Improve the performance of RISC-V vector unit-stride ld/st instructions

[RFC PATCH 5/6] accel/tcg: Inline do_ld1_mmu function

Posted by Max Chou 8 months, 2 weeks ago

Signed-off-by: Max Chou <max.chou@sifive.com>
---
 accel/tcg/user-exec.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/accel/tcg/user-exec.c b/accel/tcg/user-exec.c
index c5453810eee..803c271df11 100644
--- a/accel/tcg/user-exec.c
+++ b/accel/tcg/user-exec.c
@@ -963,8 +963,9 @@ static inline QEMU_ALWAYS_INLINE void *cpu_mmu_lookup(CPUState *cpu,
 
 #include "ldst_atomicity.c.inc"
 
-static uint8_t do_ld1_mmu(CPUState *cpu, vaddr addr, MemOpIdx oi,
-                          uintptr_t ra, MMUAccessType access_type)
+static inline QEMU_ALWAYS_INLINE uint8_t do_ld1_mmu(CPUState *cpu, vaddr addr,
+                                                    MemOpIdx oi, uintptr_t ra,
+                                                    MMUAccessType access_type)
 {
     void *haddr;
     uint8_t ret;
-- 
2.34.1

Re: [RFC PATCH 5/6] accel/tcg: Inline do_ld1_mmu function

Posted by Richard Henderson 8 months, 2 weeks ago

On 2/15/24 09:28, Max Chou wrote:
> Signed-off-by: Max Chou <max.chou@sifive.com>
> ---
>   accel/tcg/user-exec.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/accel/tcg/user-exec.c b/accel/tcg/user-exec.c
> index c5453810eee..803c271df11 100644
> --- a/accel/tcg/user-exec.c
> +++ b/accel/tcg/user-exec.c
> @@ -963,8 +963,9 @@ static inline QEMU_ALWAYS_INLINE void *cpu_mmu_lookup(CPUState *cpu,
>   
>   #include "ldst_atomicity.c.inc"
>   
> -static uint8_t do_ld1_mmu(CPUState *cpu, vaddr addr, MemOpIdx oi,
> -                          uintptr_t ra, MMUAccessType access_type)
> +static inline QEMU_ALWAYS_INLINE uint8_t do_ld1_mmu(CPUState *cpu, vaddr addr,
> +                                                    MemOpIdx oi, uintptr_t ra,
> +                                                    MMUAccessType access_type)
>   {
>       void *haddr;
>       uint8_t ret;

I expect the small functions that use this to tail-call into this common function.  I do 
not believe that inlining helps.  Same with do_st1_mmu.


r~