[v2,084/100] target/arm: Implement SVE mixed sign dot product (indexed)

Message ID 20200618042644.1685561-85-richard.henderson@linaro.org
State New
Headers show
Series
  • target/arm: Implement SVE2
Related show

Commit Message

Richard Henderson June 18, 2020, 4:26 a.m.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

---
 target/arm/cpu.h           |  5 +++
 target/arm/helper.h        |  4 +++
 target/arm/sve.decode      |  4 +++
 target/arm/translate-sve.c | 18 +++++++++++
 target/arm/vec_helper.c    | 66 ++++++++++++++++++++++++++++++++++++++
 5 files changed, 97 insertions(+)

-- 
2.25.1

Patch

diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index 331c5cdd4b..df0a3e201b 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -3877,6 +3877,11 @@  static inline bool isar_feature_aa64_sve2_bitperm(const ARMISARegisters *id)
     return FIELD_EX64(id->id_aa64zfr0, ID_AA64ZFR0, BITPERM) != 0;
 }
 
+static inline bool isar_feature_aa64_sve2_i8mm(const ARMISARegisters *id)
+{
+    return FIELD_EX64(id->id_aa64zfr0, ID_AA64ZFR0, I8MM) != 0;
+}
+
 static inline bool isar_feature_aa64_sve2_f32mm(const ARMISARegisters *id)
 {
     return FIELD_EX64(id->id_aa64zfr0, ID_AA64ZFR0, F32MM) != 0;
diff --git a/target/arm/helper.h b/target/arm/helper.h
index e9d7ab97da..6fac613dfc 100644
--- a/target/arm/helper.h
+++ b/target/arm/helper.h
@@ -587,6 +587,10 @@  DEF_HELPER_FLAGS_5(gvec_sdot_idx_h, TCG_CALL_NO_RWG,
                    void, ptr, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_5(gvec_udot_idx_h, TCG_CALL_NO_RWG,
                    void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(gvec_sudot_idx_b, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(gvec_usdot_idx_b, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
 
 DEF_HELPER_FLAGS_5(gvec_fcaddh, TCG_CALL_NO_RWG,
                    void, ptr, ptr, ptr, ptr, i32)
diff --git a/target/arm/sve.decode b/target/arm/sve.decode
index e8011fe91b..51acbfa797 100644
--- a/target/arm/sve.decode
+++ b/target/arm/sve.decode
@@ -815,6 +815,10 @@  SQRDMLSH_zzxz_h 01000100 .. 1 ..... 000101 ..... .....          @rrxr_h
 SQRDMLSH_zzxz_s 01000100 .. 1 ..... 000101 ..... .....          @rrxr_s
 SQRDMLSH_zzxz_d 01000100 .. 1 ..... 000101 ..... .....          @rrxr_d
 
+# SVE mixed sign dot product (indexed)
+USDOT_zzxw_s    01000100 .. 1 ..... 000110 ..... .....          @rrxr_s
+SUDOT_zzxw_s    01000100 .. 1 ..... 000111 ..... .....          @rrxr_s
+
 # SVE2 saturating multiply-add (indexed)
 SQDMLALB_zzxw_s 01000100 .. 1 ..... 0010.0 ..... .....          @rrxw_s
 SQDMLALB_zzxw_d 01000100 .. 1 ..... 0010.0 ..... .....          @rrxw_d
diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c
index 94c1e9aa05..fe4b4b7387 100644
--- a/target/arm/translate-sve.c
+++ b/target/arm/translate-sve.c
@@ -3842,6 +3842,24 @@  DO_RRXR(trans_SDOT_zzxw_d, gen_helper_gvec_sdot_idx_h)
 DO_RRXR(trans_UDOT_zzxw_s, gen_helper_gvec_udot_idx_b)
 DO_RRXR(trans_UDOT_zzxw_d, gen_helper_gvec_udot_idx_h)
 
+static bool trans_SUDOT_zzxw_s(DisasContext *s, arg_rrxr_esz *a)
+{
+    if (!dc_isar_feature(aa64_sve2_i8mm, s)) {
+        return false;
+    }
+    return do_zzxz_data(s, a->rd, a->rn, a->rm, a->ra,
+                        gen_helper_gvec_sudot_idx_b, a->index);
+}
+
+static bool trans_USDOT_zzxw_s(DisasContext *s, arg_rrxr_esz *a)
+{
+    if (!dc_isar_feature(aa64_sve2_i8mm, s)) {
+        return false;
+    }
+    return do_zzxz_data(s, a->rd, a->rn, a->rm, a->ra,
+                        gen_helper_gvec_usdot_idx_b, a->index);
+}
+
 #undef DO_RRXR
 
 static bool do_sve2_zzx_data(DisasContext *s, arg_rrx_esz *a,
diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c
index 8e85a16e7e..e1689d730f 100644
--- a/target/arm/vec_helper.c
+++ b/target/arm/vec_helper.c
@@ -678,6 +678,72 @@  void HELPER(gvec_udot_idx_b)(void *vd, void *vn, void *vm,
     clear_tail(d, opr_sz, simd_maxsz(desc));
 }
 
+void HELPER(gvec_sudot_idx_b)(void *vd, void *vn, void *vm,
+                              void *va, uint32_t desc)
+{
+    intptr_t i, segend, opr_sz = simd_oprsz(desc), opr_sz_4 = opr_sz / 4;
+    intptr_t index = simd_data(desc);
+    int32_t *d = vd, *a = va;
+    int8_t *n = vn;
+    uint8_t *m_indexed = (uint8_t *)vm + index * 4;
+
+    /* Notice the special case of opr_sz == 8, from aa64/aa32 advsimd.
+     * Otherwise opr_sz is a multiple of 16.
+     */
+    segend = MIN(4, opr_sz_4);
+    i = 0;
+    do {
+        uint8_t m0 = m_indexed[i * 4 + 0];
+        uint8_t m1 = m_indexed[i * 4 + 1];
+        uint8_t m2 = m_indexed[i * 4 + 2];
+        uint8_t m3 = m_indexed[i * 4 + 3];
+
+        do {
+            d[i] = (a[i] +
+                    n[i * 4 + 0] * m0 +
+                    n[i * 4 + 1] * m1 +
+                    n[i * 4 + 2] * m2 +
+                    n[i * 4 + 3] * m3);
+        } while (++i < segend);
+        segend = i + 4;
+    } while (i < opr_sz_4);
+
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}
+
+void HELPER(gvec_usdot_idx_b)(void *vd, void *vn, void *vm,
+                              void *va, uint32_t desc)
+{
+    intptr_t i, segend, opr_sz = simd_oprsz(desc), opr_sz_4 = opr_sz / 4;
+    intptr_t index = simd_data(desc);
+    uint32_t *d = vd, *a = va;
+    uint8_t *n = vn;
+    int8_t *m_indexed = (int8_t *)vm + index * 4;
+
+    /* Notice the special case of opr_sz == 8, from aa64/aa32 advsimd.
+     * Otherwise opr_sz is a multiple of 16.
+     */
+    segend = MIN(4, opr_sz_4);
+    i = 0;
+    do {
+        int8_t m0 = m_indexed[i * 4 + 0];
+        int8_t m1 = m_indexed[i * 4 + 1];
+        int8_t m2 = m_indexed[i * 4 + 2];
+        int8_t m3 = m_indexed[i * 4 + 3];
+
+        do {
+            d[i] = (a[i] +
+                    n[i * 4 + 0] * m0 +
+                    n[i * 4 + 1] * m1 +
+                    n[i * 4 + 2] * m2 +
+                    n[i * 4 + 3] * m3);
+        } while (++i < segend);
+        segend = i + 4;
+    } while (i < opr_sz_4);
+
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}
+
 void HELPER(gvec_sdot_idx_h)(void *vd, void *vn, void *vm,
                              void *va, uint32_t desc)
 {