[v4,08/11] Hexagon (target/hexagon) Add overrides for compound compare and jump

Message ID	20221108040552.22175-9-tsimpson@quicinc.com (mailing list archive)
State	New, archived
Headers	show Return-Path: <qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org> From: Taylor Simpson <tsimpson@quicinc.com> To: qemu-devel@nongnu.org Cc: tsimpson@quicinc.com, richard.henderson@linaro.org, philmd@linaro.org, ale@rev.ng, anjo@rev.ng, bcain@quicinc.com, quic_mathbern@quicinc.com Subject: [PATCH v4 08/11] Hexagon (target/hexagon) Add overrides for compound compare and jump Date: Mon, 7 Nov 2022 20:05:49 -0800 Message-Id: <20221108040552.22175-9-tsimpson@quicinc.com> In-Reply-To: <20221108040552.22175-1-tsimpson@quicinc.com> References: <20221108040552.22175-1-tsimpson@quicinc.com> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: base64 Received-SPF: pass client-ip=205.220.168.131; envelope-from=tsimpson@qualcomm.com; helo=mx0a-0031df01.pphosted.com X-Spam_score_int: -24 X-Spam_score: -2.5 X-Spam_bar: -- X-Spam_report: (-2.5 / 5.0 requ) BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, HEADER_FROM_DIFFERENT_DOMAINS=0.25, RCVD_IN_DNSWL_LOW=-0.7, SPF_HELO_NONE=0.001, SPF_PASS=-0.001 autolearn=ham autolearn_force=no X-Spam_action: no action Precedence: list Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org
Series	Hexagon (target/hexagon) performance and bug fixes \| expand [v4,00/11] Hexagon (target/hexagon) performance and bug fixes [v4,01/11] Hexagon (target/hexagon) Add pkt and insn to DisasContext [v4,02/11] Hexagon (target/hexagon) Fix predicated assignment to .tmp and .cur [v4,03/11] Hexagon (target/hexagon) Add overrides for S2_asr_r_r_sat/S2_asl_r_r_sat [v4,04/11] Hexagon (target/hexagon) Only use branch_taken when packet has multi cof [v4,05/11] Hexagon (target/hexagon) Remove PC from the runtime state [v4,06/11] Hexagon (target/hexagon) Remove next_PC from runtime state [v4,07/11] Hexagon (target/hexagon) Add overrides for direct call instructions [v4,08/11] Hexagon (target/hexagon) Add overrides for compound compare and jump [v4,09/11] Hexagon (target/hexagon) Add overrides for various forms of jump [v4,10/11] Hexagon (target/hexagon) Use direct block chaining for direct jump/branch [v4,11/11] Hexagon (target/hexagon) Use direct block chaining for tight loops

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 ffaea5a639..00df8851a2 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 =

[v4,08/11] Hexagon (target/hexagon) Add overrides for compound compare and jump

Commit Message

Patch