Apple M1 Microarchitecture Research by Dougall Johnson

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

SQDMULL2 (by element, 4S)

Test 1: uops

Code:

  sqdmull2 v0.4s, v0.8h, v1.h[1]
  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

retire (01)cycle (02)031e3f4e51inst 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)a8a9accfd5d6ddinst fetch restart (de)e0? fp/simd (ee)f5f6f7f8fd
1004303723661254825100010001000398313030183037303724153289510001000200030373037111001100010073116112630100030383038303830383038
1004303723061254825100010001000398313130183037303724153289510001000200030373037111001100000073116112630100030383038303830383038
1004308422361254825100010001000398313030183037303724153289510001000200030373037111001100000073116112630100030383038303830383038
1004303723061254825100010001000398313030183037303724153289510001000200030373037111001100000073116112630100030383038303830383038
1004303723061254825100010001000398313030183037303724153289510001000200030373037111001100000073116112630100030383038303830383038
1004303722361254825100010001000398313130183037303724153289510001000200030373037111001100000073116112630100030383038303830383038
1004303722061254825100010001000398313130183037303724153289510001000200030373037111001100000073116112630100030383038303830383038
1004303722061254825100010001000398313130183037303724193289510001000200030373037111001100000073116112630100030383038303830383038
1004303722082254825100010001000398313130183037303724153289510001000200030373037111001100000073116112630100030383038303830383038
1004303723061254825100010001000398313030183037303724153289510001000200030373037111001100000073116112630100030383038303830383038

Test 2: Latency 1->2

Code:

  sqdmull2 v0.4s, v0.8h, v1.h[1]
  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

retire (01)cycle (02)03080b18191e1f3a3f4e51inst 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)9fa0a1a8a9acc2c5cfd5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
102043003722500000006129548251010010010000100100005004277313130018300373003728265328745101002001000020020000300373003711102011009910010010000100000000071011611296340100001003003830038300383008530038
1020430037225000000020842954825101001001000010010149527427867013001830134300852826532874510100200103282022000030037300371110201100991001001000010024042226540805281122994434100001003018030465304643037430417
10204304182280098120070405419294762051021914310032144111927474288169130306304183041728289392889511339228114962272233430324304059110201100991001001000010000002245500893289422985031100001003042430419304683045330472
10204304172281149105679235227294761811020915610064154111927274289225030342300373003728265328745101002001000020020000300373003711102011009910010010000100000000071011611296340100001003003830038300383003830038
102043003722400000006129548251010010010000100100005004277313030018300373003728265328745101002001000020020000300373003711102011009910010010000100000000071011611296340100001003003830038300383003830038
102043003722500000006129548251010010010000100100005004277313130018300373003728265328745101002001000020020000300373003711102011009910010010000100000000071011611296340100001003003830038300383003830038
102043003722500000006129548251010010010000100100005004277313030018300373003728265328745101002001000020020000300373003711102011009910010010000100000000071011611296340100001003003830038300383003830038
102043003722500000006129548251010010010000100100005004277313030018300373003728265328745101002001000020020000300373003711102011009910010010000100000000071011611296340100001003003830038300383003830038
102043003722500000006129548251010010010000100100005004277313030018300373003728265328745101002001000020020000300373003711102011009910010010000100000000071011611296340100001003003830038300383003830038
102043003722500000006129548251010010010000100100005004277313030018300373003728265328745101002001000020020000300373003711102011009910010010000100000000071011611296340100001003003830038300383003830038

1000 unrolls and 10 iterations

Result (median cycles for code): 3.0037

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)9fa0a1a6a8a9acc2cdcfd5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1002430037225000000612954825100101010000101000050427731303001830037300372828732876710010201000020200003003730037111002110910101000010000000006403163329630010000103003830038300383003830038
1002430037225000000612954825100101010000101000050427731313001830037300372828732876710010201000020200003003730037111002110910101000010000000006403163329630010000103003830038300383003830038
1002430037225000000662954825100101010000101000050427731313001830037300372828732876710010201000020200003003730037111002110910101000010000100006403163329630010000103003830038300383003830038
1002430037225000000612954825100101010000101000050427731313001830037300372828732876710010201000020200003003730037111002110910101000010000000006403163329630010000103003830038300383003830038
1002430037225000000612954825100101010000101000050427731313001830037300372828732876710010201000020200003003730037111002110910101000010000000006403163329630010000103003830038300383003830038
1002430037225000000612954825100101010000101000050427731313001830037300372828732876710010201000020200003003730037111002110910101000010000000006403163329630010000103003830038300383003830038
1002430037225000000612954825100101010000101000050427731313001830037300372828732876710010201000020200003003730037111002110910101000010000000006403163329630010000103003830038300383003830038
1002430037225000000612954825100101010000101000050427731313001830037300372828732876710010201000020200003003730037111002110910101000010000000006403163329630010000103003830038300383003830038
1002430037225000000612954825100101010000101000050427731313001830037300372828732876710010201000020200003003730037111002110910101000010000000006403163329630010000103003830038300383003830038
1002430037225000000612954825100101010000101000050427731313001830037300372828732876710010201000020200003003730037111002110910101000010000000006403163329630010000103003830038300383003830038

Test 3: Latency 1->3

Code:

  sqdmull2 v0.4s, v1.8h, v0.h[1]
  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

retire (01)cycle (02)030b18191e1f3f4e51inst 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)9fa1a8a9acc2cfd0d5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
1020430037241000122121345295396510142122100161191029862842796041300900301313013328272132876310253204101672042066030037300371110201100991001001000010000253600755023211297045100001003013330085301343003830038
10204300372251110076829548251010010010000100100005004277313130018030037300372826532874510100200100002002000030037300371110201100991001001000010000000710011611296340100001003003830038300383003830038
1020430037225000006129548251010010010000100100005004277313130018030037300372826532874510100200100002002000030037300371110201100991001001000010000000710011611296340100001003003830085300383003830038
10204300372250000031229548251010010010000100101485004277313130018030037300372826532874510100200100002002000030037300371110201100991001001000010000000710011611296340100001003003830038300383003830038
1020430037225000006129548251010010010000100100005004277313130018030037300372826532874510100200100002002000030037300371110201100991001001000010000000710011611296340100001003003830038300383003830038
1020430037225000006129548251010010010000100100005004277313130018030037300372826532874510100200100002002000030037300371110201100991001001000010000000710011611296340100001003003830038300383003830038
1020430037225000006129548251010010010000100100005004277313130018030037300372826532874510100200100002002000030037300371110201100991001001000010000000710011611296340100001003003830038300383003830038
1020430037225000006129548251010010010008100100005004277313130018030037300372826532874510100200100002002000030037300371110201100991001001000010000000710011611296340100001003003830038300383003830038
1020430037225000006129548251010010010000100100005004277313130018030037300372826532874510100200100002002000030037300371110201100991001001000010000000710011611296340100001003003830038300383003830038
1020430037228000306129548251010010010000100100005004277313130018030037300372826532874510100200100002002000030037300371110201100991001001000010000000710011611296340100001003003830038300383003830038

1000 unrolls and 10 iterations

Result (median cycles for code): 3.0037

retire (01)cycle (02)030718191e1f3f4e51inst 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)9fa0a8a9acc5cfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
10024300372250000061295482510010101000010100005042773133001803003730037282873287671001020100002020000300373003711100211091010100001000000640416652963010000103003830038300383003830038
10024300372250000061295482510010101000010100005042773133001803003730037282873287671001020100002020000300373003711100211091010100001000001640516552963010000103003830038300383003830038
10024300372240000061295482510010101000010100005042773133001803003730131282873287671001020100002020000300373003711100211091010100001002000682616652963010000103003830038300383013230038
100243003722600000103295482510010101000010100005042773133001803003730131282873287671016120103242020000301313003711100211091010100001000000640516562963010000103003830038300383003830038
10024300372250000061295482510010101000010100005042773133001803003730037282873287671001020100002020000300373003711100211091010100001000000640516662963010000103003830038300383003830038
10024300372250000061295482510010101000010100005042773133001803003730037282873287671001020100002020000300373003711100211091010100001000000640616652963010000103003830038300383003830038
10024300372250000061295482510010101000010100005042773133001803003730037282873287671001020100002020000300373003711100211091010100001000000640516562963010000103003830038300383003830038
10024300372250000061295484410010101000010100005042773133001803003730037282873287671001020100002020000300373003711100211091010100001000000640616562963010000103003830038300383003830038
10024300372250000061295482510010101000010100005042773133001803003730037282873287671001020100002020000300373003711100211091010100001000000640516552963010000103003830038300383003830038
10024300372250000061295482510010101000010100005042773133001803003730037282873287671001020100002020000300373003711100211091010100001000000640616662963010000103003830038300383003830038

Test 4: throughput

Count: 8

Code:

  sqdmull2 v0.4s, v8.8h, v9.h[1]
  sqdmull2 v1.4s, v8.8h, v9.h[1]
  sqdmull2 v2.4s, v8.8h, v9.h[1]
  sqdmull2 v3.4s, v8.8h, v9.h[1]
  sqdmull2 v4.4s, v8.8h, v9.h[1]
  sqdmull2 v5.4s, v8.8h, v9.h[1]
  sqdmull2 v6.4s, v8.8h, v9.h[1]
  sqdmull2 v7.4s, v8.8h, v9.h[1]
  movi v8.16b, 9
  movi v9.16b, 10

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

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

retire (01)cycle (02)03080b18191e1f3a3f51inst 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)9fa0a6a7a8a9acc2cfd5d6ddinst fetch restart (de)e0? int output thing (e9)? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
802042005815000000004125801001008000010080000500640000120020200392003999733999780100200800002001600002003920039118020110099100100800001000002000511021611200360800001002004020040200402004020040
802042003915000000004125801001008000010080000500640000120020200392003999733999780100200800002001600002003920039118020110099100100800001000001000511011611200360800001002004020040200402004020040
802042003915000000004125801001008000010080000500640000120020200392003999733999780100200800002001600002003920039118020110099100100800001000001000511011611200360800001002004020040200402004020040
8020420039150000000041258010010080000100800005006400000200202003920039997339997801002008000020016000020039200391180201100991001008000010000039000511011611200360800001002004020040200402004020040
802042003915000000004125801001008000010080000500640000120020200392003999733999780100200800002001600002003920039118020110099100100800001000000000511011611200360800001002004020040200402004020040
802042003915000000004125801001008000010080000500640000120020200392003999733999780100200800002001600002003920039118020110099100100800001000000000511011611200360800001002004020040200402004020040
802042003915000000004125801001008000010080000500640000120020200392003999733999780100200800002001600002003920039118020110099100100800001000002000511011611200360800001002004020040200402004020040
802042003915000000004125801001008000010080000500640000120020200392003999733999780231200800002001600002003920039118020110099100100800001000000000511011611200360800001002004020040200402004020040
802042003915000000004125801001008000010080000500641632120020200392003999733999780100200800002001600002003920039118020110099100100800001000000034511011611200360800001002004020040200402004020040
802042003915000000004125801001008000010080000500640000020020200392003999733999780100200800002001600002003920039118020110099100100800001000000000511011611200360800001002004020040200402004020040

1000 unrolls and 10 iterations

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

retire (01)cycle (02)031e3f51inst 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)9fa6a8accdcfd5d6ddinst fetch restart (de)e0? fp/simd (ee)gpr retires (ef)f5f6f7f8fd
8002420048150040258001010800001080000506400001200202003920039999631001980010208000020160000200392003911800211091010800001000005020616652003680000102004020040200402004020040
8002420039150040258001010800001080000506400001200202003920039999631001980010208000020160000200392003911800211091010800001000005020516852003680000102004020040200402004020040
8002420039150045258001010800001080000506400001200202003920039999631001980010208000020160000200392003911800211091010800001000005020516572003680000102004020115200402011320040
8002420039150040258001010800001080000506400001200202003920039999631001980010208000020160000200392003911800211091010800001000005020516682003680000102004020040200402004020040
8002420039150040258001010800001080000506400001200202003920039999631001980010208000020160000200392003911800211091010800001000005020816762003680000102004020040200402004020040
8002420039150040258001010800001080000506400001200202003920039999631001980010208000020160000200392003911800211091010800001000005020516852003680000102004020040200402004020040
8002420039150040258001010800001080000506400001200202003920039999631001980010208000020160000200392003911800211091010800001001015020916562003680000102004020040200402004020040
8002420039150040258001010800001080000506416641200202003920039999631001980010208000020160000200392003911800211091010800001000305020716872003680000102004020040200402004020040
80024200391501840258001010800001080000506400001200202003920039999631001980010208000020160000200392003911800211091010800001001305020716652003680000102004020040200402004020040
8002420039150040258001010800001080000506400001200202003920039999631001980010208000020160000200392003911800211091010800001000005020716562003680000102004020040200402004020040