Re: [PATCH] RISC-V: Fix vfirst/vmsbf/vmsif/vmsof ratio attributes

2024-01-21 Thread Kito Cheng
LGTM :)

On Mon, Jan 22, 2024 at 10:49 AM Juzhe-Zhong  wrote:
>
> vfirst/vmsbf/vmsif/vmsof instructions are supposed to demand ratio instead of 
> demanding sew_lmul.
> But my previous typo makes VSETVL PASS miss honor the risc-v v spec.
>
> Consider this following simple case:
>
> int foo4 (void * in, void * out)
> {
>   vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
>   v = __riscv_vadd_vv_i32m1 (v, v, 4);
>   vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32(v);
>   mask = __riscv_vmsof_m_b32(mask, 4);
>   return __riscv_vfirst_m_b32(mask, 4);
> }
>
> Before this patch:
>
> foo4:
> vsetivlizero,4,e32,m1,ta,ma
> vle32.v v1,0(a0)
> vadd.vv v1,v1,v1
> vsetvli zero,zero,e8,mf4,ta,ma> redundant.
> vmsof.m v2,v1
> vfirst.ma0,v2
> ret
>
> After this patch:
>
> foo4:
> vsetivlizero,4,e32,m1,ta,ma
> vle32.v v1,0(a0)
> vadd.vv v1,v1,v1
> vmsof.m v2,v1
> vfirst.ma0,v2
> ret
>
> Confirm RVV spec and Clang, this patch makes VSETVL PASS match the correct 
> behavior.
>
> Tested on both RV32/RV64, no regression.
>
> gcc/ChangeLog:
>
> * config/riscv/vector.md: Fix vfirst/vmsbf/vmsof ratio attributes.
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/riscv/rvv/vsetvl/attribute-1.c: New test.
>
> ---
>  gcc/config/riscv/vector.md|  2 +-
>  .../gcc.target/riscv/rvv/vsetvl/attribute-1.c | 47 +++
>  2 files changed, 48 insertions(+), 1 deletion(-)
>  create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/vsetvl/attribute-1.c
>
> diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md
> index cfc54ae5eac..307d9a8c952 100644
> --- a/gcc/config/riscv/vector.md
> +++ b/gcc/config/riscv/vector.md
> @@ -433,7 +433,7 @@
>   vialu,vshift,vicmp,vimul,vidiv,vsalu,\
>   vext,viwalu,viwmul,vicalu,vnshift,\
>   vimuladd,vimerge,vaalu,vsmul,vsshift,\
> - vnclip,viminmax,viwmuladd,vmffs,vmsfs,\
> + vnclip,viminmax,viwmuladd,\
>   vmiota,vmidx,vfalu,vfmul,vfminmax,vfdiv,\
>   vfwalu,vfwmul,vfsqrt,vfrecp,vfsgnj,vfcmp,\
>   vfmerge,vfcvtitof,vfcvtftoi,vfwcvtitof,\
> diff --git a/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/attribute-1.c 
> b/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/attribute-1.c
> new file mode 100644
> index 000..28dcf986bac
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/attribute-1.c
> @@ -0,0 +1,47 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gcv -mabi=lp64d -O3" } */
> +
> +#include "riscv_vector.h"
> +
> +int
> +foo (void *in, void *out)
> +{
> +  vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
> +  v = __riscv_vadd_vv_i32m1 (v, v, 4);
> +  vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32 (v);
> +  return __riscv_vfirst_m_b32 (mask, 4);
> +}
> +
> +int
> +foo2 (void *in, void *out)
> +{
> +  vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
> +  v = __riscv_vadd_vv_i32m1 (v, v, 4);
> +  vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32 (v);
> +  mask = __riscv_vmsbf_m_b32 (mask, 4);
> +  return __riscv_vfirst_m_b32 (mask, 4);
> +}
> +
> +int
> +foo3 (void *in, void *out)
> +{
> +  vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
> +  v = __riscv_vadd_vv_i32m1 (v, v, 4);
> +  vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32 (v);
> +  mask = __riscv_vmsif_m_b32 (mask, 4);
> +  return __riscv_vfirst_m_b32 (mask, 4);
> +}
> +
> +int
> +foo4 (void *in, void *out)
> +{
> +  vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
> +  v = __riscv_vadd_vv_i32m1 (v, v, 4);
> +  vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32 (v);
> +  mask = __riscv_vmsof_m_b32 (mask, 4);
> +  return __riscv_vfirst_m_b32 (mask, 4);
> +}
> +
> +/* { dg-final { scan-assembler-times 
> {vsetivli\s+zero,\s*4,\s*e32,\s*m1,\s*t[au],\s*m[au]} 4 } } */
> +/* { dg-final { scan-assembler-times {vsetivli} 4 } } */
> +/* { dg-final { scan-assembler-not {vsetvli} } } */
> --
> 2.36.3
>


[PATCH] RISC-V: Fix vfirst/vmsbf/vmsif/vmsof ratio attributes

2024-01-21 Thread Juzhe-Zhong
vfirst/vmsbf/vmsif/vmsof instructions are supposed to demand ratio instead of 
demanding sew_lmul.
But my previous typo makes VSETVL PASS miss honor the risc-v v spec.

Consider this following simple case:

int foo4 (void * in, void * out)
{
  vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
  v = __riscv_vadd_vv_i32m1 (v, v, 4);
  vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32(v);
  mask = __riscv_vmsof_m_b32(mask, 4);
  return __riscv_vfirst_m_b32(mask, 4);
}

Before this patch:

foo4:
vsetivlizero,4,e32,m1,ta,ma
vle32.v v1,0(a0)
vadd.vv v1,v1,v1
vsetvli zero,zero,e8,mf4,ta,ma> redundant.
vmsof.m v2,v1
vfirst.ma0,v2
ret

After this patch:

foo4:
vsetivlizero,4,e32,m1,ta,ma
vle32.v v1,0(a0)
vadd.vv v1,v1,v1
vmsof.m v2,v1
vfirst.ma0,v2
ret

Confirm RVV spec and Clang, this patch makes VSETVL PASS match the correct 
behavior.

Tested on both RV32/RV64, no regression.

gcc/ChangeLog:

* config/riscv/vector.md: Fix vfirst/vmsbf/vmsof ratio attributes.

gcc/testsuite/ChangeLog:

* gcc.target/riscv/rvv/vsetvl/attribute-1.c: New test.

---
 gcc/config/riscv/vector.md|  2 +-
 .../gcc.target/riscv/rvv/vsetvl/attribute-1.c | 47 +++
 2 files changed, 48 insertions(+), 1 deletion(-)
 create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/vsetvl/attribute-1.c

diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md
index cfc54ae5eac..307d9a8c952 100644
--- a/gcc/config/riscv/vector.md
+++ b/gcc/config/riscv/vector.md
@@ -433,7 +433,7 @@
  vialu,vshift,vicmp,vimul,vidiv,vsalu,\
  vext,viwalu,viwmul,vicalu,vnshift,\
  vimuladd,vimerge,vaalu,vsmul,vsshift,\
- vnclip,viminmax,viwmuladd,vmffs,vmsfs,\
+ vnclip,viminmax,viwmuladd,\
  vmiota,vmidx,vfalu,vfmul,vfminmax,vfdiv,\
  vfwalu,vfwmul,vfsqrt,vfrecp,vfsgnj,vfcmp,\
  vfmerge,vfcvtitof,vfcvtftoi,vfwcvtitof,\
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/attribute-1.c 
b/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/attribute-1.c
new file mode 100644
index 000..28dcf986bac
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/attribute-1.c
@@ -0,0 +1,47 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv -mabi=lp64d -O3" } */
+
+#include "riscv_vector.h"
+
+int
+foo (void *in, void *out)
+{
+  vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
+  v = __riscv_vadd_vv_i32m1 (v, v, 4);
+  vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32 (v);
+  return __riscv_vfirst_m_b32 (mask, 4);
+}
+
+int
+foo2 (void *in, void *out)
+{
+  vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
+  v = __riscv_vadd_vv_i32m1 (v, v, 4);
+  vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32 (v);
+  mask = __riscv_vmsbf_m_b32 (mask, 4);
+  return __riscv_vfirst_m_b32 (mask, 4);
+}
+
+int
+foo3 (void *in, void *out)
+{
+  vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
+  v = __riscv_vadd_vv_i32m1 (v, v, 4);
+  vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32 (v);
+  mask = __riscv_vmsif_m_b32 (mask, 4);
+  return __riscv_vfirst_m_b32 (mask, 4);
+}
+
+int
+foo4 (void *in, void *out)
+{
+  vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
+  v = __riscv_vadd_vv_i32m1 (v, v, 4);
+  vbool32_t mask = __riscv_vreinterpret_v_i32m1_b32 (v);
+  mask = __riscv_vmsof_m_b32 (mask, 4);
+  return __riscv_vfirst_m_b32 (mask, 4);
+}
+
+/* { dg-final { scan-assembler-times 
{vsetivli\s+zero,\s*4,\s*e32,\s*m1,\s*t[au],\s*m[au]} 4 } } */
+/* { dg-final { scan-assembler-times {vsetivli} 4 } } */
+/* { dg-final { scan-assembler-not {vsetvli} } } */
-- 
2.36.3