RISC-V: Support VLS modes reduction[PR111153]

This patch supports VLS reduction vectorization.

It can optimize the current reduction vectorization codegen with current COST model.

TYPE __attribute__ ((noinline, noclone))	\
reduc_plus_##TYPE (TYPE * __restrict a, int n)		\
{						\
  TYPE r = 0;					\
  for (int i = 0; i < n; ++i)			\
    r += a[i];					\
  return r;					\
}

  T (int32_t)					\

TEST_PLUS (DEF_REDUC_PLUS)

Before this patch:

        vle32.v v2,0(a5)
        addi    a5,a5,16
        vadd.vv v1,v1,v2
        bne     a5,a4,.L4
        lui     a4,%hi(.LC0)
        lui     a5,%hi(.LC1)
        addi    a4,a4,%lo(.LC0)
        vlm.v   v0,0(a4)
        addi    a5,a5,%lo(.LC1)
        andi    a1,a1,-4
        vmv1r.v v2,v3
        vlm.v   v4,0(a5)
        vcompress.vm    v2,v1,v0
        vmv1r.v v0,v4
        vadd.vv v1,v2,v1
        vcompress.vm    v3,v1,v0
        vadd.vv v3,v3,v1
        vmv.x.s a0,v3
        sext.w  a0,a0
        beq     a3,a1,.L12

After this patch:

	vle32.v	v2,0(a5)
	addi	a5,a5,16
	vadd.vv	v1,v1,v2
	bne	a5,a4,.L4
	li	a5,0
	andi	a1,a1,-4
	vmv.s.x	v2,a5
	vredsum.vs	v1,v1,v2
	vmv.x.s	a0,v1
	beq	a3,a1,.L12

	PR target/111153

gcc/ChangeLog:

	* config/riscv/autovec.md: Add VLS modes.

gcc/testsuite/ChangeLog:

	* gcc.target/riscv/rvv/autovec/vls/def.h: Add VLS mode reduction case.
	* gcc.target/riscv/rvv/autovec/vls/reduc-1.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-10.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-11.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-12.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-13.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-14.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-15.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-16.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-17.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-18.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-19.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-2.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-20.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-21.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-3.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-4.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-5.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-6.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-7.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-8.c: New test.
	* gcc.target/riscv/rvv/autovec/vls/reduc-9.c: New test.

23 files changed, 825 insertions, 1 deletions

diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md
index fe2405f..63a8397 100644
--- a/gcc/config/riscv/autovec.md
+++ b/gcc/config/riscv/autovec.md
@@ -2241,7 +2241,7 @@
 (define_insn_and_split "fold_left_plus_<mode>"
   [(set (match_operand:<VEL> 0 "register_operand")
         (unspec:<VEL> [
-             (match_operand:VF 2 "register_operand")
+             (match_operand:V_VLSF 2 "register_operand")
              (match_operand:<VEL> 1 "register_operand")
         ] UNSPEC_REDUC_SUM_ORDERED))]
   "TARGET_VECTOR && can_create_pseudo_p ()"
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/def.h b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/def.h
index 81c4570..9ede7af 100644
--- a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/def.h
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/def.h
@@ -266,3 +266,33 @@ typedef double v512df __attribute__ ((vector_size (4096)));
     for (int i = 0; i < NUM; ++i)                                              \
       a[i] = b[i] * CALL (1.0, c[i]);                                          \
   }
+
+#define DEF_REDUC_PLUS(TYPE, NUM)                                              \
+  TYPE __attribute__ ((noinline, noclone))                                     \
+  reduc_plus_##TYPE##NUM (TYPE *__restrict a)                                  \
+  {                                                                            \
+    TYPE r = 0;                                                                \
+    for (int i = 0; i < NUM; ++i)                                              \
+      r += a[i];                                                               \
+    return r;                                                                  \
+  }
+
+#define DEF_REDUC_MAXMIN(TYPE, NAME, CMP_OP, NUM)                              \
+  TYPE __attribute__ ((noinline, noclone))                                     \
+  reduc_##NAME##_##TYPE##_##NUM (TYPE *a)                                      \
+  {                                                                            \
+    TYPE r = 13;                                                               \
+    for (int i = 0; i < NUM; ++i)                                              \
+      r = a[i] CMP_OP r ? a[i] : r;                                            \
+    return r;                                                                  \
+  }
+
+#define DEF_REDUC_BITWISE(TYPE, NAME, BIT_OP, NUM)                             \
+  TYPE __attribute__ ((noinline, noclone))                                     \
+  reduc_##NAME##_##TYPE##_##NUM (TYPE *a)                                      \
+  {                                                                            \
+    TYPE r = 13;                                                               \
+    for (int i = 0; i < NUM; ++i)                                              \
+      r BIT_OP a[i];                                                           \
+    return r;                                                                  \
+  }
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-1.c
new file mode 100644
index 0000000..2db25a2
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-1.c
@@ -0,0 +1,31 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (int8_t, 4)
+DEF_REDUC_PLUS (int8_t, 8)
+DEF_REDUC_PLUS (int8_t, 16)
+DEF_REDUC_PLUS (int8_t, 32)
+DEF_REDUC_PLUS (int8_t, 64)
+DEF_REDUC_PLUS (int8_t, 128)
+DEF_REDUC_PLUS (int8_t, 256)
+DEF_REDUC_PLUS (int8_t, 512)
+DEF_REDUC_PLUS (int8_t, 1024)
+DEF_REDUC_PLUS (int8_t, 2048)
+DEF_REDUC_PLUS (int8_t, 4096)
+
+DEF_REDUC_PLUS (uint8_t, 4)
+DEF_REDUC_PLUS (uint8_t, 8)
+DEF_REDUC_PLUS (uint8_t, 16)
+DEF_REDUC_PLUS (uint8_t, 32)
+DEF_REDUC_PLUS (uint8_t, 64)
+DEF_REDUC_PLUS (uint8_t, 128)
+DEF_REDUC_PLUS (uint8_t, 256)
+DEF_REDUC_PLUS (uint8_t, 512)
+DEF_REDUC_PLUS (uint8_t, 1024)
+DEF_REDUC_PLUS (uint8_t, 2048)
+DEF_REDUC_PLUS (uint8_t, 4096)
+
+/* { dg-final { scan-assembler-times {vredsum\.vs} 22 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-10.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-10.c
new file mode 100644
index 0000000..cdbbe11
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-10.c
@@ -0,0 +1,50 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_MAXMIN (int32_t, max, >, 4)
+DEF_REDUC_MAXMIN (int32_t, max, >, 8)
+DEF_REDUC_MAXMIN (int32_t, max, >, 16)
+DEF_REDUC_MAXMIN (int32_t, max, >, 32)
+DEF_REDUC_MAXMIN (int32_t, max, >, 64)
+DEF_REDUC_MAXMIN (int32_t, max, >, 128)
+DEF_REDUC_MAXMIN (int32_t, max, >, 256)
+DEF_REDUC_MAXMIN (int32_t, max, >, 512)
+DEF_REDUC_MAXMIN (int32_t, max, >, 1024)
+
+DEF_REDUC_MAXMIN (uint32_t, max, >, 4)
+DEF_REDUC_MAXMIN (uint32_t, max, >, 8)
+DEF_REDUC_MAXMIN (uint32_t, max, >, 16)
+DEF_REDUC_MAXMIN (uint32_t, max, >, 32)
+DEF_REDUC_MAXMIN (uint32_t, max, >, 64)
+DEF_REDUC_MAXMIN (uint32_t, max, >, 128)
+DEF_REDUC_MAXMIN (uint32_t, max, >, 256)
+DEF_REDUC_MAXMIN (uint32_t, max, >, 512)
+DEF_REDUC_MAXMIN (uint32_t, max, >, 1024)
+
+DEF_REDUC_MAXMIN (int32_t, min, <, 4)
+DEF_REDUC_MAXMIN (int32_t, min, <, 8)
+DEF_REDUC_MAXMIN (int32_t, min, <, 16)
+DEF_REDUC_MAXMIN (int32_t, min, <, 32)
+DEF_REDUC_MAXMIN (int32_t, min, <, 64)
+DEF_REDUC_MAXMIN (int32_t, min, <, 128)
+DEF_REDUC_MAXMIN (int32_t, min, <, 256)
+DEF_REDUC_MAXMIN (int32_t, min, <, 512)
+DEF_REDUC_MAXMIN (int32_t, min, <, 1024)
+
+DEF_REDUC_MAXMIN (uint32_t, min, <, 4)
+DEF_REDUC_MAXMIN (uint32_t, min, <, 8)
+DEF_REDUC_MAXMIN (uint32_t, min, <, 16)
+DEF_REDUC_MAXMIN (uint32_t, min, <, 32)
+DEF_REDUC_MAXMIN (uint32_t, min, <, 64)
+DEF_REDUC_MAXMIN (uint32_t, min, <, 128)
+DEF_REDUC_MAXMIN (uint32_t, min, <, 256)
+DEF_REDUC_MAXMIN (uint32_t, min, <, 512)
+DEF_REDUC_MAXMIN (uint32_t, min, <, 1024)
+
+/* { dg-final { scan-assembler-times {vredmax\.vs} 9 } } */
+/* { dg-final { scan-assembler-times {vredmaxu\.vs} 9 } } */
+/* { dg-final { scan-assembler-times {vredmin\.vs} 9 } } */
+/* { dg-final { scan-assembler-times {vredminu\.vs} 9 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-11.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-11.c
new file mode 100644
index 0000000..d204031
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-11.c
@@ -0,0 +1,46 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_MAXMIN (int64_t, max, >, 4)
+DEF_REDUC_MAXMIN (int64_t, max, >, 8)
+DEF_REDUC_MAXMIN (int64_t, max, >, 16)
+DEF_REDUC_MAXMIN (int64_t, max, >, 32)
+DEF_REDUC_MAXMIN (int64_t, max, >, 64)
+DEF_REDUC_MAXMIN (int64_t, max, >, 128)
+DEF_REDUC_MAXMIN (int64_t, max, >, 256)
+DEF_REDUC_MAXMIN (int64_t, max, >, 512)
+
+DEF_REDUC_MAXMIN (uint64_t, max, >, 4)
+DEF_REDUC_MAXMIN (uint64_t, max, >, 8)
+DEF_REDUC_MAXMIN (uint64_t, max, >, 16)
+DEF_REDUC_MAXMIN (uint64_t, max, >, 32)
+DEF_REDUC_MAXMIN (uint64_t, max, >, 64)
+DEF_REDUC_MAXMIN (uint64_t, max, >, 128)
+DEF_REDUC_MAXMIN (uint64_t, max, >, 256)
+DEF_REDUC_MAXMIN (uint64_t, max, >, 512)
+
+DEF_REDUC_MAXMIN (int64_t, min, <, 4)
+DEF_REDUC_MAXMIN (int64_t, min, <, 8)
+DEF_REDUC_MAXMIN (int64_t, min, <, 16)
+DEF_REDUC_MAXMIN (int64_t, min, <, 32)
+DEF_REDUC_MAXMIN (int64_t, min, <, 64)
+DEF_REDUC_MAXMIN (int64_t, min, <, 128)
+DEF_REDUC_MAXMIN (int64_t, min, <, 256)
+DEF_REDUC_MAXMIN (int64_t, min, <, 512)
+
+DEF_REDUC_MAXMIN (uint64_t, min, <, 4)
+DEF_REDUC_MAXMIN (uint64_t, min, <, 8)
+DEF_REDUC_MAXMIN (uint64_t, min, <, 16)
+DEF_REDUC_MAXMIN (uint64_t, min, <, 32)
+DEF_REDUC_MAXMIN (uint64_t, min, <, 64)
+DEF_REDUC_MAXMIN (uint64_t, min, <, 128)
+DEF_REDUC_MAXMIN (uint64_t, min, <, 256)
+DEF_REDUC_MAXMIN (uint64_t, min, <, 512)
+
+/* { dg-final { scan-assembler-times {vredmax\.vs} 8 } } */
+/* { dg-final { scan-assembler-times {vredmaxu\.vs} 8 } } */
+/* { dg-final { scan-assembler-times {vredmin\.vs} 8 } } */
+/* { dg-final { scan-assembler-times {vredminu\.vs} 8 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-12.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-12.c
new file mode 100644
index 0000000..97660d2
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-12.c
@@ -0,0 +1,30 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8 -ffast-math" } */
+
+#include "def.h"
+
+DEF_REDUC_MAXMIN (_Float16, max, >, 4)
+DEF_REDUC_MAXMIN (_Float16, max, >, 8)
+DEF_REDUC_MAXMIN (_Float16, max, >, 16)
+DEF_REDUC_MAXMIN (_Float16, max, >, 32)
+DEF_REDUC_MAXMIN (_Float16, max, >, 64)
+DEF_REDUC_MAXMIN (_Float16, max, >, 128)
+DEF_REDUC_MAXMIN (_Float16, max, >, 256)
+DEF_REDUC_MAXMIN (_Float16, max, >, 512)
+DEF_REDUC_MAXMIN (_Float16, max, >, 1024)
+DEF_REDUC_MAXMIN (_Float16, max, >, 2048)
+
+DEF_REDUC_MAXMIN (_Float16, min, <, 4)
+DEF_REDUC_MAXMIN (_Float16, min, <, 8)
+DEF_REDUC_MAXMIN (_Float16, min, <, 16)
+DEF_REDUC_MAXMIN (_Float16, min, <, 32)
+DEF_REDUC_MAXMIN (_Float16, min, <, 64)
+DEF_REDUC_MAXMIN (_Float16, min, <, 128)
+DEF_REDUC_MAXMIN (_Float16, min, <, 256)
+DEF_REDUC_MAXMIN (_Float16, min, <, 512)
+DEF_REDUC_MAXMIN (_Float16, min, <, 1024)
+DEF_REDUC_MAXMIN (_Float16, min, <, 2048)
+
+/* { dg-final { scan-assembler-times {vfredmax\.vs} 10 } } */
+/* { dg-final { scan-assembler-times {vfredmin\.vs} 10 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-13.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-13.c
new file mode 100644
index 0000000..e4bc95c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-13.c
@@ -0,0 +1,28 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8 -ffast-math" } */
+
+#include "def.h"
+
+DEF_REDUC_MAXMIN (float, max, >, 4)
+DEF_REDUC_MAXMIN (float, max, >, 8)
+DEF_REDUC_MAXMIN (float, max, >, 16)
+DEF_REDUC_MAXMIN (float, max, >, 32)
+DEF_REDUC_MAXMIN (float, max, >, 64)
+DEF_REDUC_MAXMIN (float, max, >, 128)
+DEF_REDUC_MAXMIN (float, max, >, 256)
+DEF_REDUC_MAXMIN (float, max, >, 512)
+DEF_REDUC_MAXMIN (float, max, >, 1024)
+
+DEF_REDUC_MAXMIN (float, min, <, 4)
+DEF_REDUC_MAXMIN (float, min, <, 8)
+DEF_REDUC_MAXMIN (float, min, <, 16)
+DEF_REDUC_MAXMIN (float, min, <, 32)
+DEF_REDUC_MAXMIN (float, min, <, 64)
+DEF_REDUC_MAXMIN (float, min, <, 128)
+DEF_REDUC_MAXMIN (float, min, <, 256)
+DEF_REDUC_MAXMIN (float, min, <, 512)
+DEF_REDUC_MAXMIN (float, min, <, 1024)
+
+/* { dg-final { scan-assembler-times {vfredmax\.vs} 9 } } */
+/* { dg-final { scan-assembler-times {vfredmin\.vs} 9 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-14.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-14.c
new file mode 100644
index 0000000..c90e926
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-14.c
@@ -0,0 +1,26 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8 -ffast-math" } */
+
+#include "def.h"
+
+DEF_REDUC_MAXMIN (double, max, >, 4)
+DEF_REDUC_MAXMIN (double, max, >, 8)
+DEF_REDUC_MAXMIN (double, max, >, 16)
+DEF_REDUC_MAXMIN (double, max, >, 32)
+DEF_REDUC_MAXMIN (double, max, >, 64)
+DEF_REDUC_MAXMIN (double, max, >, 128)
+DEF_REDUC_MAXMIN (double, max, >, 256)
+DEF_REDUC_MAXMIN (double, max, >, 512)
+
+DEF_REDUC_MAXMIN (double, min, <, 4)
+DEF_REDUC_MAXMIN (double, min, <, 8)
+DEF_REDUC_MAXMIN (double, min, <, 16)
+DEF_REDUC_MAXMIN (double, min, <, 32)
+DEF_REDUC_MAXMIN (double, min, <, 64)
+DEF_REDUC_MAXMIN (double, min, <, 128)
+DEF_REDUC_MAXMIN (double, min, <, 256)
+DEF_REDUC_MAXMIN (double, min, <, 512)
+
+/* { dg-final { scan-assembler-times {vfredmax\.vs} 8 } } */
+/* { dg-final { scan-assembler-times {vfredmin\.vs} 8 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-15.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-15.c
new file mode 100644
index 0000000..65863f7
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-15.c
@@ -0,0 +1,81 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_BITWISE (int8_t, and, &=, 4)
+DEF_REDUC_BITWISE (int8_t, and, &=, 8)
+DEF_REDUC_BITWISE (int8_t, and, &=, 16)
+DEF_REDUC_BITWISE (int8_t, and, &=, 32)
+DEF_REDUC_BITWISE (int8_t, and, &=, 64)
+DEF_REDUC_BITWISE (int8_t, and, &=, 128)
+DEF_REDUC_BITWISE (int8_t, and, &=, 256)
+DEF_REDUC_BITWISE (int8_t, and, &=, 512)
+DEF_REDUC_BITWISE (int8_t, and, &=, 1024)
+DEF_REDUC_BITWISE (int8_t, and, &=, 2048)
+DEF_REDUC_BITWISE (int8_t, and, &=, 4096)
+
+DEF_REDUC_BITWISE (uint8_t, and, &=, 4)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 8)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 16)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 32)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 64)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 128)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 256)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 512)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 1024)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 2048)
+DEF_REDUC_BITWISE (uint8_t, and, &=, 4096)
+
+DEF_REDUC_BITWISE (int8_t, ior, |=, 4)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 8)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 16)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 32)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 64)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 128)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 256)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 512)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 1024)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 2048)
+DEF_REDUC_BITWISE (int8_t, ior, |=, 4096)
+
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 4)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 8)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 16)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 32)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 64)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 128)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 256)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 512)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 1024)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 2048)
+DEF_REDUC_BITWISE (uint8_t, ior, |=, 4096)
+
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 4)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 8)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 16)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 32)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 64)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 128)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 256)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 512)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 1024)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 2048)
+DEF_REDUC_BITWISE (int8_t, xor, ^=, 4096)
+
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 4)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 8)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 16)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 32)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 64)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 128)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 256)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 512)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 1024)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 2048)
+DEF_REDUC_BITWISE (uint8_t, xor, ^=, 4096)
+
+/* { dg-final { scan-assembler-times {vredand\.vs} 22 } } */
+/* { dg-final { scan-assembler-times {vredor\.vs} 22 } } */
+/* { dg-final { scan-assembler-times {vredxor\.vs} 22 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-16.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-16.c
new file mode 100644
index 0000000..c6c5714
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-16.c
@@ -0,0 +1,75 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_BITWISE (int16_t, and, &=, 4)
+DEF_REDUC_BITWISE (int16_t, and, &=, 8)
+DEF_REDUC_BITWISE (int16_t, and, &=, 16)
+DEF_REDUC_BITWISE (int16_t, and, &=, 32)
+DEF_REDUC_BITWISE (int16_t, and, &=, 64)
+DEF_REDUC_BITWISE (int16_t, and, &=, 128)
+DEF_REDUC_BITWISE (int16_t, and, &=, 256)
+DEF_REDUC_BITWISE (int16_t, and, &=, 512)
+DEF_REDUC_BITWISE (int16_t, and, &=, 1024)
+DEF_REDUC_BITWISE (int16_t, and, &=, 2048)
+
+DEF_REDUC_BITWISE (uint16_t, and, &=, 4)
+DEF_REDUC_BITWISE (uint16_t, and, &=, 8)
+DEF_REDUC_BITWISE (uint16_t, and, &=, 16)
+DEF_REDUC_BITWISE (uint16_t, and, &=, 32)
+DEF_REDUC_BITWISE (uint16_t, and, &=, 64)
+DEF_REDUC_BITWISE (uint16_t, and, &=, 128)
+DEF_REDUC_BITWISE (uint16_t, and, &=, 256)
+DEF_REDUC_BITWISE (uint16_t, and, &=, 512)
+DEF_REDUC_BITWISE (uint16_t, and, &=, 1024)
+DEF_REDUC_BITWISE (uint16_t, and, &=, 2048)
+
+DEF_REDUC_BITWISE (int16_t, ior, |=, 4)
+DEF_REDUC_BITWISE (int16_t, ior, |=, 8)
+DEF_REDUC_BITWISE (int16_t, ior, |=, 16)
+DEF_REDUC_BITWISE (int16_t, ior, |=, 32)
+DEF_REDUC_BITWISE (int16_t, ior, |=, 64)
+DEF_REDUC_BITWISE (int16_t, ior, |=, 128)
+DEF_REDUC_BITWISE (int16_t, ior, |=, 256)
+DEF_REDUC_BITWISE (int16_t, ior, |=, 512)
+DEF_REDUC_BITWISE (int16_t, ior, |=, 1024)
+DEF_REDUC_BITWISE (int16_t, ior, |=, 2048)
+
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 4)
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 8)
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 16)
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 32)
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 64)
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 128)
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 256)
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 512)
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 1024)
+DEF_REDUC_BITWISE (uint16_t, ior, |=, 2048)
+
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 4)
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 8)
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 16)
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 32)
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 64)
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 128)
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 256)
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 512)
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 1024)
+DEF_REDUC_BITWISE (int16_t, xor, ^=, 2048)
+
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 4)
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 8)
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 16)
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 32)
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 64)
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 128)
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 256)
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 512)
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 1024)
+DEF_REDUC_BITWISE (uint16_t, xor, ^=, 2048)
+
+/* { dg-final { scan-assembler-times {vredand\.vs} 20 } } */
+/* { dg-final { scan-assembler-times {vredor\.vs} 20 } } */
+/* { dg-final { scan-assembler-times {vredxor\.vs} 20 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-17.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-17.c
new file mode 100644
index 0000000..7598307
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-17.c
@@ -0,0 +1,69 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_BITWISE (int32_t, and, &=, 4)
+DEF_REDUC_BITWISE (int32_t, and, &=, 8)
+DEF_REDUC_BITWISE (int32_t, and, &=, 16)
+DEF_REDUC_BITWISE (int32_t, and, &=, 32)
+DEF_REDUC_BITWISE (int32_t, and, &=, 64)
+DEF_REDUC_BITWISE (int32_t, and, &=, 128)
+DEF_REDUC_BITWISE (int32_t, and, &=, 256)
+DEF_REDUC_BITWISE (int32_t, and, &=, 512)
+DEF_REDUC_BITWISE (int32_t, and, &=, 1024)
+
+DEF_REDUC_BITWISE (uint32_t, and, &=, 4)
+DEF_REDUC_BITWISE (uint32_t, and, &=, 8)
+DEF_REDUC_BITWISE (uint32_t, and, &=, 16)
+DEF_REDUC_BITWISE (uint32_t, and, &=, 32)
+DEF_REDUC_BITWISE (uint32_t, and, &=, 64)
+DEF_REDUC_BITWISE (uint32_t, and, &=, 128)
+DEF_REDUC_BITWISE (uint32_t, and, &=, 256)
+DEF_REDUC_BITWISE (uint32_t, and, &=, 512)
+DEF_REDUC_BITWISE (uint32_t, and, &=, 1024)
+
+DEF_REDUC_BITWISE (int32_t, ior, |=, 4)
+DEF_REDUC_BITWISE (int32_t, ior, |=, 8)
+DEF_REDUC_BITWISE (int32_t, ior, |=, 16)
+DEF_REDUC_BITWISE (int32_t, ior, |=, 32)
+DEF_REDUC_BITWISE (int32_t, ior, |=, 64)
+DEF_REDUC_BITWISE (int32_t, ior, |=, 128)
+DEF_REDUC_BITWISE (int32_t, ior, |=, 256)
+DEF_REDUC_BITWISE (int32_t, ior, |=, 512)
+DEF_REDUC_BITWISE (int32_t, ior, |=, 1024)
+
+DEF_REDUC_BITWISE (uint32_t, ior, |=, 4)
+DEF_REDUC_BITWISE (uint32_t, ior, |=, 8)
+DEF_REDUC_BITWISE (uint32_t, ior, |=, 16)
+DEF_REDUC_BITWISE (uint32_t, ior, |=, 32)
+DEF_REDUC_BITWISE (uint32_t, ior, |=, 64)
+DEF_REDUC_BITWISE (uint32_t, ior, |=, 128)
+DEF_REDUC_BITWISE (uint32_t, ior, |=, 256)
+DEF_REDUC_BITWISE (uint32_t, ior, |=, 512)
+DEF_REDUC_BITWISE (uint32_t, ior, |=, 1024)
+
+DEF_REDUC_BITWISE (int32_t, xor, ^=, 4)
+DEF_REDUC_BITWISE (int32_t, xor, ^=, 8)
+DEF_REDUC_BITWISE (int32_t, xor, ^=, 16)
+DEF_REDUC_BITWISE (int32_t, xor, ^=, 32)
+DEF_REDUC_BITWISE (int32_t, xor, ^=, 64)
+DEF_REDUC_BITWISE (int32_t, xor, ^=, 128)
+DEF_REDUC_BITWISE (int32_t, xor, ^=, 256)
+DEF_REDUC_BITWISE (int32_t, xor, ^=, 512)
+DEF_REDUC_BITWISE (int32_t, xor, ^=, 1024)
+
+DEF_REDUC_BITWISE (uint32_t, xor, ^=, 4)
+DEF_REDUC_BITWISE (uint32_t, xor, ^=, 8)
+DEF_REDUC_BITWISE (uint32_t, xor, ^=, 16)
+DEF_REDUC_BITWISE (uint32_t, xor, ^=, 32)
+DEF_REDUC_BITWISE (uint32_t, xor, ^=, 64)
+DEF_REDUC_BITWISE (uint32_t, xor, ^=, 128)
+DEF_REDUC_BITWISE (uint32_t, xor, ^=, 256)
+DEF_REDUC_BITWISE (uint32_t, xor, ^=, 512)
+DEF_REDUC_BITWISE (uint32_t, xor, ^=, 1024)
+
+/* { dg-final { scan-assembler-times {vredand\.vs} 18 } } */
+/* { dg-final { scan-assembler-times {vredor\.vs} 18 } } */
+/* { dg-final { scan-assembler-times {vredxor\.vs} 18 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-18.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-18.c
new file mode 100644
index 0000000..30dfad6
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-18.c
@@ -0,0 +1,63 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_BITWISE (int64_t, and, &=, 4)
+DEF_REDUC_BITWISE (int64_t, and, &=, 8)
+DEF_REDUC_BITWISE (int64_t, and, &=, 16)
+DEF_REDUC_BITWISE (int64_t, and, &=, 32)
+DEF_REDUC_BITWISE (int64_t, and, &=, 64)
+DEF_REDUC_BITWISE (int64_t, and, &=, 128)
+DEF_REDUC_BITWISE (int64_t, and, &=, 256)
+DEF_REDUC_BITWISE (int64_t, and, &=, 512)
+
+DEF_REDUC_BITWISE (uint64_t, and, &=, 4)
+DEF_REDUC_BITWISE (uint64_t, and, &=, 8)
+DEF_REDUC_BITWISE (uint64_t, and, &=, 16)
+DEF_REDUC_BITWISE (uint64_t, and, &=, 32)
+DEF_REDUC_BITWISE (uint64_t, and, &=, 64)
+DEF_REDUC_BITWISE (uint64_t, and, &=, 128)
+DEF_REDUC_BITWISE (uint64_t, and, &=, 256)
+DEF_REDUC_BITWISE (uint64_t, and, &=, 512)
+
+DEF_REDUC_BITWISE (int64_t, ior, |=, 4)
+DEF_REDUC_BITWISE (int64_t, ior, |=, 8)
+DEF_REDUC_BITWISE (int64_t, ior, |=, 16)
+DEF_REDUC_BITWISE (int64_t, ior, |=, 32)
+DEF_REDUC_BITWISE (int64_t, ior, |=, 64)
+DEF_REDUC_BITWISE (int64_t, ior, |=, 128)
+DEF_REDUC_BITWISE (int64_t, ior, |=, 256)
+DEF_REDUC_BITWISE (int64_t, ior, |=, 512)
+
+DEF_REDUC_BITWISE (uint64_t, ior, |=, 4)
+DEF_REDUC_BITWISE (uint64_t, ior, |=, 8)
+DEF_REDUC_BITWISE (uint64_t, ior, |=, 16)
+DEF_REDUC_BITWISE (uint64_t, ior, |=, 32)
+DEF_REDUC_BITWISE (uint64_t, ior, |=, 64)
+DEF_REDUC_BITWISE (uint64_t, ior, |=, 128)
+DEF_REDUC_BITWISE (uint64_t, ior, |=, 256)
+DEF_REDUC_BITWISE (uint64_t, ior, |=, 512)
+
+DEF_REDUC_BITWISE (int64_t, xor, ^=, 4)
+DEF_REDUC_BITWISE (int64_t, xor, ^=, 8)
+DEF_REDUC_BITWISE (int64_t, xor, ^=, 16)
+DEF_REDUC_BITWISE (int64_t, xor, ^=, 32)
+DEF_REDUC_BITWISE (int64_t, xor, ^=, 64)
+DEF_REDUC_BITWISE (int64_t, xor, ^=, 128)
+DEF_REDUC_BITWISE (int64_t, xor, ^=, 256)
+DEF_REDUC_BITWISE (int64_t, xor, ^=, 512)
+
+DEF_REDUC_BITWISE (uint64_t, xor, ^=, 4)
+DEF_REDUC_BITWISE (uint64_t, xor, ^=, 8)
+DEF_REDUC_BITWISE (uint64_t, xor, ^=, 16)
+DEF_REDUC_BITWISE (uint64_t, xor, ^=, 32)
+DEF_REDUC_BITWISE (uint64_t, xor, ^=, 64)
+DEF_REDUC_BITWISE (uint64_t, xor, ^=, 128)
+DEF_REDUC_BITWISE (uint64_t, xor, ^=, 256)
+DEF_REDUC_BITWISE (uint64_t, xor, ^=, 512)
+
+/* { dg-final { scan-assembler-times {vredand\.vs} 16 } } */
+/* { dg-final { scan-assembler-times {vredor\.vs} 16 } } */
+/* { dg-final { scan-assembler-times {vredxor\.vs} 16 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-19.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-19.c
new file mode 100644
index 0000000..a2de4d1
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-19.c
@@ -0,0 +1,18 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (_Float16, 4)
+DEF_REDUC_PLUS (_Float16, 8)
+DEF_REDUC_PLUS (_Float16, 16)
+DEF_REDUC_PLUS (_Float16, 32)
+DEF_REDUC_PLUS (_Float16, 64)
+DEF_REDUC_PLUS (_Float16, 128)
+DEF_REDUC_PLUS (_Float16, 256)
+DEF_REDUC_PLUS (_Float16, 512)
+DEF_REDUC_PLUS (_Float16, 1024)
+DEF_REDUC_PLUS (_Float16, 2048)
+
+/* { dg-final { scan-assembler-times {vfredosum\.vs} 10 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-2.c
new file mode 100644
index 0000000..9ffdec9
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-2.c
@@ -0,0 +1,29 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (int16_t, 4)
+DEF_REDUC_PLUS (int16_t, 8)
+DEF_REDUC_PLUS (int16_t, 16)
+DEF_REDUC_PLUS (int16_t, 32)
+DEF_REDUC_PLUS (int16_t, 64)
+DEF_REDUC_PLUS (int16_t, 128)
+DEF_REDUC_PLUS (int16_t, 256)
+DEF_REDUC_PLUS (int16_t, 512)
+DEF_REDUC_PLUS (int16_t, 1024)
+DEF_REDUC_PLUS (int16_t, 2048)
+
+DEF_REDUC_PLUS (uint16_t, 4)
+DEF_REDUC_PLUS (uint16_t, 8)
+DEF_REDUC_PLUS (uint16_t, 16)
+DEF_REDUC_PLUS (uint16_t, 32)
+DEF_REDUC_PLUS (uint16_t, 64)
+DEF_REDUC_PLUS (uint16_t, 128)
+DEF_REDUC_PLUS (uint16_t, 256)
+DEF_REDUC_PLUS (uint16_t, 512)
+DEF_REDUC_PLUS (uint16_t, 1024)
+DEF_REDUC_PLUS (uint16_t, 2048)
+
+/* { dg-final { scan-assembler-times {vredsum\.vs} 20 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-20.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-20.c
new file mode 100644
index 0000000..8e1164a
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-20.c
@@ -0,0 +1,17 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (float, 4)
+DEF_REDUC_PLUS (float, 8)
+DEF_REDUC_PLUS (float, 16)
+DEF_REDUC_PLUS (float, 32)
+DEF_REDUC_PLUS (float, 64)
+DEF_REDUC_PLUS (float, 128)
+DEF_REDUC_PLUS (float, 256)
+DEF_REDUC_PLUS (float, 512)
+DEF_REDUC_PLUS (float, 1024)
+
+/* { dg-final { scan-assembler-times {vfredosum\.vs} 9 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-21.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-21.c
new file mode 100644
index 0000000..727cf85
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-21.c
@@ -0,0 +1,16 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (float, 4)
+DEF_REDUC_PLUS (float, 8)
+DEF_REDUC_PLUS (float, 16)
+DEF_REDUC_PLUS (float, 32)
+DEF_REDUC_PLUS (float, 64)
+DEF_REDUC_PLUS (float, 128)
+DEF_REDUC_PLUS (float, 256)
+DEF_REDUC_PLUS (float, 512)
+
+/* { dg-final { scan-assembler-times {vfredosum\.vs} 8 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-3.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-3.c
new file mode 100644
index 0000000..63f9697
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-3.c
@@ -0,0 +1,27 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (int32_t, 4)
+DEF_REDUC_PLUS (int32_t, 8)
+DEF_REDUC_PLUS (int32_t, 16)
+DEF_REDUC_PLUS (int32_t, 32)
+DEF_REDUC_PLUS (int32_t, 64)
+DEF_REDUC_PLUS (int32_t, 128)
+DEF_REDUC_PLUS (int32_t, 256)
+DEF_REDUC_PLUS (int32_t, 512)
+DEF_REDUC_PLUS (int32_t, 1024)
+
+DEF_REDUC_PLUS (uint32_t, 4)
+DEF_REDUC_PLUS (uint32_t, 8)
+DEF_REDUC_PLUS (uint32_t, 16)
+DEF_REDUC_PLUS (uint32_t, 32)
+DEF_REDUC_PLUS (uint32_t, 64)
+DEF_REDUC_PLUS (uint32_t, 128)
+DEF_REDUC_PLUS (uint32_t, 256)
+DEF_REDUC_PLUS (uint32_t, 512)
+DEF_REDUC_PLUS (uint32_t, 1024)
+
+/* { dg-final { scan-assembler-times {vredsum\.vs} 18 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-4.c
new file mode 100644
index 0000000..8cd2728
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-4.c
@@ -0,0 +1,25 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (int64_t, 4)
+DEF_REDUC_PLUS (int64_t, 8)
+DEF_REDUC_PLUS (int64_t, 16)
+DEF_REDUC_PLUS (int64_t, 32)
+DEF_REDUC_PLUS (int64_t, 64)
+DEF_REDUC_PLUS (int64_t, 128)
+DEF_REDUC_PLUS (int64_t, 256)
+DEF_REDUC_PLUS (int64_t, 512)
+
+DEF_REDUC_PLUS (uint64_t, 4)
+DEF_REDUC_PLUS (uint64_t, 8)
+DEF_REDUC_PLUS (uint64_t, 16)
+DEF_REDUC_PLUS (uint64_t, 32)
+DEF_REDUC_PLUS (uint64_t, 64)
+DEF_REDUC_PLUS (uint64_t, 128)
+DEF_REDUC_PLUS (uint64_t, 256)
+DEF_REDUC_PLUS (uint64_t, 512)
+
+/* { dg-final { scan-assembler-times {vredsum\.vs} 16 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-5.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-5.c
new file mode 100644
index 0000000..09b69fc
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-5.c
@@ -0,0 +1,18 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8 -ffast-math" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (_Float16, 4)
+DEF_REDUC_PLUS (_Float16, 8)
+DEF_REDUC_PLUS (_Float16, 16)
+DEF_REDUC_PLUS (_Float16, 32)
+DEF_REDUC_PLUS (_Float16, 64)
+DEF_REDUC_PLUS (_Float16, 128)
+DEF_REDUC_PLUS (_Float16, 256)
+DEF_REDUC_PLUS (_Float16, 512)
+DEF_REDUC_PLUS (_Float16, 1024)
+DEF_REDUC_PLUS (_Float16, 2048)
+
+/* { dg-final { scan-assembler-times {vfredusum\.vs} 10 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-6.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-6.c
new file mode 100644
index 0000000..56b13dc
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-6.c
@@ -0,0 +1,17 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8 -ffast-math" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (float, 4)
+DEF_REDUC_PLUS (float, 8)
+DEF_REDUC_PLUS (float, 16)
+DEF_REDUC_PLUS (float, 32)
+DEF_REDUC_PLUS (float, 64)
+DEF_REDUC_PLUS (float, 128)
+DEF_REDUC_PLUS (float, 256)
+DEF_REDUC_PLUS (float, 512)
+DEF_REDUC_PLUS (float, 1024)
+
+/* { dg-final { scan-assembler-times {vfredusum\.vs} 9 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-7.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-7.c
new file mode 100644
index 0000000..eb5de35
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-7.c
@@ -0,0 +1,16 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8 -ffast-math" } */
+
+#include "def.h"
+
+DEF_REDUC_PLUS (double, 4)
+DEF_REDUC_PLUS (double, 8)
+DEF_REDUC_PLUS (double, 16)
+DEF_REDUC_PLUS (double, 32)
+DEF_REDUC_PLUS (double, 64)
+DEF_REDUC_PLUS (double, 128)
+DEF_REDUC_PLUS (double, 256)
+DEF_REDUC_PLUS (double, 512)
+
+/* { dg-final { scan-assembler-times {vfredusum\.vs} 8 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-8.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-8.c
new file mode 100644
index 0000000..cf4f9cc
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-8.c
@@ -0,0 +1,58 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_MAXMIN (int8_t, max, >, 4)
+DEF_REDUC_MAXMIN (int8_t, max, >, 8)
+DEF_REDUC_MAXMIN (int8_t, max, >, 16)
+DEF_REDUC_MAXMIN (int8_t, max, >, 32)
+DEF_REDUC_MAXMIN (int8_t, max, >, 64)
+DEF_REDUC_MAXMIN (int8_t, max, >, 128)
+DEF_REDUC_MAXMIN (int8_t, max, >, 256)
+DEF_REDUC_MAXMIN (int8_t, max, >, 512)
+DEF_REDUC_MAXMIN (int8_t, max, >, 1024)
+DEF_REDUC_MAXMIN (int8_t, max, >, 2048)
+DEF_REDUC_MAXMIN (int8_t, max, >, 4096)
+
+DEF_REDUC_MAXMIN (uint8_t, max, >, 4)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 8)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 16)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 32)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 64)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 128)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 256)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 512)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 1024)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 2048)
+DEF_REDUC_MAXMIN (uint8_t, max, >, 4096)
+
+DEF_REDUC_MAXMIN (int8_t, min, <, 4)
+DEF_REDUC_MAXMIN (int8_t, min, <, 8)
+DEF_REDUC_MAXMIN (int8_t, min, <, 16)
+DEF_REDUC_MAXMIN (int8_t, min, <, 32)
+DEF_REDUC_MAXMIN (int8_t, min, <, 64)
+DEF_REDUC_MAXMIN (int8_t, min, <, 128)
+DEF_REDUC_MAXMIN (int8_t, min, <, 256)
+DEF_REDUC_MAXMIN (int8_t, min, <, 512)
+DEF_REDUC_MAXMIN (int8_t, min, <, 1024)
+DEF_REDUC_MAXMIN (int8_t, min, <, 2048)
+DEF_REDUC_MAXMIN (int8_t, min, <, 4096)
+
+DEF_REDUC_MAXMIN (uint8_t, min, <, 4)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 8)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 16)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 32)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 64)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 128)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 256)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 512)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 1024)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 2048)
+DEF_REDUC_MAXMIN (uint8_t, min, <, 4096)
+
+/* { dg-final { scan-assembler-times {vredmax\.vs} 11 } } */
+/* { dg-final { scan-assembler-times {vredmaxu\.vs} 11 } } */
+/* { dg-final { scan-assembler-times {vredmin\.vs} 11 } } */
+/* { dg-final { scan-assembler-times {vredminu\.vs} 11 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-9.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-9.c
new file mode 100644
index 0000000..6416458
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/reduc-9.c
@@ -0,0 +1,54 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -O3 --param=riscv-autovec-lmul=m8" } */
+
+#include "def.h"
+
+DEF_REDUC_MAXMIN (int16_t, max, >, 4)
+DEF_REDUC_MAXMIN (int16_t, max, >, 8)
+DEF_REDUC_MAXMIN (int16_t, max, >, 16)
+DEF_REDUC_MAXMIN (int16_t, max, >, 32)
+DEF_REDUC_MAXMIN (int16_t, max, >, 64)
+DEF_REDUC_MAXMIN (int16_t, max, >, 128)
+DEF_REDUC_MAXMIN (int16_t, max, >, 256)
+DEF_REDUC_MAXMIN (int16_t, max, >, 512)
+DEF_REDUC_MAXMIN (int16_t, max, >, 1024)
+DEF_REDUC_MAXMIN (int16_t, max, >, 2048)
+
+DEF_REDUC_MAXMIN (uint16_t, max, >, 4)
+DEF_REDUC_MAXMIN (uint16_t, max, >, 8)
+DEF_REDUC_MAXMIN (uint16_t, max, >, 16)
+DEF_REDUC_MAXMIN (uint16_t, max, >, 32)
+DEF_REDUC_MAXMIN (uint16_t, max, >, 64)
+DEF_REDUC_MAXMIN (uint16_t, max, >, 128)
+DEF_REDUC_MAXMIN (uint16_t, max, >, 256)
+DEF_REDUC_MAXMIN (uint16_t, max, >, 512)
+DEF_REDUC_MAXMIN (uint16_t, max, >, 1024)
+DEF_REDUC_MAXMIN (uint16_t, max, >, 2048)
+
+DEF_REDUC_MAXMIN (int16_t, min, <, 4)
+DEF_REDUC_MAXMIN (int16_t, min, <, 8)
+DEF_REDUC_MAXMIN (int16_t, min, <, 16)
+DEF_REDUC_MAXMIN (int16_t, min, <, 32)
+DEF_REDUC_MAXMIN (int16_t, min, <, 64)
+DEF_REDUC_MAXMIN (int16_t, min, <, 128)
+DEF_REDUC_MAXMIN (int16_t, min, <, 256)
+DEF_REDUC_MAXMIN (int16_t, min, <, 512)
+DEF_REDUC_MAXMIN (int16_t, min, <, 1024)
+DEF_REDUC_MAXMIN (int16_t, min, <, 2048)
+
+DEF_REDUC_MAXMIN (uint16_t, min, <, 4)
+DEF_REDUC_MAXMIN (uint16_t, min, <, 8)
+DEF_REDUC_MAXMIN (uint16_t, min, <, 16)
+DEF_REDUC_MAXMIN (uint16_t, min, <, 32)
+DEF_REDUC_MAXMIN (uint16_t, min, <, 64)
+DEF_REDUC_MAXMIN (uint16_t, min, <, 128)
+DEF_REDUC_MAXMIN (uint16_t, min, <, 256)
+DEF_REDUC_MAXMIN (uint16_t, min, <, 512)
+DEF_REDUC_MAXMIN (uint16_t, min, <, 1024)
+DEF_REDUC_MAXMIN (uint16_t, min, <, 2048)
+
+/* { dg-final { scan-assembler-times {vredmax\.vs} 10 } } */
+/* { dg-final { scan-assembler-times {vredmaxu\.vs} 10 } } */
+/* { dg-final { scan-assembler-times {vredmin\.vs} 10 } } */
+/* { dg-final { scan-assembler-times {vredminu\.vs} 10 } } */
+/* { dg-final { scan-assembler-not {csrr} } } */