Apple Microarchitecture Research by Dougall Johnson

M1/A14 P-core (Firestorm): Overview | Base Instructions | SIMD and FP Instructions
M1/A14 E-core (Icestorm):  Overview | Base Instructions | SIMD and FP Instructions

CSINV (32-bit)

Test 1: uops

Code:

  csinv w0, w0, w1, hi
  mov x0, 1
  mov x1, 2

(no loop instructions)

1000 unrolls and 1 iteration

Retires: 1.000

Issues: 1.000

Integer unit issues: 1.001

Load/store unit issues: 0.000

SIMD/FP unit issues: 0.000

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
100410301001100110002560310001000300010011000
100410301001100110002560310001000300010011000
100410301001100110002560310001000300010011000
100410301001100110002560310001000300010011000
100410301001100110002560310001000300010011000
100410301001100110002560310001000300010011000
100410301001100110002560310001000300010011000
100410301001100110002560310001000300010011000
100410301001100110002560310001000300010011000
100410301001100110002560310001000300010011000

Test 2: Latency 1->2

Code:

  csinv w0, w0, w1, hi
  mov x0, 1
  mov x1, 2

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 1.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
10204100301010110101101072593301010710214302421000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100

1000 unrolls and 10 iterations

Result (median cycles for code): 1.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
10024100301002110021100282596281002810036300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010

Test 3: Latency 1->3

Code:

  csinv w0, w1, w0, hi
  mov x0, 1
  mov x1, 2

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 1.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
10204100301010110101101082594151010710214302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100
10204100301010110101101072595391010710212302361000110100

1000 unrolls and 10 iterations

Result (median cycles for code): 1.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
10024100301002110021100282594901002810034300681001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010
10024100301002110021100202595911002010020300201001110010

Test 4: Latency 1->4

Chain cycles: 1

Code:

  csinv w0, w1, w2, hi
  tst x0, 1
  mov x0, 1
  mov x1, 2
  mov x2, 3

(non-fused SUB/CBNZ loop)

100 unrolls and 100 iterations

Result (median cycles for code, minus 1 chain cycle): 1.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
20204200302020120201202095190382020920216402322010110100
20204200302020120201202085194482020820216402322010110100
20204200302020120201202085194482020820216402322010110100
20204200302020120201202085194482020820216402322010110100
20204200302020120201202085194482020820216402322010110100
20204200302020120201202085194482020820216402322010110100
20204200302020120201202085194482020820216402322010110100
20204200302020120201202085194482020820216402322010110100
20204200302020120201202095194482020820216402322010110100
20204200302020120201202085194482020820216402322010110100

1000 unrolls and 10 iterations

Result (median cycles for code, minus 1 chain cycle): 1.0030

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
20025200602003520035200695196282002820036400202001110010
20024200302002120021200205195912002020020400202001110010
20024200302002120021200205195912002020020400202001110010
20024200302002120021200205195912002020020400202001110010
20024200302002120021200205195912002020020400202001110010
20024200302002120021200205195912002020020400202001110010
20024200302002120021200205195912002020020400202001110010
20024200302002120021200205195912002020020400202001110010
20024200302002120021200205195912002020020400202001110010
20024200302002120021200205195912002020020400202001110010

Test 5: throughput

Count: 8

Code:

  csinv w0, w8, w9, hi
  csinv w1, w8, w9, hi
  csinv w2, w8, w9, hi
  csinv w3, w8, w9, hi
  csinv w4, w8, w9, hi
  csinv w5, w8, w9, hi
  csinv w6, w8, w9, hi
  csinv w7, w8, w9, hi
  mov x8, 9
  mov x9, 10

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.3342

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
802042692180115801158012029163480120802222402668001580100
802042675280115801158012029323680120802242404348005780100
802042673780115801158012029323680120802242402728001580100
802042673780115801158012029323680120802242402728001580100
802042673780115801158012029323680120802242402728001580100
802042673780115801158012029323680120802242402728001580100
802042673780115801158012029323680120802242402728001580100
802042673780115801158012029323680120802242402728001580100
802042673780115801158012029323680120802242402728001580100
802042673780115801158012029323680120802242402728001580100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.3340

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)dispatch int uop (56)int uops in schedulers (59)dispatch uop (78)map int uop (7c)map int uop inputs (7f)? int output thing (e9)? int retires (ef)
800242799980038800388005438943780020800202402968007280010
800242684780021800218002042678080020800202400208001180010
800242672380021800218002035742980020800202401828005480010
800242674080021800218002042678080020800202401228002880010
800242678180021800218002042678080020800202400208001180010
800242675880021800218002042678080020800202400208001180010
800242672380021800218002042678080020800202400208001180010
800242672380021800218002042678080020800202400208001180010
800242672380021800218002042678080020800202400208001180010
800242672380021800218002042678080020800202400208001180010