Re: [PATCH 2/5] RISC-V: memmove() speed optimized: Replace macros and use fixed-width types
Marlene Fally <[email protected]>
| Newsgroups | gmane.comp.lib.newlib |
|---|---|
| Message-ID | <CAA_fkGtiF9=uzawxKOaf2gTfP=M7umrUgQHq3n8WJT+=-b5Q5g@mail.gmail.com> |
Hi Alexey, thank you for your comments! These changes makes newlib ignore "_HAVE_HW_MISALIGNED_ACCESS". I am happy to change it back to use _HAVE_HW_MISALIGNED_ACCESS if that is preferred, but since these changes are only to the risc-v port, and the other functions in this port check for either __riscv_misaligned_fast, __riscv_misaligned_slow, and/or __riscv_misaligned_avoid, I figured it would make sense to opt for one of these macros. Also, checking for __riscv_misaligned_fast allows for still making use of the source-alignment-operation if misaligned accesses are allowed but slow. What purpose of replacing macros? (I can't see advantages to this). > Is "sizeof(long)" gives something different compared to SZREG ? The purpose of using SZREG here is to have consistency between all the functions in the port. I don't think in this case sizeof(long) would give something different – as far as I'm aware the valid ABIs for risc-v have long be 4 bytes for a 32-bit architecture, and 8 bytes for a 64-bit architecture, which matches the respective SZREG values. Best, M Op di 27 mei 2025 om 08:04 schreef Alexey Lapshin < [email protected]>: > These changes makes newlib ignore "_HAVE_HW_MISALIGNED_ACCESS". > > What purpose of replacing macros? (I can't see advantages to this). > Is "sizeof(long)" gives something different compared to SZREG ? > > On Mon, 2025-05-26 at 17:05 +0200, m fally wrote: > > Remove macros or replace them with static inline functions or RISC-V > > specific macros where applicable. > > Change data types to fixed-width and/or RISC-V specific types. > > > > Reviewed-by: Christian Herber <[email protected]> > > Signed-off-by: m fally <[email protected]> > > --- > > newlib/libc/machine/riscv/memmove.c | 60 +++++++++++++++++------------ > > 1 file changed, 35 insertions(+), 25 deletions(-) > > > > diff --git a/newlib/libc/machine/riscv/memmove.c > b/newlib/libc/machine/riscv/memmove.c > > index b48da0905..2e5c6ca9b 100644 > > --- a/newlib/libc/machine/riscv/memmove.c > > +++ b/newlib/libc/machine/riscv/memmove.c > > @@ -14,19 +14,29 @@ > > #else > > > > #include "../../string/local.h" > > -#include <_ansi.h> > > +#include "sys/asm.h" > > +#include "xlenint.h" > > #include <limits.h> > > #include <stddef.h> > > #include <string.h> > > > > -/*SUPPRESS 20*/ > > +static inline uint8_t > > +__libc_fast_xlen_aligned (void *dst, const void *src) > > +{ > > +#if defined(__riscv_misaligned_fast) > > + return 1; > > +#else > > + return !(((uintxlen_t)src & (SZREG - 1)) | ((uintxlen_t)dst & (SZREG > - 1))); > > +#endif > > +} > > + > > void *__inhibit_loop_to_libcall > > memmove (void *dst_void, const void *src_void, size_t length) > > { > > - char *dst = dst_void; > > - const char *src = src_void; > > - long *aligned_dst; > > - const long *aligned_src; > > + unsigned char *dst = dst_void; > > + const unsigned char *src = src_void; > > + uintxlen_t *aligned_dst; > > + const uintxlen_t *aligned_src; > > > > if (src < dst && dst < src + length) > > { > > @@ -34,21 +44,21 @@ memmove (void *dst_void, const void *src_void, > size_t length) > > src += length; > > dst += length; > > > > - if (!TOO_SMALL_LITTLE_BLOCK (length) && !UNALIGNED_X_Y (src, dst)) > > + if (length >= SZREG && __libc_fast_xlen_aligned (dst, src)) > > { > > - aligned_dst = (long *)dst; > > - aligned_src = (long *)src; > > + aligned_dst = (uintxlen_t *)dst; > > + aligned_src = (uintxlen_t *)src; > > > > - /* Copy one long word at a time if possible. */ > > - while (!TOO_SMALL_LITTLE_BLOCK (length)) > > + /* Copy one uintxlen_t word at a time if possible. */ > > + while (length >= SZREG) > > { > > *--aligned_dst = *--aligned_src; > > - length -= LITTLE_BLOCK_SIZE; > > + length -= SZREG; > > } > > > > /* Pick up any residual with a byte copier. */ > > - dst = (char *)aligned_dst; > > - src = (char *)aligned_src; > > + dst = (unsigned char *)aligned_dst; > > + src = (unsigned char *)aligned_src; > > } > > > > while (length--) > > @@ -61,31 +71,31 @@ memmove (void *dst_void, const void *src_void, > size_t length) > > /* Use optimizing algorithm for a non-destructive copy to closely > > match memcpy. If the size is small or either SRC or DST is > unaligned, > > then punt into the byte copy loop. This should be rare. */ > > - if (!TOO_SMALL_LITTLE_BLOCK (length) && !UNALIGNED_X_Y (src, dst)) > > + if (length >= SZREG && __libc_fast_xlen_aligned (dst, src)) > > { > > - aligned_dst = (long *)dst; > > - aligned_src = (long *)src; > > + aligned_dst = (uintxlen_t *)dst; > > + aligned_src = (uintxlen_t *)src; > > > > - /* Copy 4X long words at a time if possible. */ > > - while (!TOO_SMALL_BIG_BLOCK (length)) > > + /* Copy 4X uintxlen_t words at a time if possible. */ > > + while (length >= (SZREG * 4)) > > { > > *aligned_dst++ = *aligned_src++; > > *aligned_dst++ = *aligned_src++; > > *aligned_dst++ = *aligned_src++; > > *aligned_dst++ = *aligned_src++; > > - length -= BIG_BLOCK_SIZE; > > + length -= SZREG * 4; > > } > > > > - /* Copy one long word at a time if possible. */ > > - while (!TOO_SMALL_LITTLE_BLOCK (length)) > > + /* Copy one uintxlen_t word at a time if possible. */ > > + while (length >= SZREG) > > { > > *aligned_dst++ = *aligned_src++; > > - length -= LITTLE_BLOCK_SIZE; > > + length -= SZREG; > > } > > > > /* Pick up any residual with a byte copier. */ > > - dst = (char *)aligned_dst; > > - src = (char *)aligned_src; > > + dst = (unsigned char *)aligned_dst; > > + src = (unsigned char *)aligned_src; > > } > > > > while (length--) > >