diff mbox series

[06/61] tcg: Split out tcg_gen_gvec_3_var

Message ID 20250206195715.2150758-7-richard.henderson@linaro.org
State New
Headers show
Series target/arm: Implement FEAT_SME2 | expand

Commit Message

Richard Henderson Feb. 6, 2025, 7:56 p.m. UTC 
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/tcg/tcg-op-gvec-common.h |   4 ++
 tcg/tcg-op-gvec.c                | 102 +++++++++++++++++++------------
 2 files changed, 68 insertions(+), 38 deletions(-)
diff mbox series

Patch

diff --git a/include/tcg/tcg-op-gvec-common.h b/include/tcg/tcg-op-gvec-common.h
index 877871c101..6e8fccad01 100644
--- a/include/tcg/tcg-op-gvec-common.h
+++ b/include/tcg/tcg-op-gvec-common.h
@@ -236,6 +236,10 @@  void tcg_gen_gvec_2i(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
                      uint32_t maxsz, int64_t c, const GVecGen2i *);
 void tcg_gen_gvec_2s(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
                      uint32_t maxsz, TCGv_i64 c, const GVecGen2s *);
+void tcg_gen_gvec_3_var(TCGv_ptr dbase, uint32_t dofs,
+                        TCGv_ptr abase, uint32_t aofs,
+                        TCGv_ptr bbase, uint32_t bofs,
+                        uint32_t oprsz, uint32_t maxsz, const GVecGen3 *);
 void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                     uint32_t oprsz, uint32_t maxsz, const GVecGen3 *);
 void tcg_gen_gvec_3i(uint32_t dofs, uint32_t aofs, uint32_t bofs,
diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c
index 08019421f2..3e53e43354 100644
--- a/tcg/tcg-op-gvec.c
+++ b/tcg/tcg-op-gvec.c
@@ -178,9 +178,11 @@  void tcg_gen_gvec_2i_ool(uint32_t dofs, uint32_t aofs, TCGv_i64 c,
 }
 
 /* Generate a call to a gvec-style helper with three vector operands.  */
-void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
-                        uint32_t oprsz, uint32_t maxsz, int32_t data,
-                        gen_helper_gvec_3 *fn)
+static void expand_3_ool(TCGv_ptr dbase, uint32_t dofs,
+                         TCGv_ptr abase, uint32_t aofs,
+                         TCGv_ptr bbase, uint32_t bofs,
+                         uint32_t oprsz, uint32_t maxsz,
+                         int32_t data, gen_helper_gvec_3 *fn)
 {
     TCGv_ptr a0, a1, a2;
     TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
@@ -189,9 +191,9 @@  void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     a1 = tcg_temp_ebb_new_ptr();
     a2 = tcg_temp_ebb_new_ptr();
 
-    tcg_gen_addi_ptr(a0, tcg_env, dofs);
-    tcg_gen_addi_ptr(a1, tcg_env, aofs);
-    tcg_gen_addi_ptr(a2, tcg_env, bofs);
+    tcg_gen_addi_ptr(a0, dbase, dofs);
+    tcg_gen_addi_ptr(a1, abase, aofs);
+    tcg_gen_addi_ptr(a2, bbase, bofs);
 
     fn(a0, a1, a2, desc);
 
@@ -200,6 +202,14 @@  void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     tcg_temp_free_ptr(a2);
 }
 
+void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
+                        uint32_t oprsz, uint32_t maxsz, int32_t data,
+                        gen_helper_gvec_3 *fn)
+{
+    expand_3_ool(tcg_env, dofs, tcg_env, aofs, tcg_env, bofs,
+                 oprsz, maxsz, data, fn);
+}
+
 /* Generate a call to a gvec-style helper with four vector operands.  */
 void tcg_gen_gvec_4_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                         uint32_t cofs, uint32_t oprsz, uint32_t maxsz,
@@ -789,8 +799,10 @@  static void expand_2s_i32(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
 }
 
 /* Expand OPSZ bytes worth of three-operand operations using i32 elements.  */
-static void expand_3_i32(uint32_t dofs, uint32_t aofs,
-                         uint32_t bofs, uint32_t oprsz, bool load_dest,
+static void expand_3_i32(TCGv_ptr dbase, uint32_t dofs,
+                         TCGv_ptr abase, uint32_t aofs,
+                         TCGv_ptr bbase, uint32_t bofs,
+                         uint32_t oprsz, bool load_dest,
                          void (*fni)(TCGv_i32, TCGv_i32, TCGv_i32))
 {
     TCGv_i32 t0 = tcg_temp_new_i32();
@@ -799,13 +811,13 @@  static void expand_3_i32(uint32_t dofs, uint32_t aofs,
     uint32_t i;
 
     for (i = 0; i < oprsz; i += 4) {
-        tcg_gen_ld_i32(t0, tcg_env, aofs + i);
-        tcg_gen_ld_i32(t1, tcg_env, bofs + i);
+        tcg_gen_ld_i32(t0, abase, aofs + i);
+        tcg_gen_ld_i32(t1, bbase, bofs + i);
         if (load_dest) {
-            tcg_gen_ld_i32(t2, tcg_env, dofs + i);
+            tcg_gen_ld_i32(t2, dbase, dofs + i);
         }
         fni(t2, t0, t1);
-        tcg_gen_st_i32(t2, tcg_env, dofs + i);
+        tcg_gen_st_i32(t2, dbase, dofs + i);
     }
     tcg_temp_free_i32(t2);
     tcg_temp_free_i32(t1);
@@ -953,8 +965,10 @@  static void expand_2s_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
 }
 
 /* Expand OPSZ bytes worth of three-operand operations using i64 elements.  */
-static void expand_3_i64(uint32_t dofs, uint32_t aofs,
-                         uint32_t bofs, uint32_t oprsz, bool load_dest,
+static void expand_3_i64(TCGv_ptr dbase, uint32_t dofs,
+                         TCGv_ptr abase, uint32_t aofs,
+                         TCGv_ptr bbase, uint32_t bofs,
+                         uint32_t oprsz, bool load_dest,
                          void (*fni)(TCGv_i64, TCGv_i64, TCGv_i64))
 {
     TCGv_i64 t0 = tcg_temp_new_i64();
@@ -963,13 +977,13 @@  static void expand_3_i64(uint32_t dofs, uint32_t aofs,
     uint32_t i;
 
     for (i = 0; i < oprsz; i += 8) {
-        tcg_gen_ld_i64(t0, tcg_env, aofs + i);
-        tcg_gen_ld_i64(t1, tcg_env, bofs + i);
+        tcg_gen_ld_i64(t0, abase, aofs + i);
+        tcg_gen_ld_i64(t1, bbase, bofs + i);
         if (load_dest) {
-            tcg_gen_ld_i64(t2, tcg_env, dofs + i);
+            tcg_gen_ld_i64(t2, dbase, dofs + i);
         }
         fni(t2, t0, t1);
-        tcg_gen_st_i64(t2, tcg_env, dofs + i);
+        tcg_gen_st_i64(t2, dbase, dofs + i);
     }
     tcg_temp_free_i64(t2);
     tcg_temp_free_i64(t1);
@@ -1114,8 +1128,9 @@  static void expand_2s_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
 }
 
 /* Expand OPSZ bytes worth of three-operand operations using host vectors.  */
-static void expand_3_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
-                         uint32_t bofs, uint32_t oprsz,
+static void expand_3_vec(unsigned vece, TCGv_ptr dbase, uint32_t dofs,
+                         TCGv_ptr abase, uint32_t aofs,
+                         TCGv_ptr bbase, uint32_t bofs, uint32_t oprsz,
                          uint32_t tysz, TCGType type, bool load_dest,
                          void (*fni)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec))
 {
@@ -1124,13 +1139,13 @@  static void expand_3_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
         TCGv_vec t1 = tcg_temp_new_vec(type);
         TCGv_vec t2 = tcg_temp_new_vec(type);
 
-        tcg_gen_ld_vec(t0, tcg_env, aofs + i);
-        tcg_gen_ld_vec(t1, tcg_env, bofs + i);
+        tcg_gen_ld_vec(t0, abase, aofs + i);
+        tcg_gen_ld_vec(t1, bbase, bofs + i);
         if (load_dest) {
-            tcg_gen_ld_vec(t2, tcg_env, dofs + i);
+            tcg_gen_ld_vec(t2, dbase, dofs + i);
         }
         fni(vece, t2, t0, t1);
-        tcg_gen_st_vec(t2, tcg_env, dofs + i);
+        tcg_gen_st_vec(t2, dbase, dofs + i);
     }
 }
 
@@ -1432,8 +1447,10 @@  void tcg_gen_gvec_2s(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
 }
 
 /* Expand a vector three-operand operation.  */
-void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
-                    uint32_t oprsz, uint32_t maxsz, const GVecGen3 *g)
+void tcg_gen_gvec_3_var(TCGv_ptr dbase, uint32_t dofs,
+                        TCGv_ptr abase, uint32_t aofs,
+                        TCGv_ptr bbase, uint32_t bofs,
+                        uint32_t oprsz, uint32_t maxsz, const GVecGen3 *g)
 {
     const TCGOpcode *this_list = g->opt_opc ? : vecop_list_empty;
     const TCGOpcode *hold_list = tcg_swap_vecop_list(this_list);
@@ -1441,7 +1458,7 @@  void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     uint32_t some;
 
     check_size_align(oprsz, maxsz, dofs | aofs | bofs);
-    check_overlap_3(tcg_env, dofs, tcg_env, aofs, tcg_env, bofs, maxsz);
+    check_overlap_3(dbase, dofs, abase, aofs, bbase, bofs, maxsz);
 
     type = 0;
     if (g->fniv) {
@@ -1454,8 +1471,8 @@  void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
          * that e.g. size == 80 would be expanded with 2x32 + 1x16.
          */
         some = QEMU_ALIGN_DOWN(oprsz, 32);
-        expand_3_vec(g->vece, dofs, aofs, bofs, some, 32, TCG_TYPE_V256,
-                     g->load_dest, g->fniv);
+        expand_3_vec(g->vece, dbase, dofs, abase, aofs, bbase, bofs,
+                     some, 32, TCG_TYPE_V256, g->load_dest, g->fniv);
         if (some == oprsz) {
             break;
         }
@@ -1466,23 +1483,25 @@  void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
         maxsz -= some;
         /* fallthru */
     case TCG_TYPE_V128:
-        expand_3_vec(g->vece, dofs, aofs, bofs, oprsz, 16, TCG_TYPE_V128,
-                     g->load_dest, g->fniv);
+        expand_3_vec(g->vece, dbase, dofs, abase, aofs, bbase, bofs,
+                     oprsz, 16, TCG_TYPE_V128, g->load_dest, g->fniv);
         break;
     case TCG_TYPE_V64:
-        expand_3_vec(g->vece, dofs, aofs, bofs, oprsz, 8, TCG_TYPE_V64,
-                     g->load_dest, g->fniv);
+        expand_3_vec(g->vece, dbase, dofs, abase, aofs, bbase, bofs,
+                     oprsz, 8, TCG_TYPE_V64, g->load_dest, g->fniv);
         break;
 
     case 0:
         if (g->fni8 && check_size_impl(oprsz, 8)) {
-            expand_3_i64(dofs, aofs, bofs, oprsz, g->load_dest, g->fni8);
+            expand_3_i64(dbase, dofs, abase, aofs, bbase, bofs,
+                         oprsz, g->load_dest, g->fni8);
         } else if (g->fni4 && check_size_impl(oprsz, 4)) {
-            expand_3_i32(dofs, aofs, bofs, oprsz, g->load_dest, g->fni4);
+            expand_3_i32(dbase, dofs, abase, aofs, bbase, bofs,
+                         oprsz, g->load_dest, g->fni4);
         } else {
             assert(g->fno != NULL);
-            tcg_gen_gvec_3_ool(dofs, aofs, bofs, oprsz,
-                               maxsz, g->data, g->fno);
+            expand_3_ool(dbase, dofs, abase, aofs, bbase, bofs,
+                         oprsz, maxsz, g->data, g->fno);
             oprsz = maxsz;
         }
         break;
@@ -1493,10 +1512,17 @@  void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     tcg_swap_vecop_list(hold_list);
 
     if (oprsz < maxsz) {
-        expand_clr(tcg_env, dofs + oprsz, maxsz - oprsz);
+        expand_clr(dbase, dofs + oprsz, maxsz - oprsz);
     }
 }
 
+void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
+                    uint32_t oprsz, uint32_t maxsz, const GVecGen3 *g)
+{
+    tcg_gen_gvec_3_var(tcg_env, dofs, tcg_env, aofs, tcg_env, bofs,
+                       oprsz, maxsz, g);
+}
+
 /* Expand a vector operation with three vectors and an immediate.  */
 void tcg_gen_gvec_3i(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                      uint32_t oprsz, uint32_t maxsz, int64_t c,