[RFC,2/2] softmmu: simplify helper_*_st_name with smmu_helper(do_unl_store)

From: Alvise Rigo <a.rigo@virtualopensystems.com>

Attempting to simplify the helper_*_st_name, wrap the
do_unaligned_access code into an shared inline function. As this also
removes the goto statement the inline code is expanded twice in each
helper.

Suggested-by: Jani Kokkonen <jani.kokkonen@huawei.com>
Suggested-by: Claudio Fontana <claudio.fontana@huawei.com>
CC: Alvise Rigo <a.rigo@virtualopensystems.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

---
v2
  - based on original patch from Alvise
  - uses a single shared inline function to reduce duplication
---
 softmmu_template.h | 75 ++++++++++++++++++++++++++++--------------------------
 1 file changed, 39 insertions(+), 36 deletions(-)

On Fri, Jan 8, 2016 at 4:53 PM, Alex Bennée <alex.bennee@linaro.org> wrote:
> From: Alvise Rigo <a.rigo@virtualopensystems.com>
>
> Attempting to simplify the helper_*_st_name, wrap the
> do_unaligned_access code into an shared inline function. As this also
> removes the goto statement the inline code is expanded twice in each
> helper.
>
> Suggested-by: Jani Kokkonen <jani.kokkonen@huawei.com>
> Suggested-by: Claudio Fontana <claudio.fontana@huawei.com>
> CC: Alvise Rigo <a.rigo@virtualopensystems.com>
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
>
> ---
> v2
>   - based on original patch from Alvise
>   - uses a single shared inline function to reduce duplication
> ---
>  softmmu_template.h | 75 ++++++++++++++++++++++++++++--------------------------
>  1 file changed, 39 insertions(+), 36 deletions(-)
>
> diff --git a/softmmu_template.h b/softmmu_template.h
> index 0074bd7..ac0b4ac 100644
> --- a/softmmu_template.h
> +++ b/softmmu_template.h
> @@ -159,6 +159,39 @@ static inline int smmu_helper(victim_tlb_hit) (const bool is_read, CPUArchState
>  }
>
>  #ifndef SOFTMMU_CODE_ACCESS
> +
> +static inline void smmu_helper(do_unl_store)(CPUArchState *env,
> +                                             bool little_endian,
> +                                             DATA_TYPE val,
> +                                             target_ulong addr,
> +                                             TCGMemOpIdx oi,
> +                                             unsigned mmu_idx,
> +                                             uintptr_t retaddr)
> +{
> +    int i;
> +
> +    if ((get_memop(oi) & MO_AMASK) == MO_ALIGN) {
> +        cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
> +                             mmu_idx, retaddr);
> +    }
> +    /* Note: relies on the fact that tlb_fill() does not remove the
> +     * previous page from the TLB cache.  */
> +    for (i = DATA_SIZE - 1; i >= 0; i--) {
> +        uint8_t val8;
> +        if (little_endian) {
> +            /* Little-endian extract.  */
> +            val8 = val >> (i * 8);
> +        } else {
> +            /* Big-endian extract.  */
> +            val8 = val >> (((DATA_SIZE - 1) * 8) - (i * 8));
> +        }
> +        /* Note the adjustment at the beginning of the function.
> +           Undo that for the recursion.  */
> +        glue(helper_ret_stb, MMUSUFFIX)(env, addr + i, val8,
> +                                        oi, retaddr + GETPC_ADJ);
> +    }
> +}
> +
>  static inline DATA_TYPE glue(io_read, SUFFIX)(CPUArchState *env,
>                                                CPUIOTLBEntry *iotlbentry,
>                                                target_ulong addr,
> @@ -416,7 +449,8 @@ void helper_le_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
>      if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
>          CPUIOTLBEntry *iotlbentry;
>          if ((addr & (DATA_SIZE - 1)) != 0) {
> -            goto do_unaligned_access;
> +            smmu_helper(do_unl_store)(env, true, val, addr, oi, mmu_idx, retaddr);
> +            return;
>          }
>          iotlbentry = &env->iotlb[mmu_idx][index];
>
> @@ -431,23 +465,7 @@ void helper_le_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
>      if (DATA_SIZE > 1
>          && unlikely((addr & ~TARGET_PAGE_MASK) + DATA_SIZE - 1
>                       >= TARGET_PAGE_SIZE)) {
> -        int i;
> -    do_unaligned_access:
> -        if ((get_memop(oi) & MO_AMASK) == MO_ALIGN) {
> -            cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
> -                                 mmu_idx, retaddr);
> -        }
> -        /* XXX: not efficient, but simple */
> -        /* Note: relies on the fact that tlb_fill() does not remove the
> -         * previous page from the TLB cache.  */
> -        for (i = DATA_SIZE - 1; i >= 0; i--) {
> -            /* Little-endian extract.  */
> -            uint8_t val8 = val >> (i * 8);
> -            /* Note the adjustment at the beginning of the function.
> -               Undo that for the recursion.  */
> -            glue(helper_ret_stb, MMUSUFFIX)(env, addr + i, val8,
> -                                            oi, retaddr + GETPC_ADJ);
> -        }
> +        smmu_helper(do_unl_store)(env, true, val, addr, oi, mmu_idx, retaddr);
>          return;
>      }
>
> @@ -496,7 +514,8 @@ void helper_be_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
>      if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
>          CPUIOTLBEntry *iotlbentry;
>          if ((addr & (DATA_SIZE - 1)) != 0) {
> -            goto do_unaligned_access;
> +            smmu_helper(do_unl_store)(env, false, val, addr, oi, mmu_idx, retaddr);
> +            return;
>          }
>          iotlbentry = &env->iotlb[mmu_idx][index];
>
> @@ -511,23 +530,7 @@ void helper_be_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
>      if (DATA_SIZE > 1
>          && unlikely((addr & ~TARGET_PAGE_MASK) + DATA_SIZE - 1
>                       >= TARGET_PAGE_SIZE)) {
> -        int i;
> -    do_unaligned_access:
> -        if ((get_memop(oi) & MO_AMASK) == MO_ALIGN) {
> -            cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
> -                                 mmu_idx, retaddr);
> -        }
> -        /* XXX: not efficient, but simple */
> -        /* Note: relies on the fact that tlb_fill() does not remove the
> -         * previous page from the TLB cache.  */
> -        for (i = DATA_SIZE - 1; i >= 0; i--) {
> -            /* Big-endian extract.  */
> -            uint8_t val8 = val >> (((DATA_SIZE - 1) * 8) - (i * 8));
> -            /* Note the adjustment at the beginning of the function.
> -               Undo that for the recursion.  */
> -            glue(helper_ret_stb, MMUSUFFIX)(env, addr + i, val8,
> -                                            oi, retaddr + GETPC_ADJ);
> -        }
> +        smmu_helper(do_unl_store)(env, false, val, addr, oi, mmu_idx, retaddr);
>          return;
>      }
>
> --
> 2.6.4
>

This approach makes sense to me, given that the leg of the *if*
statement is actually inlined depending on the (constant) value of
little_endian.

The thing not convincing me is the fact that we are not imposing the
inlining, but we are relying on the compiler optimizations to do it. I
wonder if this will always happen, even with other compilers (clang).

alvise