[PATCH v3 11/37] target/riscv: SIMD 16-bit Miscellaneous Instructions

From: LIU Zhiwei <zhiwei_liu@c-sky.com>
To: qemu-devel@nongnu.org, qemu-riscv@nongnu.org
Cc: palmer@dabbelt.com, bin.meng@windriver.com,
	Alistair.Francis@wdc.com, LIU Zhiwei <zhiwei_liu@c-sky.com>
Subject: [PATCH v3 11/37] target/riscv: SIMD 16-bit Miscellaneous Instructions
Date: Thu, 24 Jun 2021 18:54:55 +0800	[thread overview]
Message-ID: <20210624105521.3964-12-zhiwei_liu@c-sky.com> (raw)
In-Reply-To: <20210624105521.3964-1-zhiwei_liu@c-sky.com>

There are 11 instructions, including signed or unsigned
minimum, maximum, clip value, absolute value, and leading
zero, leading one count instructions.

Signed-off-by: LIU Zhiwei <zhiwei_liu@c-sky.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
---
 target/riscv/helper.h                   |  11 ++
 target/riscv/insn32.decode              |  11 ++
 target/riscv/insn_trans/trans_rvp.c.inc |  41 ++++++
 target/riscv/packed_helper.c            | 158 ++++++++++++++++++++++++
 4 files changed, 221 insertions(+)

diff --git a/target/riscv/helper.h b/target/riscv/helper.h
index 4d0918b9a9..88035aafad 100644
--- a/target/riscv/helper.h
+++ b/target/riscv/helper.h
@@ -1232,3 +1232,14 @@ DEF_HELPER_3(umul8, i64, env, tl, tl)
 DEF_HELPER_3(umulx8, i64, env, tl, tl)
 DEF_HELPER_3(khm8, tl, env, tl, tl)
 DEF_HELPER_3(khmx8, tl, env, tl, tl)
+
+DEF_HELPER_3(smin16, tl, env, tl, tl)
+DEF_HELPER_3(umin16, tl, env, tl, tl)
+DEF_HELPER_3(smax16, tl, env, tl, tl)
+DEF_HELPER_3(umax16, tl, env, tl, tl)
+DEF_HELPER_3(sclip16, tl, env, tl, tl)
+DEF_HELPER_3(uclip16, tl, env, tl, tl)
+DEF_HELPER_2(kabs16, tl, env, tl)
+DEF_HELPER_2(clrs16, tl, env, tl)
+DEF_HELPER_2(clz16, tl, env, tl)
+DEF_HELPER_2(clo16, tl, env, tl)
diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode
index 05c3e67477..847c796874 100644
--- a/target/riscv/insn32.decode
+++ b/target/riscv/insn32.decode
@@ -835,3 +835,14 @@ umul8      1011100  ..... ..... 000 ..... 1110111 @r
 umulx8     1011101  ..... ..... 000 ..... 1110111 @r
 khm8       1000111  ..... ..... 000 ..... 1110111 @r
 khmx8      1001111  ..... ..... 000 ..... 1110111 @r
+
+smin16     1000000  ..... ..... 000 ..... 1110111 @r
+umin16     1001000  ..... ..... 000 ..... 1110111 @r
+smax16     1000001  ..... ..... 000 ..... 1110111 @r
+umax16     1001001  ..... ..... 000 ..... 1110111 @r
+sclip16    1000010  0.... ..... 000 ..... 1110111 @sh4
+uclip16    1000010  1.... ..... 000 ..... 1110111 @sh4
+kabs16     1010110  10001 ..... 000 ..... 1110111 @r2
+clrs16     1010111  01000 ..... 000 ..... 1110111 @r2
+clz16      1010111  01001 ..... 000 ..... 1110111 @r2
+clo16      1010111  01011 ..... 000 ..... 1110111 @r2
diff --git a/target/riscv/insn_trans/trans_rvp.c.inc b/target/riscv/insn_trans/trans_rvp.c.inc
index 2188de8505..3e6307cdc3 100644
--- a/target/riscv/insn_trans/trans_rvp.c.inc
+++ b/target/riscv/insn_trans/trans_rvp.c.inc
@@ -294,3 +294,44 @@ GEN_RVP_R_D64_OOL(umul8);
 GEN_RVP_R_D64_OOL(umulx8);
 GEN_RVP_R_OOL(khm8);
 GEN_RVP_R_OOL(khmx8);
+
+/* SIMD 16-bit Miscellaneous Instructions */
+GEN_RVP_R_OOL(smin16);
+GEN_RVP_R_OOL(umin16);
+GEN_RVP_R_OOL(smax16);
+GEN_RVP_R_OOL(umax16);
+GEN_RVP_SHIFTI(sclip16, NULL, gen_helper_sclip16);
+GEN_RVP_SHIFTI(uclip16, NULL, gen_helper_uclip16);
+
+/* Out of line helpers for R2 format */
+static bool
+r2_ool(DisasContext *ctx, arg_r2 *a,
+       void (* fn)(TCGv, TCGv_ptr, TCGv))
+{
+    TCGv src1, dst;
+    if (!has_ext(ctx, RVP)) {
+        return false;
+    }
+
+    src1 = tcg_temp_new();
+    dst = tcg_temp_new();
+
+    gen_get_gpr(src1, a->rs1);
+    fn(dst, cpu_env, src1);
+    gen_set_gpr(a->rd, dst);
+
+    tcg_temp_free(src1);
+    tcg_temp_free(dst);
+    return true;
+}
+
+#define GEN_RVP_R2_OOL(NAME)                           \
+static bool trans_##NAME(DisasContext *s, arg_r2 *a)   \
+{                                                      \
+    return r2_ool(s, a, gen_helper_##NAME);            \
+}
+
+GEN_RVP_R2_OOL(kabs16);
+GEN_RVP_R2_OOL(clrs16);
+GEN_RVP_R2_OOL(clz16);
+GEN_RVP_R2_OOL(clo16);
diff --git a/target/riscv/packed_helper.c b/target/riscv/packed_helper.c
index 56baefeb8e..e4a9463135 100644
--- a/target/riscv/packed_helper.c
+++ b/target/riscv/packed_helper.c
@@ -920,3 +920,161 @@ static inline void do_khmx8(CPURISCVState *env, void *vd, void *va,
 }
 
 RVPR(khmx8, 2, 1);
+
+/* SIMD 16-bit Miscellaneous Instructions */
+static inline void do_smin16(CPURISCVState *env, void *vd, void *va,
+                             void *vb, uint8_t i)
+{
+    int16_t *d = vd, *a = va, *b = vb;
+
+    d[i] = (a[i] < b[i]) ? a[i] : b[i];
+}
+
+RVPR(smin16, 1, 2);
+
+static inline void do_umin16(CPURISCVState *env, void *vd, void *va,
+                             void *vb, uint8_t i)
+{
+    uint16_t *d = vd, *a = va, *b = vb;
+
+    d[i] = (a[i] < b[i]) ? a[i] : b[i];
+}
+
+RVPR(umin16, 1, 2);
+
+static inline void do_smax16(CPURISCVState *env, void *vd, void *va,
+                             void *vb, uint8_t i)
+{
+    int16_t *d = vd, *a = va, *b = vb;
+
+    d[i] = (a[i] > b[i]) ? a[i] : b[i];
+}
+
+RVPR(smax16, 1, 2);
+
+static inline void do_umax16(CPURISCVState *env, void *vd, void *va,
+                             void *vb, uint8_t i)
+{
+    uint16_t *d = vd, *a = va, *b = vb;
+
+    d[i] = (a[i] > b[i]) ? a[i] : b[i];
+}
+
+RVPR(umax16, 1, 2);
+
+static int64_t sat64(CPURISCVState *env, int64_t a, uint8_t shift)
+{
+    int64_t max = shift >= 64 ? INT64_MAX : (1ull << shift) - 1;
+    int64_t min = shift >= 64 ? INT64_MIN : -(1ull << shift);
+    int64_t result;
+
+    if (a > max) {
+        result = max;
+        env->vxsat = 0x1;
+    } else if (a < min) {
+        result = min;
+        env->vxsat = 0x1;
+    } else {
+        result = a;
+    }
+    return result;
+}
+
+static inline void do_sclip16(CPURISCVState *env, void *vd, void *va,
+                              void *vb, uint8_t i)
+{
+    int16_t *d = vd, *a = va;
+    uint8_t shift = *(uint8_t *)vb & 0xf;
+
+    d[i] = sat64(env, a[i], shift);
+}
+
+RVPR(sclip16, 1, 2);
+
+static uint64_t satu64(CPURISCVState *env, uint64_t a, uint8_t shift)
+{
+    uint64_t max = shift >= 64 ? UINT64_MAX : (1ull << shift) - 1;
+    uint64_t result;
+
+    if (a > max) {
+        result = max;
+        env->vxsat = 0x1;
+    } else {
+        result = a;
+    }
+    return result;
+}
+
+static inline void do_uclip16(CPURISCVState *env, void *vd, void *va,
+                              void *vb, uint8_t i)
+{
+    int16_t *d = vd, *a = va;
+    uint8_t shift = *(uint8_t *)vb & 0xf;
+
+    if (a[i] < 0) {
+        d[i] = 0;
+        env->vxsat = 0x1;
+    } else {
+        d[i] = satu64(env, a[i], shift);
+    }
+}
+
+RVPR(uclip16, 1, 2);
+
+typedef void PackedFn2i(CPURISCVState *, void *, void *, uint8_t);
+
+static inline target_ulong rvpr2(CPURISCVState *env, target_ulong a,
+                                 uint8_t step, uint8_t size, PackedFn2i *fn)
+{
+    int i, passes = sizeof(target_ulong) / size;
+    target_ulong result;
+
+    for (i = 0; i < passes; i += step) {
+        fn(env, &result, &a, i);
+    }
+    return result;
+}
+
+#define RVPR2(NAME, STEP, SIZE)                                  \
+target_ulong HELPER(NAME)(CPURISCVState *env, target_ulong a)    \
+{                                                                \
+    return rvpr2(env, a, STEP, SIZE, (PackedFn2i *)do_##NAME);   \
+}
+
+static inline void do_kabs16(CPURISCVState *env, void *vd, void *va, uint8_t i)
+{
+    int16_t *d = vd, *a = va;
+
+    if (a[i] == INT16_MIN) {
+        d[i] = INT16_MAX;
+        env->vxsat = 0x1;
+    } else {
+        d[i] = abs(a[i]);
+    }
+}
+
+RVPR2(kabs16, 1, 2);
+
+static inline void do_clrs16(CPURISCVState *env, void *vd, void *va, uint8_t i)
+{
+    int16_t *d = vd, *a = va;
+    d[i] = clrsb32(a[i]) - 16;
+}
+
+RVPR2(clrs16, 1, 2);
+
+static inline void do_clz16(CPURISCVState *env, void *vd, void *va, uint8_t i)
+{
+    int16_t *d = vd, *a = va;
+    d[i] = (a[i] < 0) ? 0 : (clz32(a[i]) - 16);
+}
+
+RVPR2(clz16, 1, 2);
+
+static inline void do_clo16(CPURISCVState *env, void *vd, void *va, uint8_t i)
+{
+    int16_t *d = vd, *a = va;
+    d[i] = (a[i] >= 0) ? 0 : (clo32(a[i]) - 16);
+}
+
+RVPR2(clo16, 1, 2);
-- 
2.17.1


