Re: [Qemu-devel] [PATCH v4 11/35] cputlb: Move most of iotlb code out of line

[Alex Bennée]

Richard Henderson  writes:

> Saves 2k code size off of a cold path.
>
> Signed-off-by: Richard Henderson 

Reviewed-by: Alex Bennée 

> ---
>  cputlb.c   | 37 +
>  softmmu_template.h | 52 ++--
>  2 files changed, 47 insertions(+), 42 deletions(-)
>
> diff --git a/cputlb.c b/cputlb.c
> index 0c9b77b..1bee47d 100644
> --- a/cputlb.c
> +++ b/cputlb.c
> @@ -498,6 +498,43 @@ tb_page_addr_t get_page_addr_code(CPUArchState *env1, 
> target_ulong addr)
>  return qemu_ram_addr_from_host_nofail(p);
>  }
>
> +static uint64_t io_readx(CPUArchState *env, CPUIOTLBEntry *iotlbentry,
> + target_ulong addr, uintptr_t retaddr, int size)
> +{
> +CPUState *cpu = ENV_GET_CPU(env);
> +hwaddr physaddr = iotlbentry->addr;
> +MemoryRegion *mr = iotlb_to_region(cpu, physaddr, iotlbentry->attrs);
> +uint64_t val;
> +
> +physaddr = (physaddr & TARGET_PAGE_MASK) + addr;
> +cpu->mem_io_pc = retaddr;
> +if (mr != _mem_rom && mr != _mem_notdirty && !cpu->can_do_io) {
> +cpu_io_recompile(cpu, retaddr);
> +}
> +
> +cpu->mem_io_vaddr = addr;
> +memory_region_dispatch_read(mr, physaddr, , size, iotlbentry->attrs);
> +return val;
> +}
> +
> +static void io_writex(CPUArchState *env, CPUIOTLBEntry *iotlbentry,
> +  uint64_t val, target_ulong addr,
> +  uintptr_t retaddr, int size)
> +{
> +CPUState *cpu = ENV_GET_CPU(env);
> +hwaddr physaddr = iotlbentry->addr;
> +MemoryRegion *mr = iotlb_to_region(cpu, physaddr, iotlbentry->attrs);
> +
> +physaddr = (physaddr & TARGET_PAGE_MASK) + addr;
> +if (mr != _mem_rom && mr != _mem_notdirty && !cpu->can_do_io) {
> +cpu_io_recompile(cpu, retaddr);
> +}
> +
> +cpu->mem_io_vaddr = addr;
> +cpu->mem_io_pc = retaddr;
> +memory_region_dispatch_write(mr, physaddr, val, size, iotlbentry->attrs);
> +}
> +
>  /* Return true if ADDR is present in the victim tlb, and has been copied
> back to the main tlb.  */
>  static bool victim_tlb_hit(CPUArchState *env, size_t mmu_idx, size_t index,
> diff --git a/softmmu_template.h b/softmmu_template.h
> index b9532a4..035ffc8 100644
> --- a/softmmu_template.h
> +++ b/softmmu_template.h
> @@ -112,25 +112,12 @@
>
>  #ifndef SOFTMMU_CODE_ACCESS
>  static inline DATA_TYPE glue(io_read, SUFFIX)(CPUArchState *env,
> -  CPUIOTLBEntry *iotlbentry,
> +  size_t mmu_idx, size_t index,
>target_ulong addr,
>uintptr_t retaddr)
>  {
> -uint64_t val;
> -CPUState *cpu = ENV_GET_CPU(env);
> -hwaddr physaddr = iotlbentry->addr;
> -MemoryRegion *mr = iotlb_to_region(cpu, physaddr, iotlbentry->attrs);
> -
> -physaddr = (physaddr & TARGET_PAGE_MASK) + addr;
> -cpu->mem_io_pc = retaddr;
> -if (mr != _mem_rom && mr != _mem_notdirty && !cpu->can_do_io) {
> -cpu_io_recompile(cpu, retaddr);
> -}
> -
> -cpu->mem_io_vaddr = addr;
> -memory_region_dispatch_read(mr, physaddr, , DATA_SIZE,
> -iotlbentry->attrs);
> -return val;
> +CPUIOTLBEntry *iotlbentry = >iotlb[mmu_idx][index];
> +return io_readx(env, iotlbentry, addr, retaddr, DATA_SIZE);
>  }
>  #endif
>
> @@ -161,15 +148,13 @@ WORD_TYPE helper_le_ld_name(CPUArchState *env, 
> target_ulong addr,
>
>  /* Handle an IO access.  */
>  if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
> -CPUIOTLBEntry *iotlbentry;
>  if ((addr & (DATA_SIZE - 1)) != 0) {
>  goto do_unaligned_access;
>  }
> -iotlbentry = >iotlb[mmu_idx][index];
>
>  /* ??? Note that the io helpers always read data in the target
> byte ordering.  We should push the LE/BE request down into io.  */
> -res = glue(io_read, SUFFIX)(env, iotlbentry, addr, retaddr);
> +res = glue(io_read, SUFFIX)(env, mmu_idx, index, addr, retaddr);
>  res = TGT_LE(res);
>  return res;
>  }
> @@ -230,15 +215,13 @@ WORD_TYPE helper_be_ld_name(CPUArchState *env, 
> target_ulong addr,
>
>  /* Handle an IO access.  */
>  if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
> -CPUIOTLBEntry *iotlbentry;
>  if ((addr & (DATA_SIZE - 1)) != 0) {
>  goto do_unaligned_access;
>  }
> -iotlbentry = >iotlb[mmu_idx][index];
>
>  /* ??? Note that the io helpers always read data in the target
> byte ordering.  We should push the LE/BE request down into io.  */
> -res = glue(io_read, SUFFIX)(env, iotlbentry, addr, retaddr);
> +res = glue(io_read, SUFFIX)(env, mmu_idx, index, addr, retaddr);
>  res = TGT_BE(res);
>  return res;
>  }
> @@ -289,24 

[Qemu-devel] [PATCH v4 11/35] cputlb: Move most of iotlb code out of line

[Richard Henderson]

Saves 2k code size off of a cold path.

Signed-off-by: Richard Henderson 
---
 cputlb.c   | 37 +
 softmmu_template.h | 52 ++--
 2 files changed, 47 insertions(+), 42 deletions(-)

diff --git a/cputlb.c b/cputlb.c
index 0c9b77b..1bee47d 100644
--- a/cputlb.c
+++ b/cputlb.c
@@ -498,6 +498,43 @@ tb_page_addr_t get_page_addr_code(CPUArchState *env1, 
target_ulong addr)
 return qemu_ram_addr_from_host_nofail(p);
 }
 
+static uint64_t io_readx(CPUArchState *env, CPUIOTLBEntry *iotlbentry,
+ target_ulong addr, uintptr_t retaddr, int size)
+{
+CPUState *cpu = ENV_GET_CPU(env);
+hwaddr physaddr = iotlbentry->addr;
+MemoryRegion *mr = iotlb_to_region(cpu, physaddr, iotlbentry->attrs);
+uint64_t val;
+
+physaddr = (physaddr & TARGET_PAGE_MASK) + addr;
+cpu->mem_io_pc = retaddr;
+if (mr != _mem_rom && mr != _mem_notdirty && !cpu->can_do_io) {
+cpu_io_recompile(cpu, retaddr);
+}
+
+cpu->mem_io_vaddr = addr;
+memory_region_dispatch_read(mr, physaddr, , size, iotlbentry->attrs);
+return val;
+}
+
+static void io_writex(CPUArchState *env, CPUIOTLBEntry *iotlbentry,
+  uint64_t val, target_ulong addr,
+  uintptr_t retaddr, int size)
+{
+CPUState *cpu = ENV_GET_CPU(env);
+hwaddr physaddr = iotlbentry->addr;
+MemoryRegion *mr = iotlb_to_region(cpu, physaddr, iotlbentry->attrs);
+
+physaddr = (physaddr & TARGET_PAGE_MASK) + addr;
+if (mr != _mem_rom && mr != _mem_notdirty && !cpu->can_do_io) {
+cpu_io_recompile(cpu, retaddr);
+}
+
+cpu->mem_io_vaddr = addr;
+cpu->mem_io_pc = retaddr;
+memory_region_dispatch_write(mr, physaddr, val, size, iotlbentry->attrs);
+}
+
 /* Return true if ADDR is present in the victim tlb, and has been copied
back to the main tlb.  */
 static bool victim_tlb_hit(CPUArchState *env, size_t mmu_idx, size_t index,
diff --git a/softmmu_template.h b/softmmu_template.h
index b9532a4..035ffc8 100644
--- a/softmmu_template.h
+++ b/softmmu_template.h
@@ -112,25 +112,12 @@
 
 #ifndef SOFTMMU_CODE_ACCESS
 static inline DATA_TYPE glue(io_read, SUFFIX)(CPUArchState *env,
-  CPUIOTLBEntry *iotlbentry,
+  size_t mmu_idx, size_t index,
   target_ulong addr,
   uintptr_t retaddr)
 {
-uint64_t val;
-CPUState *cpu = ENV_GET_CPU(env);
-hwaddr physaddr = iotlbentry->addr;
-MemoryRegion *mr = iotlb_to_region(cpu, physaddr, iotlbentry->attrs);
-
-physaddr = (physaddr & TARGET_PAGE_MASK) + addr;
-cpu->mem_io_pc = retaddr;
-if (mr != _mem_rom && mr != _mem_notdirty && !cpu->can_do_io) {
-cpu_io_recompile(cpu, retaddr);
-}
-
-cpu->mem_io_vaddr = addr;
-memory_region_dispatch_read(mr, physaddr, , DATA_SIZE,
-iotlbentry->attrs);
-return val;
+CPUIOTLBEntry *iotlbentry = >iotlb[mmu_idx][index];
+return io_readx(env, iotlbentry, addr, retaddr, DATA_SIZE);
 }
 #endif
 
@@ -161,15 +148,13 @@ WORD_TYPE helper_le_ld_name(CPUArchState *env, 
target_ulong addr,
 
 /* Handle an IO access.  */
 if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
-CPUIOTLBEntry *iotlbentry;
 if ((addr & (DATA_SIZE - 1)) != 0) {
 goto do_unaligned_access;
 }
-iotlbentry = >iotlb[mmu_idx][index];
 
 /* ??? Note that the io helpers always read data in the target
byte ordering.  We should push the LE/BE request down into io.  */
-res = glue(io_read, SUFFIX)(env, iotlbentry, addr, retaddr);
+res = glue(io_read, SUFFIX)(env, mmu_idx, index, addr, retaddr);
 res = TGT_LE(res);
 return res;
 }
@@ -230,15 +215,13 @@ WORD_TYPE helper_be_ld_name(CPUArchState *env, 
target_ulong addr,
 
 /* Handle an IO access.  */
 if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
-CPUIOTLBEntry *iotlbentry;
 if ((addr & (DATA_SIZE - 1)) != 0) {
 goto do_unaligned_access;
 }
-iotlbentry = >iotlb[mmu_idx][index];
 
 /* ??? Note that the io helpers always read data in the target
byte ordering.  We should push the LE/BE request down into io.  */
-res = glue(io_read, SUFFIX)(env, iotlbentry, addr, retaddr);
+res = glue(io_read, SUFFIX)(env, mmu_idx, index, addr, retaddr);
 res = TGT_BE(res);
 return res;
 }
@@ -289,24 +272,13 @@ WORD_TYPE helper_be_lds_name(CPUArchState *env, 
target_ulong addr,
 #endif
 
 static inline void glue(io_write, SUFFIX)(CPUArchState *env,
-  CPUIOTLBEntry *iotlbentry,
+  size_t mmu_idx, size_t index,