xor-neon.c 1.2 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243
  1. // SPDX-License-Identifier: GPL-2.0-only
  2. /*
  3. * linux/arch/arm/lib/xor-neon.c
  4. *
  5. * Copyright (C) 2013 Linaro Ltd <ard.biesheuvel@linaro.org>
  6. */
  7. #include <linux/raid/xor.h>
  8. #include <linux/module.h>
  9. MODULE_LICENSE("GPL");
  10. #ifndef __ARM_NEON__
  11. #error You should compile this file with '-march=armv7-a -mfloat-abi=softfp -mfpu=neon'
  12. #endif
  13. /*
  14. * Pull in the reference implementations while instructing GCC (through
  15. * -ftree-vectorize) to attempt to exploit implicit parallelism and emit
  16. * NEON instructions.
  17. */
  18. #if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)
  19. #pragma GCC optimize "tree-vectorize"
  20. #else
  21. /*
  22. * While older versions of GCC do not generate incorrect code, they fail to
  23. * recognize the parallel nature of these functions, and emit plain ARM code,
  24. * which is known to be slower than the optimized ARM code in asm-arm/xor.h.
  25. */
  26. #warning This code requires at least version 4.6 of GCC
  27. #endif
  28. #pragma GCC diagnostic ignored "-Wunused-variable"
  29. #include <asm-generic/xor.h>
  30. struct xor_block_template const xor_block_neon_inner = {
  31. .name = "__inner_neon__",
  32. .do_2 = xor_8regs_2,
  33. .do_3 = xor_8regs_3,
  34. .do_4 = xor_8regs_4,
  35. .do_5 = xor_8regs_5,
  36. };
  37. EXPORT_SYMBOL(xor_block_neon_inner);