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

SQSHL (immediate, scalar, S)

Test 1: uops

Code:

  sqshl s0, s0, #3
  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: 0.001

Load/store unit issues: 0.000

SIMD/FP unit issues: 1.000

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch simd uop (57)ldst uops in schedulers (5b)dispatch uop (78)map simd uop (7e)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)
1004203310011100010005024810001000100011000
1004203310011100010005024810001000100011000
1004203310011100010005024810001000100011000
1004203310011100010005024810001000100011000
1004203310011100010005024810001000100011000
1004203310011100010005024810001000100011000
1004203310011100010005024810001000100011000
1004203310011100010005024810001000100011000
1004203310011100010005024810001000100011000
1004203310011100010005024810001000100011000

Test 2: Latency 1->2

Code:

  sqshl s0, s0, #3
  movi v0.16b, 1
  movi v1.16b, 2

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 2.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)schedule ldst uop (55)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
10204200331010110110000010010000300509248101002001000620010006110000100
10204200331010110110000010010000300509248101002001000420010004110000100
10204200331010110110000010010000300509248101002001000420010004110000100
10204200331010110110000010010000300509248101002001000420010004110000100
10204200331010110110000010010000300509248101002001000420010004110000100
10204200331010110110000010010000300509248101002001000420010004110000100
10204200331010110110000010010000300509248101002001000420010004110000100
10204200331010110110000010010000300509248101002001000420010004110000100
10204200331010110110000010010000300509248101002001000420010004110000100
10205200661010910110008010010034300509248101002001000420010004110000100

1000 unrolls and 10 iterations

Result (median cycles for code): 2.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
10024200331002121100002010000705092481002020100042010000111000010
10024200331002121100002010000705092481002020100002010000111000010
10024200331002121100002010000705092481002020100002010000111000010
10024200331002121100002010000705092481002020100002010000111000010
10024200331002121100002010000705092481002020100002010000111000010
10024200331002121100002010000705092481002020100002010000111000010
10024200331002121100002010000705092481002020100002010000111000010
10024200331002121100002010000705092481002020100002010000111000010
10024200331002121100002010000705092481002020100002010000111000010
10024200331002121100002010000705092481002020100002010000111000010

Test 3: throughput

Count: 8

Code:

  sqshl s0, s8, #3
  sqshl s1, s8, #3
  sqshl s2, s8, #3
  sqshl s3, s8, #3
  sqshl s4, s8, #3
  sqshl s5, s8, #3
  sqshl s6, s8, #3
  sqshl s7, s8, #3
  movi v8.16b, 9

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.5004

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)schedule ldst uop (55)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
80204400708010510180004010080008300320036801082008001220080012180000100
80204400348010710180006010080010300320036801082008001220080012180000100
80204400348010510180004010080008300320036801082008001220080012180000100
80204400348010510180004010080008300320036801082008001220080012180000100
80204400348010510180004010080008300320036801082008001220080012180000100
80204400348010510180004010080008300320036801082008001220080012180000100
80204400348010510180004010080008300320036801082008001220080012180000100
80204400348010510180004010080008300320200801522008006020080014180000100
80204400348010510180004010080008300320200801522008006020080014180000100
80204400348010510180004010080008300320036801082008001220080012180000100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.5004

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
80024401068002921800082080012703200448003020800162080000118000010
80024400348002121800002080000703200008002020800002080000118000010
80024400348002121800002080000703200008002020800002080000118000010
80024400348002121800002080000703200008002020800002080000118000010
80024400348002121800002080000703200008002020800002080000118000010
80024400348002121800002080000703200008002020800002080000118000010
80024400348002121800002080000703200008002020800002080000118000010
80024400348002121800002080000703200008002020800002080000118000010
80024400348002121800002080000703200008002020800002080000118000010
80024400348002121800002080000703200008002020800002080066118000010