100005057069081400184003740037381300283876710010201000020300004003740037111002110910101000010000640316333947310000104003840038400384003840038
10024400372990006139407251001010100001010000505706908140018400374003738130033876710010201000020300004003740037111002110910101000010000640316343947310000104003840038400384003840038

Test 5: throughput

Count: 8

Code:

  fmadd h0, h8, h9, h10
  fmadd h1, h8, h9, h10
  fmadd h2, h8, h9, h10
  fmadd h3, h8, h9, h10
  fmadd h4, h8, h9, h10
  fmadd h5, h8, h9, h10
  fmadd h6, h8, h9, h10
  fmadd 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)03070a1e3a3f51inst 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)9fa1a8acc5branch mispredict (cb)cdcfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
80204201101501101352580108100800081008002050064013202002120040200409977699918012020080032200240096200402004011802011009910010080000100001211151205165520037800001002004120041200412004120041
80204200401501101352580108100800081008002050064013202002120040200409977699918012020080032200240096200402004011802011009910010080000100001511151206167620037800001002004120041200412004120041
80204200401501101462580100100800001008000050064000002002120040200409973399988010020080000200240000200402004011802011009910010080000100001500051128168820037800001002004120041200412004120041
80204200401501101462580100100800001008000050064336802002120040200409973399988010020080000200240000200402004011802011009910010080000100201200051126166520037800001002004120041200412004120041
802042004015011011326258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010000000051128167520037800001002004120041200412004120041
8020420040150110246258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010000900051128168820037800001002004120041200412009220041
8020420040150110146258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010000000051125165720037800001002004120041200412004120041
8020420040150110146258010010080000100800005006400000200212004020040997339998801002008000020024000020040200401180201100991001008000010003600051128168820037800001002004120041200412004120041
8020420040150110146258010010080000100800005006433320200212004020040997339998801002008000020024000020040200401180201100991001008000010004600051127167920037800001002004120041200412004120041
80204200401501101462580100100800001008000050064000002002120040200409973399988010020080000200240000200402004011802011009910010080000100001500051128165720037800001002004120041200412004120041

1000 unrolls and 10 iterations

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

retire (01)cycle (02)031e1f3a3f51inst 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)6061696d6edispatch 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)9fa8acbranch mispredict (cb)cdcfd0d2d5d6daddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
80024200401500001902580010108000010800005064000000200212004020040999631002080010208000020240000200402004011800211091010800001000005020501216010262003780000102004120041200412004120041
8002420040150000412580010108000010800005064000005200212004020040999631002080010208000020240000200402004011800211091010800001000005020501016010242003780000102004120041200412004120041
800242004015000041258001010800001080000506400001020021200402004099963100208001020800002024000020040200401180021109101080000100000502000916010192003780000102004120041200412004120041
8002420040150000412580010108000010800005064000000200212004020040999631002080010208000020240000200402004011800211091010800001000005020001116011152003780000102004120041200412004120041
8002420040150000412580010108000010800005064000005200212004020040999631002080010208000020240000200402004011800211091010800001000005020001016012192003780000102004120041200412004120041
800242004015000041258001010800001080000506400001020021200402004099963100208001020800002024000020040200401180021109101080000100000502000101609222003780000102004120041200412004120041
8002420040150000412580010108000010800005064000010200212004020040999631002080010208000020240000200402004011800211091010800001000005020041016010252003780000102004120041200412004120041
8002420040150000412580010108000010800005064000010200212004020040999631002080010208000020240000200402004011800211091010800001000005020001016011232003780000102004120041200412004120041
800242004015000041258001010800001080000506400001020021200402004099963100208001020800002024000020040200401180021109101080000100000502000916011152003780000102004120041200412004120041
800242004015000062258001010800001080000506400000020021200402004099963100208001020800002024000020040200401180021109101080000100000502050101608212003780000102004120041200412004120041