Wilco Dijkstra <wilco.dijks...@arm.com> writes:
> Change the AARCH64_EXPAND_ALIGNMENT macro into proper function calls to make
> future changes easier. Use the existing alignment settings, however avoid
> overaligning small array's or structs to 64 bits when there is no benefit.
> This gives a small reduction in data and stack size.
So just to be sure I understand: we still want to align (say) an array
of 4 chars to 32 bits so that the LDR & STR are aligned, and an array of
3 chars to 32 bits so that the LDRH & STRH for the leading two bytes are
aligned? Is that right? We don't seem to take advantage of the padding
and do an LDR & STR for the 3-byte case, either for globals or on the stack.
If so, what's the advantage of aligning (say) a 6-byte array to 64 bits
rather than 32 bits, given that we don't use a 64-bit LDR & STR?
Could we save more with size < 64 instead of size <= 32?
Thanks,
Richard
> Passes regress & bootstrap, OK for commit?
>
> gcc/ChangeLog:
>
> * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): Remove.
> (DATA_ALIGNMENT): Use aarch64_data_alignment.
> (LOCAL_ALIGNMENT): Use aarch64_stack_alignment.
> * config/aarch64/aarch64.cc (aarch64_data_alignment): New function.
> (aarch64_stack_alignment): Likewise.
> * config/aarch64/aarch64-protos.h (aarch64_data_alignment): New
> prototype.
> (aarch64_stack_alignment): Likewise.
>
> ---
>
> diff --git a/gcc/config/aarch64/aarch64-protos.h
> b/gcc/config/aarch64/aarch64-protos.h
> index
> 6da81556110c978a9de6f6fad5775c9d77771b10..4133a47693b24abca071a7f77fcdbb91d3dc261a
> 100644
> --- a/gcc/config/aarch64/aarch64-protos.h
> +++ b/gcc/config/aarch64/aarch64-protos.h
> @@ -1207,4 +1207,7 @@ extern void aarch64_adjust_reg_alloc_order ();
> bool aarch64_optimize_mode_switching (aarch64_mode_entity);
> void aarch64_restore_za (rtx);
>
> +extern unsigned aarch64_data_alignment (const_tree exp, unsigned align);
> +extern unsigned aarch64_stack_alignment (const_tree exp, unsigned align);
> +
> #endif /* GCC_AARCH64_PROTOS_H */
> diff --git a/gcc/config/aarch64/aarch64.h b/gcc/config/aarch64/aarch64.h
> index
> f07b2c49f0d9abd3309afb98499ab7eebcff05bd..64f55f6f94e37bffa6b1e7403274ec5f5d906095
> 100644
> --- a/gcc/config/aarch64/aarch64.h
> +++ b/gcc/config/aarch64/aarch64.h
> @@ -121,24 +121,11 @@
> of LSE instructions. */
> #define TARGET_OUTLINE_ATOMICS (aarch64_flag_outline_atomics)
>
> -/* Align definitions of arrays, unions and structures so that
> - initializations and copies can be made more efficient. This is not
> - ABI-changing, so it only affects places where we can see the
> - definition. Increasing the alignment tends to introduce padding,
> - so don't do this when optimizing for size/conserving stack space. */
> -#define AARCH64_EXPAND_ALIGNMENT(COND, EXP, ALIGN) \
> - (((COND) && ((ALIGN) < BITS_PER_WORD)
> \
> - && (TREE_CODE (EXP) == ARRAY_TYPE \
> - || TREE_CODE (EXP) == UNION_TYPE \
> - || TREE_CODE (EXP) == RECORD_TYPE)) ? BITS_PER_WORD : (ALIGN))
> -
> -/* Align global data. */
> -#define DATA_ALIGNMENT(EXP, ALIGN) \
> - AARCH64_EXPAND_ALIGNMENT (!optimize_size, EXP, ALIGN)
> -
> -/* Similarly, make sure that objects on the stack are sensibly aligned. */
> -#define LOCAL_ALIGNMENT(EXP, ALIGN) \
> - AARCH64_EXPAND_ALIGNMENT (!flag_conserve_stack, EXP, ALIGN)
> +/* Align global data as an optimization. */
> +#define DATA_ALIGNMENT(EXP, ALIGN) aarch64_data_alignment (EXP, ALIGN)
> +
> +/* Align stack data as an optimization. */
> +#define LOCAL_ALIGNMENT(EXP, ALIGN) aarch64_stack_alignment (EXP, ALIGN)
>
> #define STRUCTURE_SIZE_BOUNDARY 8
>
> diff --git a/gcc/config/aarch64/aarch64.cc b/gcc/config/aarch64/aarch64.cc
> index
> c78845fc27e6d6a8a1631b487b19fb3143a231ac..5369129d4a405afe5a760081149da1347e7b8842
> 100644
> --- a/gcc/config/aarch64/aarch64.cc
> +++ b/gcc/config/aarch64/aarch64.cc
> @@ -2651,6 +2651,60 @@ aarch64_constant_alignment (const_tree exp,
> HOST_WIDE_INT align)
> return align;
> }
>
> +/* Align definitions of arrays, unions and structures so that
> + initializations and copies can be made more efficient. This is not
> + ABI-changing, so it only affects places where we can see the
> + definition. Increasing the alignment tends to introduce padding,
> + so don't do this when optimizing for size/conserving stack space. */
> +
> +unsigned
> +aarch64_data_alignment (const_tree type, unsigned align)
> +{
> + if (optimize_size)
> + return align;
> +
> + if (AGGREGATE_TYPE_P (type))
> + {
> + unsigned HOST_WIDE_INT size = 0;
> +
> + if (TYPE_SIZE (type) && TREE_CODE (TYPE_SIZE (type)) == INTEGER_CST
> + && tree_fits_uhwi_p (TYPE_SIZE (type)))
> + size = tree_to_uhwi (TYPE_SIZE (type));
> +
> + /* Align small structs/arrays to 32 bits, or 64 bits if larger. */
> + if (align < 32 && size <= 32)
> + align = 32;
> + else if (align < 64)
> + align = 64;
> + }
> +
> + return align;
> +}
> +
> +unsigned
> +aarch64_stack_alignment (const_tree type, unsigned align)
> +{
> + if (flag_conserve_stack)
> + return align;
> +
> + if (AGGREGATE_TYPE_P (type))
> + {
> + unsigned HOST_WIDE_INT size = 0;
> +
> + if (TYPE_SIZE (type) && TREE_CODE (TYPE_SIZE (type)) == INTEGER_CST
> + && tree_fits_uhwi_p (TYPE_SIZE (type)))
> + size = tree_to_uhwi (TYPE_SIZE (type));
> +
> + /* Align small structs/arrays to 32 bits, or 64 bits if larger. */
> + if (align < 32 && size <= 32)
> + align = 32;
> + else if (align < 64)
> + align = 64;
> + }
> +
> + return align;
> +}
> +
> /* Return true if calls to DECL should be treated as
> long-calls (ie called via a register). */
> static bool
