In what situation would the AVX2 gather instructions be faster than individually loading the data?

by

Newer microarchitectures have shifted the odds towards gather instructions. On an Intel Xeon Gold 6138 CPU @ 2.00 GHz with Skylake microarchitecture, we get for your benchmark:

9.383e+09 8.86e+08 2.777e+09 6.915e+09 7.793e+09 8.335e+09 5.386e+09 4.92e+08 6.649e+09 1.421e+09 2.362e+09 2.7e+07 8.69e+09 5.9e+07 7.763e+09 3.926e+09 5.4e+08 3.426e+09 9.172e+09 5.736e+09 
9.383e+09 8.86e+08 2.777e+09 6.915e+09 7.793e+09 8.335e+09 5.386e+09 4.92e+08 6.649e+09 1.421e+09 2.362e+09 2.7e+07 8.69e+09 5.9e+07 7.763e+09 3.926e+09 5.4e+08 3.426e+09 9.172e+09 5.736e+09 
9.383e+09 8.86e+08 2.777e+09 6.915e+09 7.793e+09 8.335e+09 5.386e+09 4.92e+08 6.649e+09 1.421e+09 2.362e+09 2.7e+07 8.69e+09 5.9e+07 7.763e+09 3.926e+09 5.4e+08 3.426e+09 9.172e+09 5.736e+09 
9.383e+09 8.86e+08 2.777e+09 6.915e+09 7.793e+09 8.335e+09 5.386e+09 4.92e+08 6.649e+09 1.421e+09 2.362e+09 2.7e+07 8.69e+09 5.9e+07 7.763e+09 3.926e+09 5.4e+08 3.426e+09 9.172e+09 5.736e+09 
Array length 10000, function called 1000000 times.
Gcc version: 6.32353
Nonvectorized assembly implementation: 6.36922
Vectorized without gather: 5.53553
Vectorized with gather: 4.50673

showing that gathers may now be well worth the effort.

More Related Contents:

  1. Test whether a register is zero with CMP reg,0 vs OR reg,reg?
  2. Does the Intel Memory Model make SFENCE and LFENCE redundant?
  3. Where is VPERMB in AVX2?
  4. What is the trick to learn x86 assembly language on Windows PC? [closed]
  5. Why do x86-64 instructions on 32-bit registers zero the upper part of the full 64-bit register?
  6. When and why do we sign extend and use cdq with mul/div?
  7. MOVing between two memory addresses
  8. Is ADD 1 really faster than INC ? x86 [duplicate]
  9. What is the instruction that gives branchless FP min and max on x86?
  10. What is the purpose of XORing a register with itself? [duplicate]
  11. What registers must be preserved by an x86 function?
  12. What is the meaning of MOV (%r11,%r12,1), %edx?
  13. x86, difference between BYTE and BYTE PTR
  14. Why is GCC pushing an extra return address on the stack?
  15. x86 XOR opcode differences
  16. Converting bin to hex in assembly
  17. what does movsbl instruction do? [duplicate]
  18. Counting 1 bits (population count) on large data using AVX-512 or AVX-2
  19. How to determine if the registers are loaded right to left or vice versa
  20. How to write an absolute target for a near direct relative call/jmp in MASM
  21. x86 Calculating AX given AH and AL?
  22. Get size of assembly instructions
  23. Printing out a number in assembly language?
  24. What is the penalty of mixing EVEX and VEX encoded scheme?
  25. x86 Assembly – Why is [e]bx preserved in calling conventions?
  26. Second stage of bootloader prints garbage using Int 0x10/ah=0x0e
  27. What is callq instruction?
  28. MUL function in assembly
  29. How is POPCNT implemented in hardware?
  30. Solution needed for building a static IDT and GDT at assemble/compile/link time

Leave a Comment