tcg: Implement gvec support for rotate by vector

No host backend support yet, but the interfaces for rotlv
and rotrv are in place.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
v3: Drop the generic expansion from rot to shift; we can do better
    for each backend, and then this code becomes unused.

accel/tcg/tcg-runtime-gvec.c

@@ -908,6 +908,102 @@ void HELPER(gvec_sar64v)(void *d, void *a, void *b, uint32_t desc)
     clear_high(d, oprsz, desc);
 }
 
+void HELPER(gvec_rotl8v)(void *d, void *a, void *b, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint8_t)) {
+        uint8_t sh = *(uint8_t *)(b + i) & 7;
+        *(uint8_t *)(d + i) = rol8(*(uint8_t *)(a + i), sh);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotl16v)(void *d, void *a, void *b, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint16_t)) {
+        uint8_t sh = *(uint16_t *)(b + i) & 15;
+        *(uint16_t *)(d + i) = rol16(*(uint16_t *)(a + i), sh);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotl32v)(void *d, void *a, void *b, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint32_t)) {
+        uint8_t sh = *(uint32_t *)(b + i) & 31;
+        *(uint32_t *)(d + i) = rol32(*(uint32_t *)(a + i), sh);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotl64v)(void *d, void *a, void *b, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint64_t)) {
+        uint8_t sh = *(uint64_t *)(b + i) & 63;
+        *(uint64_t *)(d + i) = rol64(*(uint64_t *)(a + i), sh);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotr8v)(void *d, void *a, void *b, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint8_t)) {
+        uint8_t sh = *(uint8_t *)(b + i) & 7;
+        *(uint8_t *)(d + i) = ror8(*(uint8_t *)(a + i), sh);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotr16v)(void *d, void *a, void *b, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint16_t)) {
+        uint8_t sh = *(uint16_t *)(b + i) & 15;
+        *(uint16_t *)(d + i) = ror16(*(uint16_t *)(a + i), sh);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotr32v)(void *d, void *a, void *b, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint32_t)) {
+        uint8_t sh = *(uint32_t *)(b + i) & 31;
+        *(uint32_t *)(d + i) = ror32(*(uint32_t *)(a + i), sh);
+    }
+    clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_rotr64v)(void *d, void *a, void *b, uint32_t desc)
+{
+    intptr_t oprsz = simd_oprsz(desc);
+    intptr_t i;
+
+    for (i = 0; i < oprsz; i += sizeof(uint64_t)) {
+        uint8_t sh = *(uint64_t *)(b + i) & 63;
+        *(uint64_t *)(d + i) = ror64(*(uint64_t *)(a + i), sh);
+    }
+    clear_high(d, oprsz, desc);
+}
+
 #define DO_CMP1(NAME, TYPE, OP)                                            \
 void HELPER(NAME)(void *d, void *a, void *b, uint32_t desc)                \
 {                                                                          \

accel/tcg/tcg-runtime.h

@@ -279,6 +279,16 @@ DEF_HELPER_FLAGS_4(gvec_sar16v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(gvec_sar32v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(gvec_sar64v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 
+DEF_HELPER_FLAGS_4(gvec_rotl8v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_rotl16v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_rotl32v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_rotl64v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_4(gvec_rotr8v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_rotr16v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_rotr32v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_rotr64v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+
 DEF_HELPER_FLAGS_4(gvec_eq8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(gvec_eq16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(gvec_eq32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)

include/tcg/tcg-op-gvec.h

@@ -356,6 +356,10 @@ void tcg_gen_gvec_shrv(unsigned vece, uint32_t dofs, uint32_t aofs,
                        uint32_t bofs, uint32_t oprsz, uint32_t maxsz);
 void tcg_gen_gvec_sarv(unsigned vece, uint32_t dofs, uint32_t aofs,
                        uint32_t bofs, uint32_t oprsz, uint32_t maxsz);
+void tcg_gen_gvec_rotlv(unsigned vece, uint32_t dofs, uint32_t aofs,
+                        uint32_t bofs, uint32_t oprsz, uint32_t maxsz);
+void tcg_gen_gvec_rotrv(unsigned vece, uint32_t dofs, uint32_t aofs,
+                        uint32_t bofs, uint32_t oprsz, uint32_t maxsz);
 
 void tcg_gen_gvec_cmp(TCGCond cond, unsigned vece, uint32_t dofs,
                       uint32_t aofs, uint32_t bofs,

include/tcg/tcg-op.h

@@ -1009,6 +1009,8 @@ void tcg_gen_sars_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s);
 void tcg_gen_shlv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s);
 void tcg_gen_shrv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s);
 void tcg_gen_sarv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s);
+void tcg_gen_rotlv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s);
+void tcg_gen_rotrv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s);
 
 void tcg_gen_cmp_vec(TCGCond cond, unsigned vece, TCGv_vec r,
                      TCGv_vec a, TCGv_vec b);

include/tcg/tcg-opc.h

@@ -257,6 +257,8 @@ DEF(sars_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec))
 DEF(shlv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec))
 DEF(shrv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec))
 DEF(sarv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec))
+DEF(rotlv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_rotv_vec))
+DEF(rotrv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_rotv_vec))
 
 DEF(cmp_vec, 1, 2, 1, IMPLVEC)
 

include/tcg/tcg.h

@@ -183,6 +183,7 @@ typedef uint64_t TCGRegSet;
 #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_rotv_vec         0
 #define TCG_TARGET_HAS_shi_vec          0
 #define TCG_TARGET_HAS_shs_vec          0
 #define TCG_TARGET_HAS_shv_vec          0

tcg/README

@@ -621,8 +621,10 @@ E.g. VECL=1 -> 64 << 1 -> v128, and VECE=2 -> 1 << 2 -> i32.
 
 * shrv_vec   v0, v1, v2
 * sarv_vec   v0, v1, v2
+* rotlv_vec  v0, v1, v2
+* rotrv_vec  v0, v1, v2
 
-  Similarly for logical and arithmetic right shift.
+  Similarly for logical and arithmetic right shift, and rotates.
 
 * cmp_vec  v0, v1, v2, cond
 

tcg/aarch64/tcg-target.h

@@ -134,6 +134,7 @@ typedef enum {
 #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_rotv_vec         0
 #define TCG_TARGET_HAS_shi_vec          1
 #define TCG_TARGET_HAS_shs_vec          0
 #define TCG_TARGET_HAS_shv_vec          1

tcg/i386/tcg-target.h

@@ -184,6 +184,7 @@ extern bool have_avx2;
 #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_rotv_vec         0
 #define TCG_TARGET_HAS_shi_vec          1
 #define TCG_TARGET_HAS_shs_vec          1
 #define TCG_TARGET_HAS_shv_vec          have_avx2

tcg/ppc/tcg-target.h

@@ -162,6 +162,7 @@ extern bool have_vsx;
 #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_rotv_vec         0
 #define TCG_TARGET_HAS_shi_vec          0
 #define TCG_TARGET_HAS_shs_vec          0
 #define TCG_TARGET_HAS_shv_vec          1

tcg/tcg-op-gvec.c

@@ -3171,6 +3171,128 @@ void tcg_gen_gvec_sarv(unsigned vece, uint32_t dofs, uint32_t aofs,
     tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]);
 }
 
+/*
+ * Similarly for rotates.
+ */
+
+static void tcg_gen_rotlv_mod_vec(unsigned vece, TCGv_vec d,
+                                  TCGv_vec a, TCGv_vec b)
+{
+    TCGv_vec t = tcg_temp_new_vec_matching(d);
+
+    tcg_gen_dupi_vec(vece, t, (8 << vece) - 1);
+    tcg_gen_and_vec(vece, t, t, b);
+    tcg_gen_rotlv_vec(vece, d, a, t);
+    tcg_temp_free_vec(t);
+}
+
+static void tcg_gen_rotl_mod_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b)
+{
+    TCGv_i32 t = tcg_temp_new_i32();
+
+    tcg_gen_andi_i32(t, b, 31);
+    tcg_gen_rotl_i32(d, a, t);
+    tcg_temp_free_i32(t);
+}
+
+static void tcg_gen_rotl_mod_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
+{
+    TCGv_i64 t = tcg_temp_new_i64();
+
+    tcg_gen_andi_i64(t, b, 63);
+    tcg_gen_rotl_i64(d, a, t);
+    tcg_temp_free_i64(t);
+}
+
+void tcg_gen_gvec_rotlv(unsigned vece, uint32_t dofs, uint32_t aofs,
+                        uint32_t bofs, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop_list[] = { INDEX_op_rotlv_vec, 0 };
+    static const GVecGen3 g[4] = {
+        { .fniv = tcg_gen_rotlv_mod_vec,
+          .fno = gen_helper_gvec_rotl8v,
+          .opt_opc = vecop_list,
+          .vece = MO_8 },
+        { .fniv = tcg_gen_rotlv_mod_vec,
+          .fno = gen_helper_gvec_rotl16v,
+          .opt_opc = vecop_list,
+          .vece = MO_16 },
+        { .fni4 = tcg_gen_rotl_mod_i32,
+          .fniv = tcg_gen_rotlv_mod_vec,
+          .fno = gen_helper_gvec_rotl32v,
+          .opt_opc = vecop_list,
+          .vece = MO_32 },
+        { .fni8 = tcg_gen_rotl_mod_i64,
+          .fniv = tcg_gen_rotlv_mod_vec,
+          .fno = gen_helper_gvec_rotl64v,
+          .opt_opc = vecop_list,
+          .prefer_i64 = TCG_TARGET_REG_BITS == 64,
+          .vece = MO_64 },
+    };
+
+    tcg_debug_assert(vece <= MO_64);
+    tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]);
+}
+
+static void tcg_gen_rotrv_mod_vec(unsigned vece, TCGv_vec d,
+                                  TCGv_vec a, TCGv_vec b)
+{
+    TCGv_vec t = tcg_temp_new_vec_matching(d);
+
+    tcg_gen_dupi_vec(vece, t, (8 << vece) - 1);
+    tcg_gen_and_vec(vece, t, t, b);
+    tcg_gen_rotrv_vec(vece, d, a, t);
+    tcg_temp_free_vec(t);
+}
+
+static void tcg_gen_rotr_mod_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b)
+{
+    TCGv_i32 t = tcg_temp_new_i32();
+
+    tcg_gen_andi_i32(t, b, 31);
+    tcg_gen_rotr_i32(d, a, t);
+    tcg_temp_free_i32(t);
+}
+
+static void tcg_gen_rotr_mod_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
+{
+    TCGv_i64 t = tcg_temp_new_i64();
+
+    tcg_gen_andi_i64(t, b, 63);
+    tcg_gen_rotr_i64(d, a, t);
+    tcg_temp_free_i64(t);
+}
+
+void tcg_gen_gvec_rotrv(unsigned vece, uint32_t dofs, uint32_t aofs,
+                        uint32_t bofs, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop_list[] = { INDEX_op_rotrv_vec, 0 };
+    static const GVecGen3 g[4] = {
+        { .fniv = tcg_gen_rotrv_mod_vec,
+          .fno = gen_helper_gvec_rotr8v,
+          .opt_opc = vecop_list,
+          .vece = MO_8 },
+        { .fniv = tcg_gen_rotrv_mod_vec,
+          .fno = gen_helper_gvec_rotr16v,
+          .opt_opc = vecop_list,
+          .vece = MO_16 },
+        { .fni4 = tcg_gen_rotr_mod_i32,
+          .fniv = tcg_gen_rotrv_mod_vec,
+          .fno = gen_helper_gvec_rotr32v,
+          .opt_opc = vecop_list,
+          .vece = MO_32 },
+        { .fni8 = tcg_gen_rotr_mod_i64,
+          .fniv = tcg_gen_rotrv_mod_vec,
+          .fno = gen_helper_gvec_rotr64v,
+          .opt_opc = vecop_list,
+          .prefer_i64 = TCG_TARGET_REG_BITS == 64,
+          .vece = MO_64 },
+    };
+
+    tcg_debug_assert(vece <= MO_64);
+    tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]);
+}
+
 /* Expand OPSZ bytes worth of three-operand operations using i32 elements.  */
 static void expand_cmp_i32(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                            uint32_t oprsz, TCGCond cond)

tcg/tcg-op-vec.c

@@ -696,6 +696,16 @@ void tcg_gen_sarv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b)
     do_op3_nofail(vece, r, a, b, INDEX_op_sarv_vec);
 }
 
+void tcg_gen_rotlv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b)
+{
+    do_op3_nofail(vece, r, a, b, INDEX_op_rotlv_vec);
+}
+
+void tcg_gen_rotrv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b)
+{
+    do_op3_nofail(vece, r, a, b, INDEX_op_rotrv_vec);
+}
+
 static void do_shifts(unsigned vece, TCGv_vec r, TCGv_vec a,
                       TCGv_i32 s, TCGOpcode opc_s, TCGOpcode opc_v)
 {

tcg/tcg.c

@@ -1663,6 +1663,9 @@ bool tcg_op_supported(TCGOpcode op)
         return have_vec && TCG_TARGET_HAS_shv_vec;
     case INDEX_op_rotli_vec:
         return have_vec && TCG_TARGET_HAS_roti_vec;
+    case INDEX_op_rotlv_vec:
+    case INDEX_op_rotrv_vec:
+        return have_vec && TCG_TARGET_HAS_rotv_vec;
     case INDEX_op_ssadd_vec:
     case INDEX_op_usadd_vec:
     case INDEX_op_sssub_vec: