[PULL,50/63] target/riscv: vector widening floating-point reduction instructions

Commit Message

Alistair Francis June 26, 2020, 9:43 p.m. UTC

From: LIU Zhiwei <[e-mail]>

Signed-off-by: LIU Zhiwei <[e-mail]>
Reviewed-by: Richard Henderson <[e-mail]>
Reviewed-by: Alistair Francis <[e-mail]>
Message-id: [e-mail]
Signed-off-by: Alistair Francis <[e-mail]>
---
 target/riscv/helper.h                   |  3 ++
 target/riscv/insn32.decode              |  2 ++
 target/riscv/insn_trans/trans_rvv.inc.c |  3 ++
 target/riscv/vector_helper.c            | 46 +++++++++++++++++++++++++
 4 files changed, 54 insertions(+)

Patch

diff --git a/target/riscv/helper.h b/target/riscv/helper.h
index 8e8b8d0407..217f09a55c 100644
--- a/target/riscv/helper.h
+++ b/target/riscv/helper.h
@@ -1089,3 +1089,6 @@  DEF_HELPER_6(vfredmax_vs_d, void, ptr, ptr, ptr, ptr, env, i32)
 DEF_HELPER_6(vfredmin_vs_h, void, ptr, ptr, ptr, ptr, env, i32)
 DEF_HELPER_6(vfredmin_vs_w, void, ptr, ptr, ptr, ptr, env, i32)
 DEF_HELPER_6(vfredmin_vs_d, void, ptr, ptr, ptr, ptr, env, i32)
+
+DEF_HELPER_6(vfwredsum_vs_h, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(vfwredsum_vs_w, void, ptr, ptr, ptr, ptr, env, i32)
diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode
index 986308e99a..2668d483a7 100644
--- a/target/riscv/insn32.decode
+++ b/target/riscv/insn32.decode
@@ -545,6 +545,8 @@  vwredsum_vs     110001 . ..... ..... 000 ..... 1010111 @r_vm
 vfredsum_vs     0000-1 . ..... ..... 001 ..... 1010111 @r_vm
 vfredmin_vs     000101 . ..... ..... 001 ..... 1010111 @r_vm
 vfredmax_vs     000111 . ..... ..... 001 ..... 1010111 @r_vm
+# Vector widening ordered and unordered float reduction sum
+vfwredsum_vs    1100-1 . ..... ..... 001 ..... 1010111 @r_vm
 
 vsetvli         0 ........... ..... 111 ..... 1010111  @r2_zimm
 vsetvl          1000000 ..... ..... 111 ..... 1010111  @r
diff --git a/target/riscv/insn_trans/trans_rvv.inc.c b/target/riscv/insn_trans/trans_rvv.inc.c
index 3dd4171898..b78829be36 100644
--- a/target/riscv/insn_trans/trans_rvv.inc.c
+++ b/target/riscv/insn_trans/trans_rvv.inc.c
@@ -2351,3 +2351,6 @@  GEN_OPIVV_WIDEN_TRANS(vwredsumu_vs, reduction_check)
 GEN_OPFVV_TRANS(vfredsum_vs, reduction_check)
 GEN_OPFVV_TRANS(vfredmax_vs, reduction_check)
 GEN_OPFVV_TRANS(vfredmin_vs, reduction_check)
+
+/* Vector Widening Floating-Point Reduction Instructions */
+GEN_OPFVV_WIDEN_TRANS(vfwredsum_vs, reduction_check)
diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
index 0b395ad8d1..61d169398a 100644
--- a/target/riscv/vector_helper.c
+++ b/target/riscv/vector_helper.c
@@ -4456,3 +4456,49 @@  GEN_VEXT_FRED(vfredmax_vs_d, uint64_t, uint64_t, H8, H8, float64_maxnum, clearq)
 GEN_VEXT_FRED(vfredmin_vs_h, uint16_t, uint16_t, H2, H2, float16_minnum, clearh)
 GEN_VEXT_FRED(vfredmin_vs_w, uint32_t, uint32_t, H4, H4, float32_minnum, clearl)
 GEN_VEXT_FRED(vfredmin_vs_d, uint64_t, uint64_t, H8, H8, float64_minnum, clearq)
+
+/* Vector Widening Floating-Point Reduction Instructions */
+/* Unordered reduce 2*SEW = 2*SEW + sum(promote(SEW)) */
+void HELPER(vfwredsum_vs_h)(void *vd, void *v0, void *vs1,
+                            void *vs2, CPURISCVState *env, uint32_t desc)
+{
+    uint32_t mlen = vext_mlen(desc);
+    uint32_t vm = vext_vm(desc);
+    uint32_t vl = env->vl;
+    uint32_t i;
+    uint32_t tot = env_archcpu(env)->cfg.vlen / 8;
+    uint32_t s1 =  *((uint32_t *)vs1 + H4(0));
+
+    for (i = 0; i < vl; i++) {
+        uint16_t s2 = *((uint16_t *)vs2 + H2(i));
+        if (!vm && !vext_elem_mask(v0, mlen, i)) {
+            continue;
+        }
+        s1 = float32_add(s1, float16_to_float32(s2, true, &env->fp_status),
+                         &env->fp_status);
+    }
+    *((uint32_t *)vd + H4(0)) = s1;
+    clearl(vd, 1, sizeof(uint32_t), tot);
+}
+
+void HELPER(vfwredsum_vs_w)(void *vd, void *v0, void *vs1,
+                            void *vs2, CPURISCVState *env, uint32_t desc)
+{
+    uint32_t mlen = vext_mlen(desc);
+    uint32_t vm = vext_vm(desc);
+    uint32_t vl = env->vl;
+    uint32_t i;
+    uint32_t tot = env_archcpu(env)->cfg.vlen / 8;
+    uint64_t s1 =  *((uint64_t *)vs1);
+
+    for (i = 0; i < vl; i++) {
+        uint32_t s2 = *((uint32_t *)vs2 + H4(i));
+        if (!vm && !vext_elem_mask(v0, mlen, i)) {
+            continue;
+        }
+        s1 = float64_add(s1, float32_to_float64(s2, &env->fp_status),
+                         &env->fp_status);
+    }
+    *((uint64_t *)vd) = s1;
+    clearq(vd, 1, sizeof(uint64_t), tot);
+}