[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[PULL 08/11] Hexagon (target/hexagon) Add overrides for compound compare
|
From: |
Taylor Simpson |
|
Subject: |
[PULL 08/11] Hexagon (target/hexagon) Add overrides for compound compare and jump |
|
Date: |
Thu, 10 Nov 2022 16:52:11 -0800 |
Acked-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Message-Id: <20221108162906.3166-9-tsimpson@quicinc.com>
---
target/hexagon/gen_tcg.h | 177 +++++++++++++++++++++++++++++++++++++++
target/hexagon/genptr.c | 90 ++++++++++++++++++++
2 files changed, 267 insertions(+)
diff --git a/target/hexagon/gen_tcg.h b/target/hexagon/gen_tcg.h
index 1bdc787a02..506b454e4e 100644
--- a/target/hexagon/gen_tcg.h
+++ b/target/hexagon/gen_tcg.h
@@ -620,6 +620,183 @@
#define fGEN_TCG_J2_callf(SHORTCODE) \
gen_cond_call(ctx, PuV, TCG_COND_NE, riV)
+/*
+ * Compound compare and jump instructions
+ * Here is a primer to understand the tag names
+ *
+ * Comparison
+ * cmpeqi compare equal to an immediate
+ * cmpgti compare greater than an immediate
+ * cmpgtiu compare greater than an unsigned immediate
+ * cmpeqn1 compare equal to negative 1
+ * cmpgtn1 compare greater than negative 1
+ * cmpeq compare equal (two registers)
+ * cmpgtu compare greater than unsigned (two registers)
+ * tstbit0 test bit zero
+ *
+ * Condition
+ * tp0 p0 is true p0 = cmp.eq(r0,#5); if (p0.new) jump:nt address
+ * fp0 p0 is false p0 = cmp.eq(r0,#5); if (!p0.new) jump:nt
address
+ * tp1 p1 is true p1 = cmp.eq(r0,#5); if (p1.new) jump:nt address
+ * fp1 p1 is false p1 = cmp.eq(r0,#5); if (!p1.new) jump:nt
address
+ *
+ * Prediction (not modelled in qemu)
+ * _nt not taken
+ * _t taken
+ */
+#define fGEN_TCG_J4_cmpeq_tp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 0, TCG_COND_EQ, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpeq_tp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 0, TCG_COND_EQ, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpeq_fp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 0, TCG_COND_EQ, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpeq_fp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 0, TCG_COND_EQ, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpeq_tp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 1, TCG_COND_EQ, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpeq_tp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 1, TCG_COND_EQ, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpeq_fp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 1, TCG_COND_EQ, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpeq_fp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 1, TCG_COND_EQ, RsV, RtV, riV)
+
+#define fGEN_TCG_J4_cmpgt_tp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 0, TCG_COND_GT, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgt_tp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 0, TCG_COND_GT, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgt_fp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 0, TCG_COND_GT, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgt_fp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 0, TCG_COND_GT, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgt_tp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 1, TCG_COND_GT, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgt_tp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 1, TCG_COND_GT, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgt_fp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 1, TCG_COND_GT, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgt_fp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 1, TCG_COND_GT, RsV, RtV, riV)
+
+#define fGEN_TCG_J4_cmpgtu_tp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 0, TCG_COND_GTU, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgtu_tp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 0, TCG_COND_GTU, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgtu_fp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 0, TCG_COND_GTU, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgtu_fp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 0, TCG_COND_GTU, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgtu_tp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 1, TCG_COND_GTU, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgtu_tp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_t(ctx, 1, TCG_COND_GTU, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgtu_fp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 1, TCG_COND_GTU, RsV, RtV, riV)
+#define fGEN_TCG_J4_cmpgtu_fp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_jmp_f(ctx, 1, TCG_COND_GTU, RsV, RtV, riV)
+
+#define fGEN_TCG_J4_cmpeqi_tp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 0, TCG_COND_EQ, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpeqi_tp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 0, TCG_COND_EQ, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpeqi_fp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 0, TCG_COND_EQ, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpeqi_fp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 0, TCG_COND_EQ, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpeqi_tp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 1, TCG_COND_EQ, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpeqi_tp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 1, TCG_COND_EQ, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpeqi_fp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 1, TCG_COND_EQ, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpeqi_fp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 1, TCG_COND_EQ, RsV, UiV, riV)
+
+#define fGEN_TCG_J4_cmpgti_tp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 0, TCG_COND_GT, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgti_tp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 0, TCG_COND_GT, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgti_fp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 0, TCG_COND_GT, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgti_fp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 0, TCG_COND_GT, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgti_tp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 1, TCG_COND_GT, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgti_tp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 1, TCG_COND_GT, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgti_fp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 1, TCG_COND_GT, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgti_fp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 1, TCG_COND_GT, RsV, UiV, riV)
+
+#define fGEN_TCG_J4_cmpgtui_tp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 0, TCG_COND_GTU, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgtui_tp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 0, TCG_COND_GTU, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgtui_fp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 0, TCG_COND_GTU, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgtui_fp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 0, TCG_COND_GTU, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgtui_tp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 1, TCG_COND_GTU, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgtui_tp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_t(ctx, 1, TCG_COND_GTU, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgtui_fp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 1, TCG_COND_GTU, RsV, UiV, riV)
+#define fGEN_TCG_J4_cmpgtui_fp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmpi_jmp_f(ctx, 1, TCG_COND_GTU, RsV, UiV, riV)
+
+#define fGEN_TCG_J4_cmpeqn1_tp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_t(ctx, 0, TCG_COND_EQ, RsV, riV)
+#define fGEN_TCG_J4_cmpeqn1_tp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_t(ctx, 0, TCG_COND_EQ, RsV, riV)
+#define fGEN_TCG_J4_cmpeqn1_fp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_f(ctx, 0, TCG_COND_EQ, RsV, riV)
+#define fGEN_TCG_J4_cmpeqn1_fp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_f(ctx, 0, TCG_COND_EQ, RsV, riV)
+#define fGEN_TCG_J4_cmpeqn1_tp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_t(ctx, 1, TCG_COND_EQ, RsV, riV)
+#define fGEN_TCG_J4_cmpeqn1_tp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_t(ctx, 1, TCG_COND_EQ, RsV, riV)
+#define fGEN_TCG_J4_cmpeqn1_fp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_f(ctx, 1, TCG_COND_EQ, RsV, riV)
+#define fGEN_TCG_J4_cmpeqn1_fp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_f(ctx, 1, TCG_COND_EQ, RsV, riV)
+
+#define fGEN_TCG_J4_cmpgtn1_tp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_t(ctx, 0, TCG_COND_GT, RsV, riV)
+#define fGEN_TCG_J4_cmpgtn1_tp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_t(ctx, 0, TCG_COND_GT, RsV, riV)
+#define fGEN_TCG_J4_cmpgtn1_fp0_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_f(ctx, 0, TCG_COND_GT, RsV, riV)
+#define fGEN_TCG_J4_cmpgtn1_fp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_f(ctx, 0, TCG_COND_GT, RsV, riV)
+#define fGEN_TCG_J4_cmpgtn1_tp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_t(ctx, 1, TCG_COND_GT, RsV, riV)
+#define fGEN_TCG_J4_cmpgtn1_tp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_t(ctx, 1, TCG_COND_GT, RsV, riV)
+#define fGEN_TCG_J4_cmpgtn1_fp1_jump_t(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_f(ctx, 1, TCG_COND_GT, RsV, riV)
+#define fGEN_TCG_J4_cmpgtn1_fp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_cmp_n1_jmp_f(ctx, 1, TCG_COND_GT, RsV, riV)
+
+#define fGEN_TCG_J4_tstbit0_tp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_tstbit0_jmp(ctx, 0, RsV, TCG_COND_EQ, riV)
+#define fGEN_TCG_J4_tstbit0_tp0_jump_t(SHORTCODE) \
+ gen_cmpnd_tstbit0_jmp(ctx, 0, RsV, TCG_COND_EQ, riV)
+#define fGEN_TCG_J4_tstbit0_fp0_jump_nt(SHORTCODE) \
+ gen_cmpnd_tstbit0_jmp(ctx, 0, RsV, TCG_COND_NE, riV)
+#define fGEN_TCG_J4_tstbit0_fp0_jump_t(SHORTCODE) \
+ gen_cmpnd_tstbit0_jmp(ctx, 0, RsV, TCG_COND_NE, riV)
+#define fGEN_TCG_J4_tstbit0_tp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_tstbit0_jmp(ctx, 1, RsV, TCG_COND_EQ, riV)
+#define fGEN_TCG_J4_tstbit0_tp1_jump_t(SHORTCODE) \
+ gen_cmpnd_tstbit0_jmp(ctx, 1, RsV, TCG_COND_EQ, riV)
+#define fGEN_TCG_J4_tstbit0_fp1_jump_nt(SHORTCODE) \
+ gen_cmpnd_tstbit0_jmp(ctx, 1, RsV, TCG_COND_NE, riV)
+#define fGEN_TCG_J4_tstbit0_fp1_jump_t(SHORTCODE) \
+ gen_cmpnd_tstbit0_jmp(ctx, 1, RsV, TCG_COND_NE, riV)
+
#define fGEN_TCG_J2_pause(SHORTCODE) \
do { \
uiV = uiV; \
diff --git a/target/hexagon/genptr.c b/target/hexagon/genptr.c
index 580c403879..c003800247 100644
--- a/target/hexagon/genptr.c
+++ b/target/hexagon/genptr.c
@@ -487,6 +487,96 @@ static void gen_write_new_pc_pcrel(DisasContext *ctx, int
pc_off,
gen_write_new_pc_addr(ctx, tcg_constant_tl(dest), cond, pred);
}
+static void gen_compare(TCGCond cond, TCGv res, TCGv arg1, TCGv arg2)
+{
+ TCGv one = tcg_constant_tl(0xff);
+ TCGv zero = tcg_constant_tl(0);
+
+ tcg_gen_movcond_tl(cond, res, arg1, arg2, one, zero);
+}
+
+static void gen_cond_jump(DisasContext *ctx, TCGCond cond, TCGv pred,
+ int pc_off)
+{
+ gen_write_new_pc_pcrel(ctx, pc_off, cond, pred);
+}
+
+static void gen_cmpnd_cmp_jmp(DisasContext *ctx,
+ int pnum, TCGCond cond1, TCGv arg1, TCGv arg2,
+ TCGCond cond2, int pc_off)
+{
+ if (ctx->insn->part1) {
+ TCGv pred = tcg_temp_new();
+ gen_compare(cond1, pred, arg1, arg2);
+ gen_log_pred_write(ctx, pnum, pred);
+ tcg_temp_free(pred);
+ } else {
+ TCGv pred = tcg_temp_new();
+ tcg_gen_mov_tl(pred, hex_new_pred_value[pnum]);
+ gen_cond_jump(ctx, cond2, pred, pc_off);
+ tcg_temp_free(pred);
+ }
+}
+
+static void gen_cmpnd_cmp_jmp_t(DisasContext *ctx,
+ int pnum, TCGCond cond, TCGv arg1, TCGv arg2,
+ int pc_off)
+{
+ gen_cmpnd_cmp_jmp(ctx, pnum, cond, arg1, arg2, TCG_COND_EQ, pc_off);
+}
+
+static void gen_cmpnd_cmp_jmp_f(DisasContext *ctx,
+ int pnum, TCGCond cond, TCGv arg1, TCGv arg2,
+ int pc_off)
+{
+ gen_cmpnd_cmp_jmp(ctx, pnum, cond, arg1, arg2, TCG_COND_NE, pc_off);
+}
+
+static void gen_cmpnd_cmpi_jmp_t(DisasContext *ctx,
+ int pnum, TCGCond cond, TCGv arg1, int arg2,
+ int pc_off)
+{
+ TCGv tmp = tcg_constant_tl(arg2);
+ gen_cmpnd_cmp_jmp(ctx, pnum, cond, arg1, tmp, TCG_COND_EQ, pc_off);
+}
+
+static void gen_cmpnd_cmpi_jmp_f(DisasContext *ctx,
+ int pnum, TCGCond cond, TCGv arg1, int arg2,
+ int pc_off)
+{
+ TCGv tmp = tcg_constant_tl(arg2);
+ gen_cmpnd_cmp_jmp(ctx, pnum, cond, arg1, tmp, TCG_COND_NE, pc_off);
+}
+
+static void gen_cmpnd_cmp_n1_jmp_t(DisasContext *ctx, int pnum, TCGCond cond,
+ TCGv arg, int pc_off)
+{
+ gen_cmpnd_cmpi_jmp_t(ctx, pnum, cond, arg, -1, pc_off);
+}
+
+static void gen_cmpnd_cmp_n1_jmp_f(DisasContext *ctx, int pnum, TCGCond cond,
+ TCGv arg, int pc_off)
+{
+ gen_cmpnd_cmpi_jmp_f(ctx, pnum, cond, arg, -1, pc_off);
+}
+
+static void gen_cmpnd_tstbit0_jmp(DisasContext *ctx,
+ int pnum, TCGv arg, TCGCond cond, int pc_off)
+{
+ if (ctx->insn->part1) {
+ TCGv pred = tcg_temp_new();
+ tcg_gen_andi_tl(pred, arg, 1);
+ gen_8bitsof(pred, pred);
+ gen_log_pred_write(ctx, pnum, pred);
+ tcg_temp_free(pred);
+ } else {
+ TCGv pred = tcg_temp_new();
+ tcg_gen_mov_tl(pred, hex_new_pred_value[pnum]);
+ gen_cond_jump(ctx, cond, pred, pc_off);
+ tcg_temp_free(pred);
+ }
+}
+
static void gen_call(DisasContext *ctx, int pc_off)
{
TCGv next_PC =
--
2.17.1
- [PULL 00/11] Hexagon bug fixes and performance improvement, Taylor Simpson, 2022/11/10
- [PULL 03/11] Hexagon (target/hexagon) Add overrides for S2_asr_r_r_sat/S2_asl_r_r_sat, Taylor Simpson, 2022/11/10
- [PULL 02/11] Hexagon (target/hexagon) Fix predicated assignment to .tmp and .cur, Taylor Simpson, 2022/11/10
- [PULL 05/11] Hexagon (target/hexagon) Remove PC from the runtime state, Taylor Simpson, 2022/11/10
- [PULL 08/11] Hexagon (target/hexagon) Add overrides for compound compare and jump,
Taylor Simpson <=
- [PULL 01/11] Hexagon (target/hexagon) Add pkt and insn to DisasContext, Taylor Simpson, 2022/11/10
- [PULL 09/11] Hexagon (target/hexagon) Add overrides for various forms of jump, Taylor Simpson, 2022/11/10
- [PULL 11/11] Hexagon (target/hexagon) Use direct block chaining for tight loops, Taylor Simpson, 2022/11/10
- [PULL 07/11] Hexagon (target/hexagon) Add overrides for direct call instructions, Taylor Simpson, 2022/11/10
- [PULL 06/11] Hexagon (target/hexagon) Remove next_PC from runtime state, Taylor Simpson, 2022/11/10
- [PULL 04/11] Hexagon (target/hexagon) Only use branch_taken when packet has multi cof, Taylor Simpson, 2022/11/10
- [PULL 10/11] Hexagon (target/hexagon) Use direct block chaining for direct jump/branch, Taylor Simpson, 2022/11/10
- Re: [PULL 00/11] Hexagon bug fixes and performance improvement, Stefan Hajnoczi, 2022/11/10