[PATCH 23/31] target/arm: Create arm_gen_gvec_[us]sra

[Richard Henderson]

The functions eliminate duplication of the special cases for
this operation.  They match up with the GVecGen2iFn typedef.

Add out-of-line helpers.  We got away with only having inline
expanders because the neon vector size is only 16 bytes, and
we know that the inline expansion will always succeed.
When we reuse this for SVE, tcg-gvec-op may decide to use an
out-of-line helper due to longer vector lengths.

Signed-off-by: Richard Henderson <address@hidden>
---
 target/arm/helper.h        |  10 +++
 target/arm/translate.h     |   7 +-
 target/arm/translate-a64.c |  15 +---
 target/arm/translate.c     | 161 ++++++++++++++++++++++---------------
 target/arm/vec_helper.c    |  25 ++++++
 5 files changed, 139 insertions(+), 79 deletions(-)

diff --git a/target/arm/helper.h b/target/arm/helper.h
index 938fdbc362..dc6a43dbd8 100644
--- a/target/arm/helper.h
+++ b/target/arm/helper.h
@@ -708,6 +708,16 @@ DEF_HELPER_FLAGS_4(gvec_pmull_q, TCG_CALL_NO_RWG, void, 
ptr, ptr, ptr, i32)
 
 DEF_HELPER_FLAGS_4(neon_pmull_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 
+DEF_HELPER_FLAGS_3(gvec_ssra_b, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(gvec_ssra_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(gvec_ssra_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(gvec_ssra_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_3(gvec_usra_b, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(gvec_usra_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(gvec_usra_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(gvec_usra_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+
 #ifdef TARGET_AARCH64
 #include "helper-a64.h"
 #include "helper-sve.h"
diff --git a/target/arm/translate.h b/target/arm/translate.h
index 5552ee5a94..1c5cdf13e3 100644
--- a/target/arm/translate.h
+++ b/target/arm/translate.h
@@ -291,8 +291,6 @@ extern const GVecGen3 mls_op[4];
 extern const GVecGen3 cmtst_op[4];
 extern const GVecGen3 sshl_op[4];
 extern const GVecGen3 ushl_op[4];
-extern const GVecGen2i ssra_op[4];
-extern const GVecGen2i usra_op[4];
 extern const GVecGen2i sri_op[4];
 extern const GVecGen2i sli_op[4];
 extern const GVecGen4 uqadd_op[4];
@@ -305,6 +303,11 @@ void gen_sshl_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b);
 void gen_ushl_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b);
 void gen_sshl_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b);
 
+void arm_gen_gvec_ssra(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
+                       int64_t shift, uint32_t opr_sz, uint32_t max_sz);
+void arm_gen_gvec_usra(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
+                       int64_t shift, uint32_t opr_sz, uint32_t max_sz);
+
 /*
  * Forward to the isar_feature_* tests given a DisasContext pointer.
  */
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 2bcf643069..d50207fcfb 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -10682,19 +10682,8 @@ static void handle_vec_simd_shri(DisasContext *s, bool 
is_q, bool is_u,
 
     switch (opcode) {
     case 0x02: /* SSRA / USRA (accumulate) */
-        if (is_u) {
-            /* Shift count same as element size produces zero to add.  */
-            if (shift == 8 << size) {
-                goto done;
-            }
-            gen_gvec_op2i(s, is_q, rd, rn, shift, &usra_op[size]);
-        } else {
-            /* Shift count same as element size produces all sign to add.  */
-            if (shift == 8 << size) {
-                shift -= 1;
-            }
-            gen_gvec_op2i(s, is_q, rd, rn, shift, &ssra_op[size]);
-        }
+        gen_gvec_fn2i(s, is_q, rd, rn, shift,
+                      is_u ? arm_gen_gvec_usra : arm_gen_gvec_ssra, size);
         return;
     case 0x08: /* SRI */
         /* Shift count same as element size is valid but does nothing.  */
diff --git a/target/arm/translate.c b/target/arm/translate.c
index cba84987db..f5768014d1 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -3947,33 +3947,51 @@ static void gen_ssra_vec(unsigned vece, TCGv_vec d, 
TCGv_vec a, int64_t sh)
     tcg_gen_add_vec(vece, d, d, a);
 }
 
-static const TCGOpcode vecop_list_ssra[] = {
-    INDEX_op_sari_vec, INDEX_op_add_vec, 0
-};
+void arm_gen_gvec_ssra(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
+                       int64_t shift, uint32_t opr_sz, uint32_t max_sz)
+{
+    static const TCGOpcode vecop_list[] = {
+        INDEX_op_sari_vec, INDEX_op_add_vec, 0
+    };
+    static const GVecGen2i ops[4] = {
+        { .fni8 = gen_ssra8_i64,
+          .fniv = gen_ssra_vec,
+          .fno = gen_helper_gvec_ssra_b,
+          .load_dest = true,
+          .opt_opc = vecop_list,
+          .vece = MO_8 },
+        { .fni8 = gen_ssra16_i64,
+          .fniv = gen_ssra_vec,
+          .fno = gen_helper_gvec_ssra_h,
+          .load_dest = true,
+          .opt_opc = vecop_list,
+          .vece = MO_16 },
+        { .fni4 = gen_ssra32_i32,
+          .fniv = gen_ssra_vec,
+          .fno = gen_helper_gvec_ssra_s,
+          .load_dest = true,
+          .opt_opc = vecop_list,
+          .vece = MO_32 },
+        { .fni8 = gen_ssra64_i64,
+          .fniv = gen_ssra_vec,
+          .fno = gen_helper_gvec_ssra_b,
+          .prefer_i64 = TCG_TARGET_REG_BITS == 64,
+          .opt_opc = vecop_list,
+          .load_dest = true,
+          .vece = MO_64 },
+    };
 
-const GVecGen2i ssra_op[4] = {
-    { .fni8 = gen_ssra8_i64,
-      .fniv = gen_ssra_vec,
-      .load_dest = true,
-      .opt_opc = vecop_list_ssra,
-      .vece = MO_8 },
-    { .fni8 = gen_ssra16_i64,
-      .fniv = gen_ssra_vec,
-      .load_dest = true,
-      .opt_opc = vecop_list_ssra,
-      .vece = MO_16 },
-    { .fni4 = gen_ssra32_i32,
-      .fniv = gen_ssra_vec,
-      .load_dest = true,
-      .opt_opc = vecop_list_ssra,
-      .vece = MO_32 },
-    { .fni8 = gen_ssra64_i64,
-      .fniv = gen_ssra_vec,
-      .prefer_i64 = TCG_TARGET_REG_BITS == 64,
-      .opt_opc = vecop_list_ssra,
-      .load_dest = true,
-      .vece = MO_64 },
-};
+    /* tszimm encoding produces immediates in the range [1..esize]. */
+    tcg_debug_assert(shift > 0);
+    tcg_debug_assert(shift <= (8 << vece));
+
+    /*
+     * Shifts larger than the element size are architecturally valid.
+     * Signed results in all sign bits.
+     */
+    shift = MIN(shift, (8 << vece) - 1);
+    tcg_gen_gvec_2i(rd_ofs, rm_ofs, opr_sz, max_sz, shift, &ops[vece]);
+}
 
 static void gen_usra8_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift)
 {
@@ -4005,33 +4023,55 @@ static void gen_usra_vec(unsigned vece, TCGv_vec d, 
TCGv_vec a, int64_t sh)
     tcg_gen_add_vec(vece, d, d, a);
 }
 
-static const TCGOpcode vecop_list_usra[] = {
-    INDEX_op_shri_vec, INDEX_op_add_vec, 0
-};
+void arm_gen_gvec_usra(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
+                       int64_t shift, uint32_t opr_sz, uint32_t max_sz)
+{
+    static const TCGOpcode vecop_list[] = {
+        INDEX_op_shri_vec, INDEX_op_add_vec, 0
+    };
+    static const GVecGen2i ops[4] = {
+        { .fni8 = gen_usra8_i64,
+          .fniv = gen_usra_vec,
+          .fno = gen_helper_gvec_usra_b,
+          .load_dest = true,
+          .opt_opc = vecop_list,
+          .vece = MO_8, },
+        { .fni8 = gen_usra16_i64,
+          .fniv = gen_usra_vec,
+          .fno = gen_helper_gvec_usra_h,
+          .load_dest = true,
+          .opt_opc = vecop_list,
+          .vece = MO_16, },
+        { .fni4 = gen_usra32_i32,
+          .fniv = gen_usra_vec,
+          .fno = gen_helper_gvec_usra_s,
+          .load_dest = true,
+          .opt_opc = vecop_list,
+          .vece = MO_32, },
+        { .fni8 = gen_usra64_i64,
+          .fniv = gen_usra_vec,
+          .fno = gen_helper_gvec_usra_d,
+          .prefer_i64 = TCG_TARGET_REG_BITS == 64,
+          .load_dest = true,
+          .opt_opc = vecop_list,
+          .vece = MO_64, },
+    };
 
-const GVecGen2i usra_op[4] = {
-    { .fni8 = gen_usra8_i64,
-      .fniv = gen_usra_vec,
-      .load_dest = true,
-      .opt_opc = vecop_list_usra,
-      .vece = MO_8, },
-    { .fni8 = gen_usra16_i64,
-      .fniv = gen_usra_vec,
-      .load_dest = true,
-      .opt_opc = vecop_list_usra,
-      .vece = MO_16, },
-    { .fni4 = gen_usra32_i32,
-      .fniv = gen_usra_vec,
-      .load_dest = true,
-      .opt_opc = vecop_list_usra,
-      .vece = MO_32, },
-    { .fni8 = gen_usra64_i64,
-      .fniv = gen_usra_vec,
-      .prefer_i64 = TCG_TARGET_REG_BITS == 64,
-      .load_dest = true,
-      .opt_opc = vecop_list_usra,
-      .vece = MO_64, },
-};
+    /* tszimm encoding produces immediates in the range [1..esize]. */
+    tcg_debug_assert(shift > 0);
+    tcg_debug_assert(shift <= (8 << vece));
+
+    /*
+     * Shifts larger than the element size are architecturally valid.
+     * Unsigned results in all zeros as input to accumulate: nop.
+     */
+    if (shift < (8 << vece)) {
+        tcg_gen_gvec_2i(rd_ofs, rm_ofs, opr_sz, max_sz, shift, &ops[vece]);
+    } else {
+        /* Nop, but we do need to clear the tail. */
+        tcg_gen_gvec_mov(vece, rd_ofs, rd_ofs, opr_sz, max_sz);
+    }
+}
 
 static void gen_shr8_ins_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift)
 {
@@ -5396,19 +5436,12 @@ static int disas_neon_data_insn(DisasContext *s, 
uint32_t insn)
                 case 1:  /* VSRA */
                     /* Right shift comes here negative.  */
                     shift = -shift;
-                    /* Shifts larger than the element size are architecturally
-                     * valid.  Unsigned results in all zeros; signed results
-                     * in all sign bits.
-                     */
-                    if (!u) {
-                        tcg_gen_gvec_2i(rd_ofs, rm_ofs, vec_size, vec_size,
-                                        MIN(shift, (8 << size) - 1),
-                                        &ssra_op[size]);
-                    } else if (shift >= 8 << size) {
-                        /* rd += 0 */
+                    if (u) {
+                        arm_gen_gvec_usra(size, rd_ofs, rm_ofs, shift,
+                                          vec_size, vec_size);
                     } else {
-                        tcg_gen_gvec_2i(rd_ofs, rm_ofs, vec_size, vec_size,
-                                        shift, &usra_op[size]);
+                        arm_gen_gvec_ssra(size, rd_ofs, rm_ofs, shift,
+                                          vec_size, vec_size);
                     }
                     return 0;
 
diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c
index 154d32518a..aaaccc0a2d 100644
--- a/target/arm/vec_helper.c
+++ b/target/arm/vec_helper.c
@@ -899,6 +899,31 @@ void HELPER(gvec_sqsub_d)(void *vd, void *vq, void *vn,
     clear_tail(d, oprsz, simd_maxsz(desc));
 }
 
+
+#define DO_SRA(NAME, TYPE)                              \
+void HELPER(NAME)(void *vd, void *vn, uint32_t desc)    \
+{                                                       \
+    intptr_t i, oprsz = simd_oprsz(desc);               \
+    int shift = simd_data(desc);                        \
+    TYPE *d = vd, *n = vn;                              \
+    for (i = 0; i < oprsz / sizeof(TYPE); i++) {        \
+        d[i] += n[i] >> shift;                          \
+    }                                                   \
+    clear_tail(d, oprsz, simd_maxsz(desc));             \
+}
+
+DO_SRA(gvec_ssra_b, int8_t)
+DO_SRA(gvec_ssra_h, int16_t)
+DO_SRA(gvec_ssra_s, int32_t)
+DO_SRA(gvec_ssra_d, int64_t)
+
+DO_SRA(gvec_usra_b, uint8_t)
+DO_SRA(gvec_usra_h, uint16_t)
+DO_SRA(gvec_usra_s, uint32_t)
+DO_SRA(gvec_usra_d, uint64_t)
+
+#undef DO_SRA
+
 /*
  * Convert float16 to float32, raising no exceptions and
  * preserving exceptional values, including SNaN.
-- 
2.20.1