qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
* [PATCH 1/2] target/riscv: optimize condition assign for scale < 0
@ 2022-03-25  8:59 Weiwei Li
  2022-03-25  8:59 ` [PATCH 2/2] target/riscv: optimize helper for vmv<nr>r.v Weiwei Li
                   ` (3 more replies)
  0 siblings, 4 replies; 7+ messages in thread
From: Weiwei Li @ 2022-03-25  8:59 UTC (permalink / raw)
  To: richard.henderson, palmer, alistair.francis, bin.meng, qemu-riscv,
	qemu-devel
  Cc: wangjunqiang, Weiwei Li, lazyparser

for some cases, scale is always equal or less than 0, since lmul is not larger than 3

Signed-off-by: Weiwei Li <liweiwei@iscas.ac.cn>
Signed-off-by: Junqiang Wang <wangjunqiang@iscas.ac.cn>
---
 target/riscv/insn_trans/trans_rvv.c.inc | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc
index 4ea7e41e1a..2878ca3132 100644
--- a/target/riscv/insn_trans/trans_rvv.c.inc
+++ b/target/riscv/insn_trans/trans_rvv.c.inc
@@ -1198,7 +1198,7 @@ GEN_LDST_WHOLE_TRANS(vs8r_v, 8, true)
 static inline uint32_t MAXSZ(DisasContext *s)
 {
     int scale = s->lmul - 3;
-    return scale < 0 ? s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen << scale;
+    return s->cfg_ptr->vlen >> -scale;
 }
 
 static bool opivv_check(DisasContext *s, arg_rmrr *a)
@@ -3597,8 +3597,7 @@ static bool trans_vrgather_vx(DisasContext *s, arg_rmrr *a)
 
     if (a->vm && s->vl_eq_vlmax) {
         int scale = s->lmul - (s->sew + 3);
-        int vlmax = scale < 0 ?
-                       s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen << scale;
+        int vlmax = s->cfg_ptr->vlen >> -scale;
         TCGv_i64 dest = tcg_temp_new_i64();
 
         if (a->rs1 == 0) {
@@ -3630,8 +3629,7 @@ static bool trans_vrgather_vi(DisasContext *s, arg_rmrr *a)
 
     if (a->vm && s->vl_eq_vlmax) {
         int scale = s->lmul - (s->sew + 3);
-        int vlmax = scale < 0 ?
-                       s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen << scale;
+        int vlmax = s->cfg_ptr->vlen >> -scale;
         if (a->rs1 >= vlmax) {
             tcg_gen_gvec_dup_imm(MO_64, vreg_ofs(s, a->rd),
                                  MAXSZ(s), MAXSZ(s), 0);
-- 
2.17.1



^ permalink raw reply related	[flat|nested] 7+ messages in thread

* [PATCH 2/2] target/riscv: optimize helper for vmv<nr>r.v
  2022-03-25  8:59 [PATCH 1/2] target/riscv: optimize condition assign for scale < 0 Weiwei Li
@ 2022-03-25  8:59 ` Weiwei Li
  2022-03-25  9:26   ` Frank Chang
  2022-03-28  1:11   ` Alistair Francis
  2022-03-25  9:26 ` [PATCH 1/2] target/riscv: optimize condition assign for scale < 0 Frank Chang
                   ` (2 subsequent siblings)
  3 siblings, 2 replies; 7+ messages in thread
From: Weiwei Li @ 2022-03-25  8:59 UTC (permalink / raw)
  To: richard.henderson, palmer, alistair.francis, bin.meng, qemu-riscv,
	qemu-devel
  Cc: wangjunqiang, Weiwei Li, lazyparser

LEN is not used for GEN_VEXT_VMV_WHOLE macro, so vmv<nr>r.v can share
the same helper 

Signed-off-by: Weiwei Li <liweiwei@iscas.ac.cn>
Signed-off-by: Junqiang Wang <wangjunqiang@iscas.ac.cn>
---
 target/riscv/helper.h                   |  5 +----
 target/riscv/insn_trans/trans_rvv.c.inc | 17 +++++----------
 target/riscv/vector_helper.c            | 29 ++++++++++---------------
 3 files changed, 18 insertions(+), 33 deletions(-)

diff --git a/target/riscv/helper.h b/target/riscv/helper.h
index 26bbab2fab..a669d0187b 100644
--- a/target/riscv/helper.h
+++ b/target/riscv/helper.h
@@ -1086,10 +1086,7 @@ DEF_HELPER_6(vcompress_vm_h, void, ptr, ptr, ptr, ptr, env, i32)
 DEF_HELPER_6(vcompress_vm_w, void, ptr, ptr, ptr, ptr, env, i32)
 DEF_HELPER_6(vcompress_vm_d, void, ptr, ptr, ptr, ptr, env, i32)
 
-DEF_HELPER_4(vmv1r_v, void, ptr, ptr, env, i32)
-DEF_HELPER_4(vmv2r_v, void, ptr, ptr, env, i32)
-DEF_HELPER_4(vmv4r_v, void, ptr, ptr, env, i32)
-DEF_HELPER_4(vmv8r_v, void, ptr, ptr, env, i32)
+DEF_HELPER_4(vmvr_v, void, ptr, ptr, env, i32)
 
 DEF_HELPER_5(vzext_vf2_h, void, ptr, ptr, ptr, env, i32)
 DEF_HELPER_5(vzext_vf2_w, void, ptr, ptr, ptr, env, i32)
diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc
index 2878ca3132..ec7c0e0d36 100644
--- a/target/riscv/insn_trans/trans_rvv.c.inc
+++ b/target/riscv/insn_trans/trans_rvv.c.inc
@@ -3695,7 +3695,7 @@ static bool trans_vcompress_vm(DisasContext *s, arg_r *a)
  * Whole Vector Register Move Instructions ignore vtype and vl setting.
  * Thus, we don't need to check vill bit. (Section 16.6)
  */
-#define GEN_VMV_WHOLE_TRANS(NAME, LEN, SEQ)                             \
+#define GEN_VMV_WHOLE_TRANS(NAME, LEN)                             \
 static bool trans_##NAME(DisasContext *s, arg_##NAME * a)               \
 {                                                                       \
     if (require_rvv(s) &&                                               \
@@ -3710,13 +3710,8 @@ static bool trans_##NAME(DisasContext *s, arg_##NAME * a)               \
         } else {                                                        \
             TCGLabel *over = gen_new_label();                           \
             tcg_gen_brcondi_tl(TCG_COND_GEU, cpu_vstart, maxsz, over);  \
-                                                                        \
-            static gen_helper_gvec_2_ptr * const fns[4] = {             \
-                gen_helper_vmv1r_v, gen_helper_vmv2r_v,                 \
-                gen_helper_vmv4r_v, gen_helper_vmv8r_v,                 \
-            };                                                          \
             tcg_gen_gvec_2_ptr(vreg_ofs(s, a->rd), vreg_ofs(s, a->rs2), \
-                               cpu_env, maxsz, maxsz, 0, fns[SEQ]);     \
+                               cpu_env, maxsz, maxsz, 0, gen_helper_vmvr_v); \
             mark_vs_dirty(s);                                           \
             gen_set_label(over);                                        \
         }                                                               \
@@ -3725,10 +3720,10 @@ static bool trans_##NAME(DisasContext *s, arg_##NAME * a)               \
     return false;                                                       \
 }
 
-GEN_VMV_WHOLE_TRANS(vmv1r_v, 1, 0)
-GEN_VMV_WHOLE_TRANS(vmv2r_v, 2, 1)
-GEN_VMV_WHOLE_TRANS(vmv4r_v, 4, 2)
-GEN_VMV_WHOLE_TRANS(vmv8r_v, 8, 3)
+GEN_VMV_WHOLE_TRANS(vmv1r_v, 1)
+GEN_VMV_WHOLE_TRANS(vmv2r_v, 2)
+GEN_VMV_WHOLE_TRANS(vmv4r_v, 4)
+GEN_VMV_WHOLE_TRANS(vmv8r_v, 8)
 
 static bool int_ext_check(DisasContext *s, arg_rmr *a, uint8_t div)
 {
diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
index 3bd4aac9c9..1d4982ef7f 100644
--- a/target/riscv/vector_helper.c
+++ b/target/riscv/vector_helper.c
@@ -4888,25 +4888,18 @@ GEN_VEXT_VCOMPRESS_VM(vcompress_vm_w, uint32_t, H4)
 GEN_VEXT_VCOMPRESS_VM(vcompress_vm_d, uint64_t, H8)
 
 /* Vector Whole Register Move */
-#define GEN_VEXT_VMV_WHOLE(NAME, LEN)                      \
-void HELPER(NAME)(void *vd, void *vs2, CPURISCVState *env, \
-                  uint32_t desc)                           \
-{                                                          \
-    /* EEW = 8 */                                          \
-    uint32_t maxsz = simd_maxsz(desc);                     \
-    uint32_t i = env->vstart;                              \
-                                                           \
-    memcpy((uint8_t *)vd + H1(i),                          \
-           (uint8_t *)vs2 + H1(i),                         \
-           maxsz - env->vstart);                           \
-                                                           \
-    env->vstart = 0;                                       \
-}
+void HELPER(vmvr_v)(void *vd, void *vs2, CPURISCVState *env, uint32_t desc)
+{
+    /* EEW = 8 */
+    uint32_t maxsz = simd_maxsz(desc);
+    uint32_t i = env->vstart;
+
+    memcpy((uint8_t *)vd + H1(i),
+           (uint8_t *)vs2 + H1(i),
+           maxsz - env->vstart);
 
-GEN_VEXT_VMV_WHOLE(vmv1r_v, 1)
-GEN_VEXT_VMV_WHOLE(vmv2r_v, 2)
-GEN_VEXT_VMV_WHOLE(vmv4r_v, 4)
-GEN_VEXT_VMV_WHOLE(vmv8r_v, 8)
+    env->vstart = 0;
+}
 
 /* Vector Integer Extension */
 #define GEN_VEXT_INT_EXT(NAME, ETYPE, DTYPE, HD, HS1)            \
-- 
2.17.1



^ permalink raw reply related	[flat|nested] 7+ messages in thread

* Re: [PATCH 2/2] target/riscv: optimize helper for vmv<nr>r.v
  2022-03-25  8:59 ` [PATCH 2/2] target/riscv: optimize helper for vmv<nr>r.v Weiwei Li
@ 2022-03-25  9:26   ` Frank Chang
  2022-03-28  1:11   ` Alistair Francis
  1 sibling, 0 replies; 7+ messages in thread
From: Frank Chang @ 2022-03-25  9:26 UTC (permalink / raw)
  To: Weiwei Li
  Cc: lazyparser, open list:RISC-V, wangjunqiang, Bin Meng,
	Richard Henderson, qemu-devel@nongnu.org Developers,
	Palmer Dabbelt, Alistair Francis

[-- Attachment #1: Type: text/plain, Size: 5709 bytes --]

Reviewed-by: Frank Chang <frank.chang@sifive.com>

On Fri, Mar 25, 2022 at 5:00 PM Weiwei Li <liweiwei@iscas.ac.cn> wrote:

> LEN is not used for GEN_VEXT_VMV_WHOLE macro, so vmv<nr>r.v can share
> the same helper
>
> Signed-off-by: Weiwei Li <liweiwei@iscas.ac.cn>
> Signed-off-by: Junqiang Wang <wangjunqiang@iscas.ac.cn>
> ---
>  target/riscv/helper.h                   |  5 +----
>  target/riscv/insn_trans/trans_rvv.c.inc | 17 +++++----------
>  target/riscv/vector_helper.c            | 29 ++++++++++---------------
>  3 files changed, 18 insertions(+), 33 deletions(-)
>
> diff --git a/target/riscv/helper.h b/target/riscv/helper.h
> index 26bbab2fab..a669d0187b 100644
> --- a/target/riscv/helper.h
> +++ b/target/riscv/helper.h
> @@ -1086,10 +1086,7 @@ DEF_HELPER_6(vcompress_vm_h, void, ptr, ptr, ptr,
> ptr, env, i32)
>  DEF_HELPER_6(vcompress_vm_w, void, ptr, ptr, ptr, ptr, env, i32)
>  DEF_HELPER_6(vcompress_vm_d, void, ptr, ptr, ptr, ptr, env, i32)
>
> -DEF_HELPER_4(vmv1r_v, void, ptr, ptr, env, i32)
> -DEF_HELPER_4(vmv2r_v, void, ptr, ptr, env, i32)
> -DEF_HELPER_4(vmv4r_v, void, ptr, ptr, env, i32)
> -DEF_HELPER_4(vmv8r_v, void, ptr, ptr, env, i32)
> +DEF_HELPER_4(vmvr_v, void, ptr, ptr, env, i32)
>
>  DEF_HELPER_5(vzext_vf2_h, void, ptr, ptr, ptr, env, i32)
>  DEF_HELPER_5(vzext_vf2_w, void, ptr, ptr, ptr, env, i32)
> diff --git a/target/riscv/insn_trans/trans_rvv.c.inc
> b/target/riscv/insn_trans/trans_rvv.c.inc
> index 2878ca3132..ec7c0e0d36 100644
> --- a/target/riscv/insn_trans/trans_rvv.c.inc
> +++ b/target/riscv/insn_trans/trans_rvv.c.inc
> @@ -3695,7 +3695,7 @@ static bool trans_vcompress_vm(DisasContext *s,
> arg_r *a)
>   * Whole Vector Register Move Instructions ignore vtype and vl setting.
>   * Thus, we don't need to check vill bit. (Section 16.6)
>   */
> -#define GEN_VMV_WHOLE_TRANS(NAME, LEN, SEQ)                             \
> +#define GEN_VMV_WHOLE_TRANS(NAME, LEN)                             \
>  static bool trans_##NAME(DisasContext *s, arg_##NAME * a)               \
>  {                                                                       \
>      if (require_rvv(s) &&                                               \
> @@ -3710,13 +3710,8 @@ static bool trans_##NAME(DisasContext *s,
> arg_##NAME * a)               \
>          } else {                                                        \
>              TCGLabel *over = gen_new_label();                           \
>              tcg_gen_brcondi_tl(TCG_COND_GEU, cpu_vstart, maxsz, over);  \
> -                                                                        \
> -            static gen_helper_gvec_2_ptr * const fns[4] = {             \
> -                gen_helper_vmv1r_v, gen_helper_vmv2r_v,                 \
> -                gen_helper_vmv4r_v, gen_helper_vmv8r_v,                 \
> -            };                                                          \
>              tcg_gen_gvec_2_ptr(vreg_ofs(s, a->rd), vreg_ofs(s, a->rs2), \
> -                               cpu_env, maxsz, maxsz, 0, fns[SEQ]);     \
> +                               cpu_env, maxsz, maxsz, 0,
> gen_helper_vmvr_v); \
>              mark_vs_dirty(s);                                           \
>              gen_set_label(over);                                        \
>          }                                                               \
> @@ -3725,10 +3720,10 @@ static bool trans_##NAME(DisasContext *s,
> arg_##NAME * a)               \
>      return false;                                                       \
>  }
>
> -GEN_VMV_WHOLE_TRANS(vmv1r_v, 1, 0)
> -GEN_VMV_WHOLE_TRANS(vmv2r_v, 2, 1)
> -GEN_VMV_WHOLE_TRANS(vmv4r_v, 4, 2)
> -GEN_VMV_WHOLE_TRANS(vmv8r_v, 8, 3)
> +GEN_VMV_WHOLE_TRANS(vmv1r_v, 1)
> +GEN_VMV_WHOLE_TRANS(vmv2r_v, 2)
> +GEN_VMV_WHOLE_TRANS(vmv4r_v, 4)
> +GEN_VMV_WHOLE_TRANS(vmv8r_v, 8)
>
>  static bool int_ext_check(DisasContext *s, arg_rmr *a, uint8_t div)
>  {
> diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
> index 3bd4aac9c9..1d4982ef7f 100644
> --- a/target/riscv/vector_helper.c
> +++ b/target/riscv/vector_helper.c
> @@ -4888,25 +4888,18 @@ GEN_VEXT_VCOMPRESS_VM(vcompress_vm_w, uint32_t, H4)
>  GEN_VEXT_VCOMPRESS_VM(vcompress_vm_d, uint64_t, H8)
>
>  /* Vector Whole Register Move */
> -#define GEN_VEXT_VMV_WHOLE(NAME, LEN)                      \
> -void HELPER(NAME)(void *vd, void *vs2, CPURISCVState *env, \
> -                  uint32_t desc)                           \
> -{                                                          \
> -    /* EEW = 8 */                                          \
> -    uint32_t maxsz = simd_maxsz(desc);                     \
> -    uint32_t i = env->vstart;                              \
> -                                                           \
> -    memcpy((uint8_t *)vd + H1(i),                          \
> -           (uint8_t *)vs2 + H1(i),                         \
> -           maxsz - env->vstart);                           \
> -                                                           \
> -    env->vstart = 0;                                       \
> -}
> +void HELPER(vmvr_v)(void *vd, void *vs2, CPURISCVState *env, uint32_t
> desc)
> +{
> +    /* EEW = 8 */
> +    uint32_t maxsz = simd_maxsz(desc);
> +    uint32_t i = env->vstart;
> +
> +    memcpy((uint8_t *)vd + H1(i),
> +           (uint8_t *)vs2 + H1(i),
> +           maxsz - env->vstart);
>
> -GEN_VEXT_VMV_WHOLE(vmv1r_v, 1)
> -GEN_VEXT_VMV_WHOLE(vmv2r_v, 2)
> -GEN_VEXT_VMV_WHOLE(vmv4r_v, 4)
> -GEN_VEXT_VMV_WHOLE(vmv8r_v, 8)
> +    env->vstart = 0;
> +}
>
>  /* Vector Integer Extension */
>  #define GEN_VEXT_INT_EXT(NAME, ETYPE, DTYPE, HD, HS1)            \
> --
> 2.17.1
>
>
>

[-- Attachment #2: Type: text/html, Size: 7346 bytes --]

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH 1/2] target/riscv: optimize condition assign for scale < 0
  2022-03-25  8:59 [PATCH 1/2] target/riscv: optimize condition assign for scale < 0 Weiwei Li
  2022-03-25  8:59 ` [PATCH 2/2] target/riscv: optimize helper for vmv<nr>r.v Weiwei Li
@ 2022-03-25  9:26 ` Frank Chang
  2022-03-28  1:11 ` Alistair Francis
  2022-03-28  4:01 ` Alistair Francis
  3 siblings, 0 replies; 7+ messages in thread
From: Frank Chang @ 2022-03-25  9:26 UTC (permalink / raw)
  To: Weiwei Li
  Cc: lazyparser, open list:RISC-V, wangjunqiang, Bin Meng,
	Richard Henderson, qemu-devel@nongnu.org Developers,
	Palmer Dabbelt, Alistair Francis

[-- Attachment #1: Type: text/plain, Size: 1952 bytes --]

Reviewed-by: Frank Chang <frank.chang@sifive.com>

On Fri, Mar 25, 2022 at 5:00 PM Weiwei Li <liweiwei@iscas.ac.cn> wrote:

> for some cases, scale is always equal or less than 0, since lmul is not
> larger than 3
>
> Signed-off-by: Weiwei Li <liweiwei@iscas.ac.cn>
> Signed-off-by: Junqiang Wang <wangjunqiang@iscas.ac.cn>
> ---
>  target/riscv/insn_trans/trans_rvv.c.inc | 8 +++-----
>  1 file changed, 3 insertions(+), 5 deletions(-)
>
> diff --git a/target/riscv/insn_trans/trans_rvv.c.inc
> b/target/riscv/insn_trans/trans_rvv.c.inc
> index 4ea7e41e1a..2878ca3132 100644
> --- a/target/riscv/insn_trans/trans_rvv.c.inc
> +++ b/target/riscv/insn_trans/trans_rvv.c.inc
> @@ -1198,7 +1198,7 @@ GEN_LDST_WHOLE_TRANS(vs8r_v, 8, true)
>  static inline uint32_t MAXSZ(DisasContext *s)
>  {
>      int scale = s->lmul - 3;
> -    return scale < 0 ? s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen <<
> scale;
> +    return s->cfg_ptr->vlen >> -scale;
>  }
>
>  static bool opivv_check(DisasContext *s, arg_rmrr *a)
> @@ -3597,8 +3597,7 @@ static bool trans_vrgather_vx(DisasContext *s,
> arg_rmrr *a)
>
>      if (a->vm && s->vl_eq_vlmax) {
>          int scale = s->lmul - (s->sew + 3);
> -        int vlmax = scale < 0 ?
> -                       s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen <<
> scale;
> +        int vlmax = s->cfg_ptr->vlen >> -scale;
>          TCGv_i64 dest = tcg_temp_new_i64();
>
>          if (a->rs1 == 0) {
> @@ -3630,8 +3629,7 @@ static bool trans_vrgather_vi(DisasContext *s,
> arg_rmrr *a)
>
>      if (a->vm && s->vl_eq_vlmax) {
>          int scale = s->lmul - (s->sew + 3);
> -        int vlmax = scale < 0 ?
> -                       s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen <<
> scale;
> +        int vlmax = s->cfg_ptr->vlen >> -scale;
>          if (a->rs1 >= vlmax) {
>              tcg_gen_gvec_dup_imm(MO_64, vreg_ofs(s, a->rd),
>                                   MAXSZ(s), MAXSZ(s), 0);
> --
> 2.17.1
>
>
>

[-- Attachment #2: Type: text/html, Size: 2847 bytes --]

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH 2/2] target/riscv: optimize helper for vmv<nr>r.v
  2022-03-25  8:59 ` [PATCH 2/2] target/riscv: optimize helper for vmv<nr>r.v Weiwei Li
  2022-03-25  9:26   ` Frank Chang
@ 2022-03-28  1:11   ` Alistair Francis
  1 sibling, 0 replies; 7+ messages in thread
From: Alistair Francis @ 2022-03-28  1:11 UTC (permalink / raw)
  To: Weiwei Li
  Cc: Wei Wu (吴伟), open list:RISC-V, wangjunqiang,
	Bin Meng, Richard Henderson, qemu-devel@nongnu.org Developers,
	Palmer Dabbelt, Alistair Francis

On Fri, Mar 25, 2022 at 7:01 PM Weiwei Li <liweiwei@iscas.ac.cn> wrote:
>
> LEN is not used for GEN_VEXT_VMV_WHOLE macro, so vmv<nr>r.v can share
> the same helper
>
> Signed-off-by: Weiwei Li <liweiwei@iscas.ac.cn>
> Signed-off-by: Junqiang Wang <wangjunqiang@iscas.ac.cn>

Reviewed-by: Alistair Francis <alistair.francis@wdc.com>

Alistair

> ---
>  target/riscv/helper.h                   |  5 +----
>  target/riscv/insn_trans/trans_rvv.c.inc | 17 +++++----------
>  target/riscv/vector_helper.c            | 29 ++++++++++---------------
>  3 files changed, 18 insertions(+), 33 deletions(-)
>
> diff --git a/target/riscv/helper.h b/target/riscv/helper.h
> index 26bbab2fab..a669d0187b 100644
> --- a/target/riscv/helper.h
> +++ b/target/riscv/helper.h
> @@ -1086,10 +1086,7 @@ DEF_HELPER_6(vcompress_vm_h, void, ptr, ptr, ptr, ptr, env, i32)
>  DEF_HELPER_6(vcompress_vm_w, void, ptr, ptr, ptr, ptr, env, i32)
>  DEF_HELPER_6(vcompress_vm_d, void, ptr, ptr, ptr, ptr, env, i32)
>
> -DEF_HELPER_4(vmv1r_v, void, ptr, ptr, env, i32)
> -DEF_HELPER_4(vmv2r_v, void, ptr, ptr, env, i32)
> -DEF_HELPER_4(vmv4r_v, void, ptr, ptr, env, i32)
> -DEF_HELPER_4(vmv8r_v, void, ptr, ptr, env, i32)
> +DEF_HELPER_4(vmvr_v, void, ptr, ptr, env, i32)
>
>  DEF_HELPER_5(vzext_vf2_h, void, ptr, ptr, ptr, env, i32)
>  DEF_HELPER_5(vzext_vf2_w, void, ptr, ptr, ptr, env, i32)
> diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc
> index 2878ca3132..ec7c0e0d36 100644
> --- a/target/riscv/insn_trans/trans_rvv.c.inc
> +++ b/target/riscv/insn_trans/trans_rvv.c.inc
> @@ -3695,7 +3695,7 @@ static bool trans_vcompress_vm(DisasContext *s, arg_r *a)
>   * Whole Vector Register Move Instructions ignore vtype and vl setting.
>   * Thus, we don't need to check vill bit. (Section 16.6)
>   */
> -#define GEN_VMV_WHOLE_TRANS(NAME, LEN, SEQ)                             \
> +#define GEN_VMV_WHOLE_TRANS(NAME, LEN)                             \
>  static bool trans_##NAME(DisasContext *s, arg_##NAME * a)               \
>  {                                                                       \
>      if (require_rvv(s) &&                                               \
> @@ -3710,13 +3710,8 @@ static bool trans_##NAME(DisasContext *s, arg_##NAME * a)               \
>          } else {                                                        \
>              TCGLabel *over = gen_new_label();                           \
>              tcg_gen_brcondi_tl(TCG_COND_GEU, cpu_vstart, maxsz, over);  \
> -                                                                        \
> -            static gen_helper_gvec_2_ptr * const fns[4] = {             \
> -                gen_helper_vmv1r_v, gen_helper_vmv2r_v,                 \
> -                gen_helper_vmv4r_v, gen_helper_vmv8r_v,                 \
> -            };                                                          \
>              tcg_gen_gvec_2_ptr(vreg_ofs(s, a->rd), vreg_ofs(s, a->rs2), \
> -                               cpu_env, maxsz, maxsz, 0, fns[SEQ]);     \
> +                               cpu_env, maxsz, maxsz, 0, gen_helper_vmvr_v); \
>              mark_vs_dirty(s);                                           \
>              gen_set_label(over);                                        \
>          }                                                               \
> @@ -3725,10 +3720,10 @@ static bool trans_##NAME(DisasContext *s, arg_##NAME * a)               \
>      return false;                                                       \
>  }
>
> -GEN_VMV_WHOLE_TRANS(vmv1r_v, 1, 0)
> -GEN_VMV_WHOLE_TRANS(vmv2r_v, 2, 1)
> -GEN_VMV_WHOLE_TRANS(vmv4r_v, 4, 2)
> -GEN_VMV_WHOLE_TRANS(vmv8r_v, 8, 3)
> +GEN_VMV_WHOLE_TRANS(vmv1r_v, 1)
> +GEN_VMV_WHOLE_TRANS(vmv2r_v, 2)
> +GEN_VMV_WHOLE_TRANS(vmv4r_v, 4)
> +GEN_VMV_WHOLE_TRANS(vmv8r_v, 8)
>
>  static bool int_ext_check(DisasContext *s, arg_rmr *a, uint8_t div)
>  {
> diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
> index 3bd4aac9c9..1d4982ef7f 100644
> --- a/target/riscv/vector_helper.c
> +++ b/target/riscv/vector_helper.c
> @@ -4888,25 +4888,18 @@ GEN_VEXT_VCOMPRESS_VM(vcompress_vm_w, uint32_t, H4)
>  GEN_VEXT_VCOMPRESS_VM(vcompress_vm_d, uint64_t, H8)
>
>  /* Vector Whole Register Move */
> -#define GEN_VEXT_VMV_WHOLE(NAME, LEN)                      \
> -void HELPER(NAME)(void *vd, void *vs2, CPURISCVState *env, \
> -                  uint32_t desc)                           \
> -{                                                          \
> -    /* EEW = 8 */                                          \
> -    uint32_t maxsz = simd_maxsz(desc);                     \
> -    uint32_t i = env->vstart;                              \
> -                                                           \
> -    memcpy((uint8_t *)vd + H1(i),                          \
> -           (uint8_t *)vs2 + H1(i),                         \
> -           maxsz - env->vstart);                           \
> -                                                           \
> -    env->vstart = 0;                                       \
> -}
> +void HELPER(vmvr_v)(void *vd, void *vs2, CPURISCVState *env, uint32_t desc)
> +{
> +    /* EEW = 8 */
> +    uint32_t maxsz = simd_maxsz(desc);
> +    uint32_t i = env->vstart;
> +
> +    memcpy((uint8_t *)vd + H1(i),
> +           (uint8_t *)vs2 + H1(i),
> +           maxsz - env->vstart);
>
> -GEN_VEXT_VMV_WHOLE(vmv1r_v, 1)
> -GEN_VEXT_VMV_WHOLE(vmv2r_v, 2)
> -GEN_VEXT_VMV_WHOLE(vmv4r_v, 4)
> -GEN_VEXT_VMV_WHOLE(vmv8r_v, 8)
> +    env->vstart = 0;
> +}
>
>  /* Vector Integer Extension */
>  #define GEN_VEXT_INT_EXT(NAME, ETYPE, DTYPE, HD, HS1)            \
> --
> 2.17.1
>
>


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH 1/2] target/riscv: optimize condition assign for scale < 0
  2022-03-25  8:59 [PATCH 1/2] target/riscv: optimize condition assign for scale < 0 Weiwei Li
  2022-03-25  8:59 ` [PATCH 2/2] target/riscv: optimize helper for vmv<nr>r.v Weiwei Li
  2022-03-25  9:26 ` [PATCH 1/2] target/riscv: optimize condition assign for scale < 0 Frank Chang
@ 2022-03-28  1:11 ` Alistair Francis
  2022-03-28  4:01 ` Alistair Francis
  3 siblings, 0 replies; 7+ messages in thread
From: Alistair Francis @ 2022-03-28  1:11 UTC (permalink / raw)
  To: Weiwei Li
  Cc: Wei Wu (吴伟), open list:RISC-V, wangjunqiang,
	Bin Meng, Richard Henderson, qemu-devel@nongnu.org Developers,
	Palmer Dabbelt, Alistair Francis

On Fri, Mar 25, 2022 at 7:03 PM Weiwei Li <liweiwei@iscas.ac.cn> wrote:
>
> for some cases, scale is always equal or less than 0, since lmul is not larger than 3
>
> Signed-off-by: Weiwei Li <liweiwei@iscas.ac.cn>
> Signed-off-by: Junqiang Wang <wangjunqiang@iscas.ac.cn>

Acked-by: Alistair Francis <alistair.francis@wdc.com>

Alistair

> ---
>  target/riscv/insn_trans/trans_rvv.c.inc | 8 +++-----
>  1 file changed, 3 insertions(+), 5 deletions(-)
>
> diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc
> index 4ea7e41e1a..2878ca3132 100644
> --- a/target/riscv/insn_trans/trans_rvv.c.inc
> +++ b/target/riscv/insn_trans/trans_rvv.c.inc
> @@ -1198,7 +1198,7 @@ GEN_LDST_WHOLE_TRANS(vs8r_v, 8, true)
>  static inline uint32_t MAXSZ(DisasContext *s)
>  {
>      int scale = s->lmul - 3;
> -    return scale < 0 ? s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen << scale;
> +    return s->cfg_ptr->vlen >> -scale;
>  }
>
>  static bool opivv_check(DisasContext *s, arg_rmrr *a)
> @@ -3597,8 +3597,7 @@ static bool trans_vrgather_vx(DisasContext *s, arg_rmrr *a)
>
>      if (a->vm && s->vl_eq_vlmax) {
>          int scale = s->lmul - (s->sew + 3);
> -        int vlmax = scale < 0 ?
> -                       s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen << scale;
> +        int vlmax = s->cfg_ptr->vlen >> -scale;
>          TCGv_i64 dest = tcg_temp_new_i64();
>
>          if (a->rs1 == 0) {
> @@ -3630,8 +3629,7 @@ static bool trans_vrgather_vi(DisasContext *s, arg_rmrr *a)
>
>      if (a->vm && s->vl_eq_vlmax) {
>          int scale = s->lmul - (s->sew + 3);
> -        int vlmax = scale < 0 ?
> -                       s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen << scale;
> +        int vlmax = s->cfg_ptr->vlen >> -scale;
>          if (a->rs1 >= vlmax) {
>              tcg_gen_gvec_dup_imm(MO_64, vreg_ofs(s, a->rd),
>                                   MAXSZ(s), MAXSZ(s), 0);
> --
> 2.17.1
>
>


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH 1/2] target/riscv: optimize condition assign for scale < 0
  2022-03-25  8:59 [PATCH 1/2] target/riscv: optimize condition assign for scale < 0 Weiwei Li
                   ` (2 preceding siblings ...)
  2022-03-28  1:11 ` Alistair Francis
@ 2022-03-28  4:01 ` Alistair Francis
  3 siblings, 0 replies; 7+ messages in thread
From: Alistair Francis @ 2022-03-28  4:01 UTC (permalink / raw)
  To: Weiwei Li
  Cc: Wei Wu (吴伟), open list:RISC-V, wangjunqiang,
	Bin Meng, Richard Henderson, qemu-devel@nongnu.org Developers,
	Palmer Dabbelt, Alistair Francis

On Fri, Mar 25, 2022 at 7:03 PM Weiwei Li <liweiwei@iscas.ac.cn> wrote:
>
> for some cases, scale is always equal or less than 0, since lmul is not larger than 3
>
> Signed-off-by: Weiwei Li <liweiwei@iscas.ac.cn>
> Signed-off-by: Junqiang Wang <wangjunqiang@iscas.ac.cn>

Thanks!

Applied to riscv-to-apply.next

Alistair

> ---
>  target/riscv/insn_trans/trans_rvv.c.inc | 8 +++-----
>  1 file changed, 3 insertions(+), 5 deletions(-)
>
> diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc
> index 4ea7e41e1a..2878ca3132 100644
> --- a/target/riscv/insn_trans/trans_rvv.c.inc
> +++ b/target/riscv/insn_trans/trans_rvv.c.inc
> @@ -1198,7 +1198,7 @@ GEN_LDST_WHOLE_TRANS(vs8r_v, 8, true)
>  static inline uint32_t MAXSZ(DisasContext *s)
>  {
>      int scale = s->lmul - 3;
> -    return scale < 0 ? s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen << scale;
> +    return s->cfg_ptr->vlen >> -scale;
>  }
>
>  static bool opivv_check(DisasContext *s, arg_rmrr *a)
> @@ -3597,8 +3597,7 @@ static bool trans_vrgather_vx(DisasContext *s, arg_rmrr *a)
>
>      if (a->vm && s->vl_eq_vlmax) {
>          int scale = s->lmul - (s->sew + 3);
> -        int vlmax = scale < 0 ?
> -                       s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen << scale;
> +        int vlmax = s->cfg_ptr->vlen >> -scale;
>          TCGv_i64 dest = tcg_temp_new_i64();
>
>          if (a->rs1 == 0) {
> @@ -3630,8 +3629,7 @@ static bool trans_vrgather_vi(DisasContext *s, arg_rmrr *a)
>
>      if (a->vm && s->vl_eq_vlmax) {
>          int scale = s->lmul - (s->sew + 3);
> -        int vlmax = scale < 0 ?
> -                       s->cfg_ptr->vlen >> -scale : s->cfg_ptr->vlen << scale;
> +        int vlmax = s->cfg_ptr->vlen >> -scale;
>          if (a->rs1 >= vlmax) {
>              tcg_gen_gvec_dup_imm(MO_64, vreg_ofs(s, a->rd),
>                                   MAXSZ(s), MAXSZ(s), 0);
> --
> 2.17.1
>
>


^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2022-03-28  4:08 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2022-03-25  8:59 [PATCH 1/2] target/riscv: optimize condition assign for scale < 0 Weiwei Li
2022-03-25  8:59 ` [PATCH 2/2] target/riscv: optimize helper for vmv<nr>r.v Weiwei Li
2022-03-25  9:26   ` Frank Chang
2022-03-28  1:11   ` Alistair Francis
2022-03-25  9:26 ` [PATCH 1/2] target/riscv: optimize condition assign for scale < 0 Frank Chang
2022-03-28  1:11 ` Alistair Francis
2022-03-28  4:01 ` Alistair Francis

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).