@@ -31,6 +31,12 @@
#include "gen_tcg_hvx.h"
#include "genptr.h"
+TCGv gen_read_reg(TCGv result, int num)
+{
+ tcg_gen_mov_tl(result, hex_gpr[num]);
+ return result;
+}
+
TCGv gen_read_preg(TCGv pred, uint8_t num)
{
tcg_gen_mov_tl(pred, hex_pred[num]);
@@ -399,18 +405,19 @@ static inline void gen_store_conditional8(DisasContext *ctx,
tcg_gen_movi_tl(hex_llsc_addr, ~0);
}
-void gen_store32(TCGv vaddr, TCGv src, int width, uint32_t slot)
+void gen_store32(DisasContext *ctx, TCGv vaddr, TCGv src, tcg_target_long width,
+ uint32_t slot)
{
tcg_gen_mov_tl(hex_store_addr[slot], vaddr);
tcg_gen_movi_tl(hex_store_width[slot], width);
tcg_gen_mov_tl(hex_store_val32[slot], src);
+ ctx->store_width[slot] = width;
}
void gen_store1(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx,
uint32_t slot)
{
- gen_store32(vaddr, src, 1, slot);
- ctx->store_width[slot] = 1;
+ gen_store32(ctx, vaddr, src, 1, slot);
}
void gen_store1i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx,
@@ -423,8 +430,7 @@ void gen_store1i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx,
void gen_store2(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx,
uint32_t slot)
{
- gen_store32(vaddr, src, 2, slot);
- ctx->store_width[slot] = 2;
+ gen_store32(ctx, vaddr, src, 2, slot);
}
void gen_store2i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx,
@@ -437,8 +443,7 @@ void gen_store2i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx,
void gen_store4(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx,
uint32_t slot)
{
- gen_store32(vaddr, src, 4, slot);
- ctx->store_width[slot] = 4;
+ gen_store32(ctx, vaddr, src, 4, slot);
}
void gen_store4i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx,
@@ -643,5 +648,153 @@ static void vec_to_qvec(size_t size, intptr_t dstoff, intptr_t srcoff)
tcg_temp_free_i64(mask);
}
+void gen_set_usr_field(int field, TCGv val)
+{
+ /*
+ * Apparently tcg_gen_deposit_i32/64 doesn't OR the input value
+ * with the previously present one, as deposit32/64 in bitops.h
+ * does. We therefore copy the old value to `old_usr` to later
+ * OR with it to replicate this behavior.
+ */
+ TCGv_i32 old_usr = tcg_temp_new_i32();
+ tcg_gen_mov_i32(old_usr, hex_new_value[HEX_REG_USR]);
+ tcg_gen_deposit_tl(hex_new_value[HEX_REG_USR], hex_new_value[HEX_REG_USR],
+ val,
+ reg_field_info[field].offset,
+ reg_field_info[field].width);
+ tcg_gen_or_i32(hex_new_value[HEX_REG_USR],
+ old_usr,
+ hex_new_value[HEX_REG_USR]);
+ tcg_temp_free(old_usr);
+}
+
+void gen_set_usr_fieldi(int field, int x)
+{
+ TCGv val = tcg_constant_tl(x);
+ gen_set_usr_field(field, val);
+}
+
+void gen_write_new_pc(TCGv addr)
+{
+ /* If there are multiple branches in a packet, ignore the second one */
+ TCGv zero = tcg_constant_tl(0);
+ tcg_gen_movcond_tl(TCG_COND_NE, hex_next_PC, hex_branch_taken, zero,
+ hex_next_PC, addr);
+ tcg_gen_movi_tl(hex_branch_taken, 1);
+}
+
+void gen_sat_i32(TCGv dest, TCGv source, int width)
+{
+ TCGv max_val = tcg_constant_tl((1 << (width - 1)) - 1);
+ TCGv min_val = tcg_constant_tl(-(1 << (width - 1)));
+ tcg_gen_smin_tl(dest, source, max_val);
+ tcg_gen_smax_tl(dest, dest, min_val);
+}
+
+void gen_sat_i32_ovfl(TCGv ovfl, TCGv dest, TCGv source, int width)
+{
+ gen_sat_i32(dest, source, width);
+ tcg_gen_setcond_tl(TCG_COND_NE, ovfl, source, dest);
+}
+
+void gen_satu_i32(TCGv dest, TCGv source, int width)
+{
+ TCGv max_val = tcg_constant_tl((1 << width) - 1);
+ TCGv zero = tcg_constant_tl(0);
+ tcg_gen_movcond_tl(TCG_COND_GTU, dest, source, max_val, max_val, source);
+ tcg_gen_movcond_tl(TCG_COND_LT, dest, source, zero, zero, dest);
+}
+
+void gen_satu_i32_ovfl(TCGv ovfl, TCGv dest, TCGv source, int width)
+{
+ gen_satu_i32(dest, source, width);
+ tcg_gen_setcond_tl(TCG_COND_NE, ovfl, source, dest);
+}
+
+void gen_sat_i64(TCGv_i64 dest, TCGv_i64 source, int width)
+{
+ TCGv_i64 max_val = tcg_constant_i64((1LL << (width - 1)) - 1LL);
+ TCGv_i64 min_val = tcg_constant_i64(-(1LL << (width - 1)));
+ tcg_gen_smin_i64(dest, source, max_val);
+ tcg_gen_smax_i64(dest, dest, min_val);
+}
+
+void gen_sat_i64_ovfl(TCGv ovfl, TCGv_i64 dest, TCGv_i64 source, int width)
+{
+ TCGv_i64 ovfl_64;
+ gen_sat_i64(dest, source, width);
+ ovfl_64 = tcg_temp_new_i64();
+ tcg_gen_setcond_i64(TCG_COND_NE, ovfl_64, dest, source);
+ tcg_gen_trunc_i64_tl(ovfl, ovfl_64);
+ tcg_temp_free_i64(ovfl_64);
+}
+
+void gen_satu_i64(TCGv_i64 dest, TCGv_i64 source, int width)
+{
+ TCGv_i64 max_val = tcg_constant_i64((1LL << width) - 1LL);
+ TCGv_i64 zero = tcg_constant_i64(0);
+ tcg_gen_movcond_i64(TCG_COND_GTU, dest, source, max_val, max_val, source);
+ tcg_gen_movcond_i64(TCG_COND_LT, dest, source, zero, zero, dest);
+}
+
+void gen_satu_i64_ovfl(TCGv ovfl, TCGv_i64 dest, TCGv_i64 source, int width)
+{
+ TCGv_i64 ovfl_64;
+ gen_satu_i64(dest, source, width);
+ ovfl_64 = tcg_temp_new_i64();
+ tcg_gen_setcond_i64(TCG_COND_NE, ovfl_64, dest, source);
+ tcg_gen_trunc_i64_tl(ovfl, ovfl_64);
+ tcg_temp_free_i64(ovfl_64);
+}
+
+/* Implements the fADDSAT64 macro in TCG */
+void gen_add_sat_i64(TCGv_i64 ret, TCGv_i64 a, TCGv_i64 b)
+{
+ TCGv_i64 sum = tcg_temp_local_new_i64();
+ TCGv_i64 xor = tcg_temp_new_i64();
+ TCGv_i64 cond1 = tcg_temp_new_i64();
+ TCGv_i64 cond2 = tcg_temp_local_new_i64();
+ TCGv_i64 cond3 = tcg_temp_new_i64();
+ TCGv_i64 mask = tcg_constant_i64(0x8000000000000000ULL);
+ TCGv_i64 max_pos = tcg_constant_i64(0x7FFFFFFFFFFFFFFFLL);
+ TCGv_i64 max_neg = tcg_constant_i64(0x8000000000000000LL);
+ TCGv_i64 zero = tcg_constant_i64(0);
+ TCGLabel *no_ovfl_label = gen_new_label();
+ TCGLabel *ovfl_label = gen_new_label();
+ TCGLabel *ret_label = gen_new_label();
+
+ tcg_gen_add_i64(sum, a, b);
+ tcg_gen_xor_i64(xor, a, b);
+
+ /* if (xor & mask) */
+ tcg_gen_and_i64(cond1, xor, mask);
+ tcg_temp_free_i64(xor);
+ tcg_gen_brcondi_i64(TCG_COND_NE, cond1, 0, no_ovfl_label);
+ tcg_temp_free_i64(cond1);
+
+ /* else if ((a ^ sum) & mask) */
+ tcg_gen_xor_i64(cond2, a, sum);
+ tcg_gen_and_i64(cond2, cond2, mask);
+ tcg_gen_brcondi_i64(TCG_COND_NE, cond2, 0, ovfl_label);
+ tcg_temp_free_i64(cond2);
+ /* fallthrough to no_ovfl_label branch */
+
+ /* if branch */
+ gen_set_label(no_ovfl_label);
+ tcg_gen_mov_i64(ret, sum);
+ tcg_gen_br(ret_label);
+
+ /* else if branch */
+ gen_set_label(ovfl_label);
+ tcg_gen_and_i64(cond3, sum, mask);
+ tcg_temp_free_i64(mask);
+ tcg_temp_free_i64(sum);
+ tcg_gen_movcond_i64(TCG_COND_NE, ret, cond3, zero, max_pos, max_neg);
+ tcg_temp_free_i64(cond3);
+ SET_USR_FIELD(USR_OVF, 1);
+
+ gen_set_label(ret_label);
+}
+
#include "tcg_funcs_generated.c.inc"
#include "tcg_func_table_generated.c.inc"
@@ -24,7 +24,8 @@
extern const SemanticInsn opcode_genptr[];
-void gen_store32(TCGv vaddr, TCGv src, int width, uint32_t slot);
+void gen_store32(DisasContext *ctx, TCGv vaddr, TCGv src, tcg_target_long width,
+ uint32_t slot);
void gen_store1(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx,
uint32_t slot);
void gen_store2(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx,
@@ -44,6 +45,18 @@ void gen_store8i(TCGv_env cpu_env, TCGv vaddr, int64_t src, DisasContext *ctx,
TCGv gen_read_preg(TCGv pred, uint8_t num);
void gen_log_reg_write(int rnum, TCGv val);
void gen_log_pred_write(DisasContext *ctx, int pnum, TCGv val);
+void gen_write_new_pc(TCGv addr);
+void gen_set_usr_field(int field, TCGv val);
+void gen_set_usr_fieldi(int field, int x);
+void gen_sat_i32(TCGv dest, TCGv source, int width);
+void gen_sat_i32_ovfl(TCGv ovfl, TCGv dest, TCGv source, int width);
+void gen_satu_i32(TCGv dest, TCGv source, int width);
+void gen_satu_i32_ovfl(TCGv ovfl, TCGv dest, TCGv source, int width);
+void gen_sat_i64(TCGv_i64 dest, TCGv_i64 source, int width);
+void gen_sat_i64_ovfl(TCGv ovfl, TCGv_i64 dest, TCGv_i64 source, int width);
+void gen_satu_i64(TCGv_i64 dest, TCGv_i64 source, int width);
+void gen_satu_i64_ovfl(TCGv ovfl, TCGv_i64 dest, TCGv_i64 source, int width);
+void gen_add_sat_i64(TCGv_i64 ret, TCGv_i64 a, TCGv_i64 b);
TCGv gen_8bitsof(TCGv result, TCGv value);
void gen_set_byte_i64(int N, TCGv_i64 result, TCGv src);
TCGv gen_get_byte(TCGv result, int N, TCGv src, bool sign);
@@ -51,5 +64,6 @@ TCGv gen_get_byte_i64(TCGv result, int N, TCGv_i64 src, bool sign);
TCGv gen_get_half(TCGv result, int N, TCGv src, bool sign);
void gen_set_half(int N, TCGv result, TCGv src);
void gen_set_half_i64(int N, TCGv_i64 result, TCGv src);
+TCGv gen_read_reg(TCGv result, int num);
#endif
@@ -180,7 +180,16 @@
#define MEM_STORE8(VA, DATA, SLOT) log_store64(env, VA, DATA, 8, SLOT)
#endif
+#ifdef QEMU_GENERATE
+static inline void gen_cancel(uint32_t slot)
+{
+ tcg_gen_ori_tl(hex_slot_cancelled, hex_slot_cancelled, 1 << slot);
+}
+
+#define CANCEL gen_cancel(slot);
+#else
#define CANCEL cancel_slot(env, slot)
+#endif
#define LOAD_CANCEL(EA) do { CANCEL; } while (0)