tcg: Implement gvec support for rotate by immediate

No host backend support yet, but the interfaces for rotli
are in place.  Canonicalize immediate rotate to the left,
based on a survey of architectures, but provide both left
and right shift interfaces to the translators.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

13 files changed, 150 insertions, 1 deletions

diff --git a/accel/tcg/tcg-runtime-gvec.c b/accel/tcg/tcg-runtime-gvec.c
index ca44970..34b1030 100644
--- a/accel/tcg/tcg-runtime-gvec.c
+++ b/accel/tcg/tcg-runtime-gvec.c
@@ -716,6 +716,54 @@ void HELPER(gvec_sar64i)(void *d, void *a, uint32_t desc)
     clear_high(d, oprsz, desc);
 }
 
+void HELPER(gvec_rotl8i)(void *d, void *a, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    int shift = simd_data(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint8_t)) {
+        *(uint8_t *)(d + i) = rol8(*(uint8_t *)(a + i), shift);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotl16i)(void *d, void *a, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    int shift = simd_data(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint16_t)) {
+        *(uint16_t *)(d + i) = rol16(*(uint16_t *)(a + i), shift);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotl32i)(void *d, void *a, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    int shift = simd_data(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint32_t)) {
+        *(uint32_t *)(d + i) = rol32(*(uint32_t *)(a + i), shift);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotl64i)(void *d, void *a, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    int shift = simd_data(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint64_t)) {
+        *(uint64_t *)(d + i) = rol64(*(uint64_t *)(a + i), shift);
+    }
+    clear_high(d, oprsz, desc);
+}
+
 void HELPER(gvec_shl8v)(void *d, void *a, void *b, uint32_t desc)
 {
     intptr_t oprsz = simd_oprsz(desc);
diff --git a/accel/tcg/tcg-runtime.h b/accel/tcg/tcg-runtime.h
index 4fa61b4..cf10c83 100644
--- a/accel/tcg/tcg-runtime.h
+++ b/accel/tcg/tcg-runtime.h
@@ -259,6 +259,11 @@ DEF_HELPER_FLAGS_3(gvec_sar16i, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
 DEF_HELPER_FLAGS_3(gvec_sar32i, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
 DEF_HELPER_FLAGS_3(gvec_sar64i, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
 
+DEF_HELPER_FLAGS_3(gvec_rotl8i, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(gvec_rotl16i, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(gvec_rotl32i, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(gvec_rotl64i, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+
 DEF_HELPER_FLAGS_4(gvec_shl8v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(gvec_shl16v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(gvec_shl32v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
diff --git a/include/tcg/tcg-op-gvec.h b/include/tcg/tcg-op-gvec.h
index cea6497..1afc3eb 100644
--- a/include/tcg/tcg-op-gvec.h
+++ b/include/tcg/tcg-op-gvec.h
@@ -334,6 +334,10 @@ void tcg_gen_gvec_shri(unsigned vece, uint32_t dofs, uint32_t aofs,
                        int64_t shift, uint32_t oprsz, uint32_t maxsz);
 void tcg_gen_gvec_sari(unsigned vece, uint32_t dofs, uint32_t aofs,
                        int64_t shift, uint32_t oprsz, uint32_t maxsz);
+void tcg_gen_gvec_rotli(unsigned vece, uint32_t dofs, uint32_t aofs,
+                        int64_t shift, uint32_t oprsz, uint32_t maxsz);
+void tcg_gen_gvec_rotri(unsigned vece, uint32_t dofs, uint32_t aofs,
+                        int64_t shift, uint32_t oprsz, uint32_t maxsz);
 
 void tcg_gen_gvec_shls(unsigned vece, uint32_t dofs, uint32_t aofs,
                        TCGv_i32 shift, uint32_t oprsz, uint32_t maxsz);
@@ -388,5 +392,7 @@ void tcg_gen_vec_shr8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t);
 void tcg_gen_vec_shr16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t);
 void tcg_gen_vec_sar8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t);
 void tcg_gen_vec_sar16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t);
+void tcg_gen_vec_rotl8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c);
+void tcg_gen_vec_rotl16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c);
 
 #endif
diff --git a/include/tcg/tcg-op.h b/include/tcg/tcg-op.h
index e3399d6..848d91c 100644
--- a/include/tcg/tcg-op.h
+++ b/include/tcg/tcg-op.h
@@ -999,6 +999,8 @@ void tcg_gen_umax_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b);
 void tcg_gen_shli_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i);
 void tcg_gen_shri_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i);
 void tcg_gen_sari_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i);
+void tcg_gen_rotli_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i);
+void tcg_gen_rotri_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i);
 
 void tcg_gen_shls_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s);
 void tcg_gen_shrs_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s);
diff --git a/include/tcg/tcg-opc.h b/include/tcg/tcg-opc.h
index 9288a04..a6ab4a1 100644
--- a/include/tcg/tcg-opc.h
+++ b/include/tcg/tcg-opc.h
@@ -248,6 +248,7 @@ DEF(not_vec, 1, 1, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_not_vec))
 DEF(shli_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_shi_vec))
 DEF(shri_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_shi_vec))
 DEF(sari_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_shi_vec))
+DEF(rotli_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_roti_vec))
 
 DEF(shls_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec))
 DEF(shrs_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec))
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index c48bd76..56bedbd 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -182,6 +182,7 @@ typedef uint64_t TCGRegSet;
 #define TCG_TARGET_HAS_not_vec          0
 #define TCG_TARGET_HAS_andc_vec         0
 #define TCG_TARGET_HAS_orc_vec          0
+#define TCG_TARGET_HAS_roti_vec         0
 #define TCG_TARGET_HAS_shi_vec          0
 #define TCG_TARGET_HAS_shs_vec          0
 #define TCG_TARGET_HAS_shv_vec          0
diff --git a/tcg/README b/tcg/README
index bfa2e4e..1e3e465 100644
--- a/tcg/README
+++ b/tcg/README
@@ -605,10 +605,11 @@ E.g. VECL=1 -> 64 << 1 -> v128, and VECE=2 -> 1 << 2 -> i32.
 
 * shri_vec   v0, v1, i2
 * sari_vec   v0, v1, i2
+* rotli_vec  v0, v1, i2
 * shrs_vec   v0, v1, s2
 * sars_vec   v0, v1, s2
 
-  Similarly for logical and arithmetic right shift.
+  Similarly for logical and arithmetic right shift, and left rotate.
 
 * shlv_vec   v0, v1, v2
 
diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h
index ca214f6..225a597 100644
--- a/tcg/aarch64/tcg-target.h
+++ b/tcg/aarch64/tcg-target.h
@@ -133,6 +133,7 @@ typedef enum {
 #define TCG_TARGET_HAS_not_vec          1
 #define TCG_TARGET_HAS_neg_vec          1
 #define TCG_TARGET_HAS_abs_vec          1
+#define TCG_TARGET_HAS_roti_vec         0
 #define TCG_TARGET_HAS_shi_vec          1
 #define TCG_TARGET_HAS_shs_vec          0
 #define TCG_TARGET_HAS_shv_vec          1
diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h
index bfb3f5f..23aabde 100644
--- a/tcg/i386/tcg-target.h
+++ b/tcg/i386/tcg-target.h
@@ -183,6 +183,7 @@ extern bool have_avx2;
 #define TCG_TARGET_HAS_not_vec          0
 #define TCG_TARGET_HAS_neg_vec          0
 #define TCG_TARGET_HAS_abs_vec          1
+#define TCG_TARGET_HAS_roti_vec         0
 #define TCG_TARGET_HAS_shi_vec          1
 #define TCG_TARGET_HAS_shs_vec          1
 #define TCG_TARGET_HAS_shv_vec          have_avx2
diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h
index 4fa21f0..e57b891 100644
--- a/tcg/ppc/tcg-target.h
+++ b/tcg/ppc/tcg-target.h
@@ -161,6 +161,7 @@ extern bool have_vsx;
 #define TCG_TARGET_HAS_not_vec          1
 #define TCG_TARGET_HAS_neg_vec          have_isa_3_00
 #define TCG_TARGET_HAS_abs_vec          0
+#define TCG_TARGET_HAS_roti_vec         0
 #define TCG_TARGET_HAS_shi_vec          0
 #define TCG_TARGET_HAS_shs_vec          0
 #define TCG_TARGET_HAS_shv_vec          1
diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c
index 049a55e..25300b1 100644
--- a/tcg/tcg-op-gvec.c
+++ b/tcg/tcg-op-gvec.c
@@ -2694,6 +2694,74 @@ void tcg_gen_gvec_sari(unsigned vece, uint32_t dofs, uint32_t aofs,
     }
 }
 
+void tcg_gen_vec_rotl8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c)
+{
+    uint64_t mask = dup_const(MO_8, 0xff << c);
+
+    tcg_gen_shli_i64(d, a, c);
+    tcg_gen_shri_i64(a, a, 8 - c);
+    tcg_gen_andi_i64(d, d, mask);
+    tcg_gen_andi_i64(a, a, ~mask);
+    tcg_gen_or_i64(d, d, a);
+}
+
+void tcg_gen_vec_rotl16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c)
+{
+    uint64_t mask = dup_const(MO_16, 0xffff << c);
+
+    tcg_gen_shli_i64(d, a, c);
+    tcg_gen_shri_i64(a, a, 16 - c);
+    tcg_gen_andi_i64(d, d, mask);
+    tcg_gen_andi_i64(a, a, ~mask);
+    tcg_gen_or_i64(d, d, a);
+}
+
+void tcg_gen_gvec_rotli(unsigned vece, uint32_t dofs, uint32_t aofs,
+                        int64_t shift, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop_list[] = { INDEX_op_rotli_vec, 0 };
+    static const GVecGen2i g[4] = {
+        { .fni8 = tcg_gen_vec_rotl8i_i64,
+          .fniv = tcg_gen_rotli_vec,
+          .fno = gen_helper_gvec_rotl8i,
+          .opt_opc = vecop_list,
+          .vece = MO_8 },
+        { .fni8 = tcg_gen_vec_rotl16i_i64,
+          .fniv = tcg_gen_rotli_vec,
+          .fno = gen_helper_gvec_rotl16i,
+          .opt_opc = vecop_list,
+          .vece = MO_16 },
+        { .fni4 = tcg_gen_rotli_i32,
+          .fniv = tcg_gen_rotli_vec,
+          .fno = gen_helper_gvec_rotl32i,
+          .opt_opc = vecop_list,
+          .vece = MO_32 },
+        { .fni8 = tcg_gen_rotli_i64,
+          .fniv = tcg_gen_rotli_vec,
+          .fno = gen_helper_gvec_rotl64i,
+          .opt_opc = vecop_list,
+          .prefer_i64 = TCG_TARGET_REG_BITS == 64,
+          .vece = MO_64 },
+    };
+
+    tcg_debug_assert(vece <= MO_64);
+    tcg_debug_assert(shift >= 0 && shift < (8 << vece));
+    if (shift == 0) {
+        tcg_gen_gvec_mov(vece, dofs, aofs, oprsz, maxsz);
+    } else {
+        tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, shift, &g[vece]);
+    }
+}
+
+void tcg_gen_gvec_rotri(unsigned vece, uint32_t dofs, uint32_t aofs,
+                        int64_t shift, uint32_t oprsz, uint32_t maxsz)
+{
+    tcg_debug_assert(vece <= MO_64);
+    tcg_debug_assert(shift >= 0 && shift < (8 << vece));
+    tcg_gen_gvec_rotli(vece, dofs, aofs, -shift & ((8 << vece) - 1),
+                       oprsz, maxsz);
+}
+
 /*
  * Specialized generation vector shifts by a non-constant scalar.
  */
diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c
index b6937e8..660ad9b 100644
--- a/tcg/tcg-op-vec.c
+++ b/tcg/tcg-op-vec.c
@@ -545,6 +545,18 @@ void tcg_gen_sari_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i)
     do_shifti(INDEX_op_sari_vec, vece, r, a, i);
 }
 
+void tcg_gen_rotli_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i)
+{
+    do_shifti(INDEX_op_rotli_vec, vece, r, a, i);
+}
+
+void tcg_gen_rotri_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i)
+{
+    int bits = 8 << vece;
+    tcg_debug_assert(i >= 0 && i < bits);
+    do_shifti(INDEX_op_rotli_vec, vece, r, a, -i & (bits - 1));
+}
+
 void tcg_gen_cmp_vec(TCGCond cond, unsigned vece,
                      TCGv_vec r, TCGv_vec a, TCGv_vec b)
 {
diff --git a/tcg/tcg.c b/tcg/tcg.c
index a2268d9..73a3a4f 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -1661,6 +1661,8 @@ bool tcg_op_supported(TCGOpcode op)
     case INDEX_op_shrv_vec:
     case INDEX_op_sarv_vec:
         return have_vec && TCG_TARGET_HAS_shv_vec;
+    case INDEX_op_rotli_vec:
+        return have_vec && TCG_TARGET_HAS_roti_vec;
     case INDEX_op_ssadd_vec:
     case INDEX_op_usadd_vec:
     case INDEX_op_sssub_vec: