gcc/testsuite/gcc.target/riscv/rvv/base/pr112431-24.c

   1 /* { dg-do compile } */
   2 /* { dg-options "-march=rv64gcv -mabi=lp64d -O3" } */
   3
   4 #include "riscv_vector.h"
   5
   6 size_t __attribute__ ((noinline))
   7 sumation (size_t sum0, size_t sum1, size_t sum2, size_t sum3)
   8 {
   9   return sum0 + sum1 + sum2 + sum3;
  10 }
  11
  12 size_t
  13 foo (char const *buf, size_t len)
  14 {
  15   size_t sum = 0;
  16   size_t vl = __riscv_vsetvlmax_e8m8 ();
  17   size_t step = vl * 4;
  18   const char *it = buf, *end = buf + len;
  19   for (; it + step <= end;)
  20     {
  21       vint8m4_t v0 = __riscv_vle8_v_i8m4 ((void *) it, vl);
  22       it += vl;
  23       vint8m4_t v1 = __riscv_vle8_v_i8m4 ((void *) it, vl);
  24       it += vl;
  25       vint8m4_t v2 = __riscv_vle8_v_i8m4 ((void *) it, vl);
  26       it += vl;
  27       vint8m4_t v3 = __riscv_vle8_v_i8m4 ((void *) it, vl);
  28       it += vl;
  29
  30       asm volatile("nop" ::: "memory");
  31       vint16m8_t vw0 = __riscv_vwadd_vx_i16m8 (v0, 66, vl);
  32       vint16m8_t vw1 = __riscv_vwadd_vx_i16m8 (v1, 66, vl);
  33       vint16m8_t vw2 = __riscv_vwadd_vx_i16m8 (v2, 66, vl);
  34       vint16m8_t vw3 = __riscv_vwadd_vx_i16m8 (v3, 66, vl);
  35
  36       asm volatile("nop" ::: "memory");
  37       size_t sum0 = __riscv_vmv_x_s_i16m8_i16 (vw0);
  38       size_t sum1 = __riscv_vmv_x_s_i16m8_i16 (vw1);
  39       size_t sum2 = __riscv_vmv_x_s_i16m8_i16 (vw2);
  40       size_t sum3 = __riscv_vmv_x_s_i16m8_i16 (vw3);
  41
  42       sum += sumation (sum0, sum1, sum2, sum3);
  43     }
  44   return sum;
  45 }
  46
  47 size_t
  48 foo2 (char const *buf, size_t len)
  49 {
  50   size_t sum = 0;
  51   size_t vl = __riscv_vsetvlmax_e8m8 ();
  52   size_t step = vl * 4;
  53   const char *it = buf, *end = buf + len;
  54   for (; it + step <= end;)
  55     {
  56       vint8m4_t v0 = __riscv_vle8_v_i8m4 ((void *) it, vl);
  57       it += vl;
  58       vint8m4_t v1 = __riscv_vle8_v_i8m4 ((void *) it, vl);
  59       it += vl;
  60       vint8m4_t v2 = __riscv_vle8_v_i8m4 ((void *) it, vl);
  61       it += vl;
  62       vint8m4_t v3 = __riscv_vle8_v_i8m4 ((void *) it, vl);
  63       it += vl;
  64
  65       asm volatile("nop" ::: "memory");
  66       vint16m8_t vw0 = __riscv_vwmulsu_vx_i16m8 (v0, 66, vl);
  67       vint16m8_t vw1 = __riscv_vwmulsu_vx_i16m8 (v1, 66, vl);
  68       vint16m8_t vw2 = __riscv_vwmulsu_vx_i16m8 (v2, 66, vl);
  69       vint16m8_t vw3 = __riscv_vwmulsu_vx_i16m8 (v3, 66, vl);
  70
  71       asm volatile("nop" ::: "memory");
  72       size_t sum0 = __riscv_vmv_x_s_i16m8_i16 (vw0);
  73       size_t sum1 = __riscv_vmv_x_s_i16m8_i16 (vw1);
  74       size_t sum2 = __riscv_vmv_x_s_i16m8_i16 (vw2);
  75       size_t sum3 = __riscv_vmv_x_s_i16m8_i16 (vw3);
  76
  77       sum += sumation (sum0, sum1, sum2, sum3);
  78     }
  79   return sum;
  80 }
  81
  82 /* { dg-final { scan-assembler-not {vmv1r} } } */
  83 /* { dg-final { scan-assembler-not {vmv2r} } } */
  84 /* { dg-final { scan-assembler-not {vmv4r} } } */
  85 /* { dg-final { scan-assembler-not {vmv8r} } } */
  86 /* { dg-final { scan-assembler-not {csrr} } } */