From nobody Thu Dec 18 19:36:01 2025 Delivered-To: importer@patchew.org Authentication-Results: mx.zohomail.com; dkim=fail; spf=pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org; dmarc=fail(p=none dis=none) header.from=linaro.org Return-Path: Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) by mx.zohomail.com with SMTPS id 1635485462334568.5734659321575; Thu, 28 Oct 2021 22:31:02 -0700 (PDT) Received: from localhost ([::1]:42486 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1mgKTl-00060w-5Z for importer@patchew.org; Fri, 29 Oct 2021 01:31:01 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]:52192) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1mgJdn-0002Zq-Tb for qemu-devel@nongnu.org; Fri, 29 Oct 2021 00:37:19 -0400 Received: from mail-pg1-x534.google.com ([2607:f8b0:4864:20::534]:44568) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_128_GCM_SHA256:128) (Exim 4.90_1) (envelope-from ) id 1mgJdj-0000Nr-KP for qemu-devel@nongnu.org; Fri, 29 Oct 2021 00:37:19 -0400 Received: by mail-pg1-x534.google.com with SMTP id c4so8674000pgv.11 for ; Thu, 28 Oct 2021 21:37:14 -0700 (PDT) Received: from localhost.localdomain (174-21-75-75.tukw.qwest.net. [174.21.75.75]) by smtp.gmail.com with ESMTPSA id b8sm5228232pfv.56.2021.10.28.21.37.13 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 28 Oct 2021 21:37:13 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=BKLftSTQsQ1gAc3QxL60hYM2F6IdjBMip2wzW+AEFac=; b=wBqXpHQ5V/CVtrRPaM4HIyjlFrbgs1kdXHvSF1IV2QRGODj/oA2PZ5CXJGlv/l3RAV 3u1VkXfvFRTUyTJVRw71IfPEmG6IbtBvELWfL0LldueHD9goUX8vtH4tw9WZTQJbMh8l TCYANzdFTVU5gXM6mmkxeNhA0K1mZSUXXNY52zZu67FY73TO1JdqwdWSwYZtVG6Y+e2i JHoX7hTA4FSZPRxzq+hwm/+KzsxeG62bJoggQLNDXo9a2S2jOkG+4a38IYECQKphatYg CtT81NnXxSSG6itC5h/37XP0oJ35shJVdJWN4mgM1RfotFidls0yMk21oG75zd4IqHK+ v2iw== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=BKLftSTQsQ1gAc3QxL60hYM2F6IdjBMip2wzW+AEFac=; b=q/w67dMmImE7Yu9Z9pZERAret3/v/8piCFtV6tMJfnuVS/l9A6cUjd2EaYgHtapBRj xetc2qxGrwu9oe94XBezWVrTXJdyM+PpYUj7D3U4D7tWwYPGmeBo1L6d6YL4GM9782x2 fzzYYmGsXAwP2UoHmTo95ysMBCACrgNRg8k+KxNYgwo1fe2y5Pc1yqEZdWPzGbWs6QwA 6pj5vz49FtJrhV5BS1Q/wS4YYOPI0VOb39zoa7Vn94G8lOeDx6V2wwlenQl3D0C1kTQs Nzig48JMqsvOrDf3zIDKKjsf23NoLkcylV9WFYSflO+gRY2rGTJHwM2pJfgxKeGvAmn+ Z8AA== X-Gm-Message-State: AOAM532HEQzINKryOtuJISSTJLQ3ZvPKlx2kq5kF3ieH+ZoJmZenOq9J 158KpyRaU1ZOruuIEnJy90BdUsBw8ujNmA== X-Google-Smtp-Source: ABdhPJzbnoE3EVzr5jbBFBIQJqwbtp1P0hc98hfkXH1YQHul4zAHY/ZkRsvK+8iHE8pRVfsqeoYJew== X-Received: by 2002:a62:8689:0:b0:47b:dc40:19cc with SMTP id x131-20020a628689000000b0047bdc4019ccmr8422088pfd.77.1635482233908; Thu, 28 Oct 2021 21:37:13 -0700 (PDT) From: Richard Henderson To: qemu-devel@nongnu.org Subject: [PULL v2 53/60] tcg/optimize: Optimize sign extensions Date: Thu, 28 Oct 2021 21:33:22 -0700 Message-Id: <20211029043329.1518029-54-richard.henderson@linaro.org> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20211029043329.1518029-1-richard.henderson@linaro.org> References: <20211029043329.1518029-1-richard.henderson@linaro.org> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Received-SPF: pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) client-ip=209.51.188.17; envelope-from=qemu-devel-bounces+importer=patchew.org@nongnu.org; helo=lists.gnu.org; Received-SPF: pass client-ip=2607:f8b0:4864:20::534; envelope-from=richard.henderson@linaro.org; helo=mail-pg1-x534.google.com X-Spam_score_int: -20 X-Spam_score: -2.1 X-Spam_bar: -- X-Spam_report: (-2.1 / 5.0 requ) BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_NONE=-0.0001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Luis Pires , =?UTF-8?q?Alex=20Benn=C3=A9e?= Errors-To: qemu-devel-bounces+importer=patchew.org@nongnu.org Sender: "Qemu-devel" X-ZohoMail-DKIM: fail (Header signature does not verify) X-ZM-MESSAGEID: 1635485463676100001 Certain targets, like riscv, produce signed 32-bit results. This can lead to lots of redundant extensions as values are manipulated. Begin by tracking only the obvious sign-extensions, and converting them to simple copies when possible. Reviewed-by: Alex Benn=C3=A9e Reviewed-by: Luis Pires Signed-off-by: Richard Henderson --- tcg/optimize.c | 123 ++++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 102 insertions(+), 21 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index 7ac63c9231..ef202abbcb 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -43,6 +43,7 @@ typedef struct TempOptInfo { TCGTemp *next_copy; uint64_t val; uint64_t z_mask; /* mask bit is 0 if and only if value bit is 0 */ + uint64_t s_mask; /* a left-aligned mask of clrsb(value) bits. */ } TempOptInfo; =20 typedef struct OptContext { @@ -53,9 +54,37 @@ typedef struct OptContext { /* In flight values from optimization. */ uint64_t a_mask; /* mask bit is 0 iff value identical to first input = */ uint64_t z_mask; /* mask bit is 0 iff value bit is 0 */ + uint64_t s_mask; /* mask of clrsb(value) bits */ TCGType type; } OptContext; =20 +/* Calculate the smask for a specific value. */ +static uint64_t smask_from_value(uint64_t value) +{ + int rep =3D clrsb64(value); + return ~(~0ull >> rep); +} + +/* + * Calculate the smask for a given set of known-zeros. + * If there are lots of zeros on the left, we can consider the remainder + * an unsigned field, and thus the corresponding signed field is one bit + * larger. + */ +static uint64_t smask_from_zmask(uint64_t zmask) +{ + /* + * Only the 0 bits are significant for zmask, thus the msb itself + * must be zero, else we have no sign information. + */ + int rep =3D clz64(zmask); + if (rep =3D=3D 0) { + return 0; + } + rep -=3D 1; + return ~(~0ull >> rep); +} + static inline TempOptInfo *ts_info(TCGTemp *ts) { return ts->state_ptr; @@ -94,6 +123,7 @@ static void reset_ts(TCGTemp *ts) ti->prev_copy =3D ts; ti->is_const =3D false; ti->z_mask =3D -1; + ti->s_mask =3D 0; } =20 static void reset_temp(TCGArg arg) @@ -124,9 +154,11 @@ static void init_ts_info(OptContext *ctx, TCGTemp *ts) ti->is_const =3D true; ti->val =3D ts->val; ti->z_mask =3D ts->val; + ti->s_mask =3D smask_from_value(ts->val); } else { ti->is_const =3D false; ti->z_mask =3D -1; + ti->s_mask =3D 0; } } =20 @@ -220,6 +252,7 @@ static bool tcg_opt_gen_mov(OptContext *ctx, TCGOp *op,= TCGArg dst, TCGArg src) op->args[1] =3D src; =20 di->z_mask =3D si->z_mask; + di->s_mask =3D si->s_mask; =20 if (src_ts->type =3D=3D dst_ts->type) { TempOptInfo *ni =3D ts_info(si->next_copy); @@ -658,13 +691,15 @@ static void finish_folding(OptContext *ctx, TCGOp *op) =20 nb_oargs =3D def->nb_oargs; for (i =3D 0; i < nb_oargs; i++) { - reset_temp(op->args[i]); + TCGTemp *ts =3D arg_temp(op->args[i]); + reset_ts(ts); /* - * Save the corresponding known-zero bits mask for the + * Save the corresponding known-zero/sign bits mask for the * first output argument (only one supported so far). */ if (i =3D=3D 0) { - arg_info(op->args[i])->z_mask =3D ctx->z_mask; + ts_info(ts)->z_mask =3D ctx->z_mask; + ts_info(ts)->s_mask =3D ctx->s_mask; } } } @@ -714,6 +749,7 @@ static bool fold_masks(OptContext *ctx, TCGOp *op) { uint64_t a_mask =3D ctx->a_mask; uint64_t z_mask =3D ctx->z_mask; + uint64_t s_mask =3D ctx->s_mask; =20 /* * 32-bit ops generate 32-bit results, which for the purpose of @@ -725,7 +761,9 @@ static bool fold_masks(OptContext *ctx, TCGOp *op) if (ctx->type =3D=3D TCG_TYPE_I32) { a_mask =3D (int32_t)a_mask; z_mask =3D (int32_t)z_mask; + s_mask |=3D MAKE_64BIT_MASK(32, 32); ctx->z_mask =3D z_mask; + ctx->s_mask =3D s_mask; } =20 if (z_mask =3D=3D 0) { @@ -1072,7 +1110,7 @@ static bool fold_brcond2(OptContext *ctx, TCGOp *op) =20 static bool fold_bswap(OptContext *ctx, TCGOp *op) { - uint64_t z_mask, sign; + uint64_t z_mask, s_mask, sign; =20 if (arg_is_const(op->args[1])) { uint64_t t =3D arg_info(op->args[1])->val; @@ -1082,6 +1120,7 @@ static bool fold_bswap(OptContext *ctx, TCGOp *op) } =20 z_mask =3D arg_info(op->args[1])->z_mask; + switch (op->opc) { case INDEX_op_bswap16_i32: case INDEX_op_bswap16_i64: @@ -1100,6 +1139,7 @@ static bool fold_bswap(OptContext *ctx, TCGOp *op) default: g_assert_not_reached(); } + s_mask =3D smask_from_zmask(z_mask); =20 switch (op->args[2] & (TCG_BSWAP_OZ | TCG_BSWAP_OS)) { case TCG_BSWAP_OZ: @@ -1108,14 +1148,17 @@ static bool fold_bswap(OptContext *ctx, TCGOp *op) /* If the sign bit may be 1, force all the bits above to 1. */ if (z_mask & sign) { z_mask |=3D sign; + s_mask =3D sign << 1; } break; default: /* The high bits are undefined: force all bits above the sign to 1= . */ z_mask |=3D sign << 1; + s_mask =3D 0; break; } ctx->z_mask =3D z_mask; + ctx->s_mask =3D s_mask; =20 return fold_masks(ctx, op); } @@ -1263,21 +1306,24 @@ static bool fold_eqv(OptContext *ctx, TCGOp *op) static bool fold_extract(OptContext *ctx, TCGOp *op) { uint64_t z_mask_old, z_mask; + int pos =3D op->args[2]; + int len =3D op->args[3]; =20 if (arg_is_const(op->args[1])) { uint64_t t; =20 t =3D arg_info(op->args[1])->val; - t =3D extract64(t, op->args[2], op->args[3]); + t =3D extract64(t, pos, len); return tcg_opt_gen_movi(ctx, op, op->args[0], t); } =20 z_mask_old =3D arg_info(op->args[1])->z_mask; - z_mask =3D extract64(z_mask_old, op->args[2], op->args[3]); - if (op->args[2] =3D=3D 0) { + z_mask =3D extract64(z_mask_old, pos, len); + if (pos =3D=3D 0) { ctx->a_mask =3D z_mask_old ^ z_mask; } ctx->z_mask =3D z_mask; + ctx->s_mask =3D smask_from_zmask(z_mask); =20 return fold_masks(ctx, op); } @@ -1303,14 +1349,16 @@ static bool fold_extract2(OptContext *ctx, TCGOp *o= p) =20 static bool fold_exts(OptContext *ctx, TCGOp *op) { - uint64_t z_mask_old, z_mask, sign; + uint64_t s_mask_old, s_mask, z_mask, sign; bool type_change =3D false; =20 if (fold_const1(ctx, op)) { return true; } =20 - z_mask_old =3D z_mask =3D arg_info(op->args[1])->z_mask; + z_mask =3D arg_info(op->args[1])->z_mask; + s_mask =3D arg_info(op->args[1])->s_mask; + s_mask_old =3D s_mask; =20 switch (op->opc) { CASE_OP_32_64(ext8s): @@ -1334,10 +1382,14 @@ static bool fold_exts(OptContext *ctx, TCGOp *op) =20 if (z_mask & sign) { z_mask |=3D sign; - } else if (!type_change) { - ctx->a_mask =3D z_mask_old ^ z_mask; } + s_mask |=3D sign << 1; + ctx->z_mask =3D z_mask; + ctx->s_mask =3D s_mask; + if (!type_change) { + ctx->a_mask =3D s_mask & ~s_mask_old; + } =20 return fold_masks(ctx, op); } @@ -1376,6 +1428,7 @@ static bool fold_extu(OptContext *ctx, TCGOp *op) } =20 ctx->z_mask =3D z_mask; + ctx->s_mask =3D smask_from_zmask(z_mask); if (!type_change) { ctx->a_mask =3D z_mask_old ^ z_mask; } @@ -1606,8 +1659,12 @@ static bool fold_qemu_ld(OptContext *ctx, TCGOp *op) MemOp mop =3D get_memop(oi); int width =3D 8 * memop_size(mop); =20 - if (!(mop & MO_SIGN) && width < 64) { - ctx->z_mask =3D MAKE_64BIT_MASK(0, width); + if (width < 64) { + ctx->s_mask =3D MAKE_64BIT_MASK(width, 64 - width); + if (!(mop & MO_SIGN)) { + ctx->z_mask =3D MAKE_64BIT_MASK(0, width); + ctx->s_mask <<=3D 1; + } } =20 /* Opcodes that touch guest memory stop the mb optimization. */ @@ -1726,23 +1783,31 @@ static bool fold_setcond2(OptContext *ctx, TCGOp *o= p) =20 static bool fold_sextract(OptContext *ctx, TCGOp *op) { - int64_t z_mask_old, z_mask; + uint64_t z_mask, s_mask, s_mask_old; + int pos =3D op->args[2]; + int len =3D op->args[3]; =20 if (arg_is_const(op->args[1])) { uint64_t t; =20 t =3D arg_info(op->args[1])->val; - t =3D sextract64(t, op->args[2], op->args[3]); + t =3D sextract64(t, pos, len); return tcg_opt_gen_movi(ctx, op, op->args[0], t); } =20 - z_mask_old =3D arg_info(op->args[1])->z_mask; - z_mask =3D sextract64(z_mask_old, op->args[2], op->args[3]); - if (op->args[2] =3D=3D 0 && z_mask >=3D 0) { - ctx->a_mask =3D z_mask_old ^ z_mask; - } + z_mask =3D arg_info(op->args[1])->z_mask; + z_mask =3D sextract64(z_mask, pos, len); ctx->z_mask =3D z_mask; =20 + s_mask_old =3D arg_info(op->args[1])->s_mask; + s_mask =3D sextract64(s_mask_old, pos, len); + s_mask |=3D MAKE_64BIT_MASK(len, 64 - len); + ctx->s_mask =3D s_mask; + + if (pos =3D=3D 0) { + ctx->a_mask =3D s_mask & ~s_mask_old; + } + return fold_masks(ctx, op); } =20 @@ -1819,14 +1884,26 @@ static bool fold_tcg_ld(OptContext *ctx, TCGOp *op) { /* We can't do any folding with a load, but we can record bits. */ switch (op->opc) { + CASE_OP_32_64(ld8s): + ctx->s_mask =3D MAKE_64BIT_MASK(8, 56); + break; CASE_OP_32_64(ld8u): ctx->z_mask =3D MAKE_64BIT_MASK(0, 8); + ctx->s_mask =3D MAKE_64BIT_MASK(9, 55); + break; + CASE_OP_32_64(ld16s): + ctx->s_mask =3D MAKE_64BIT_MASK(16, 48); break; CASE_OP_32_64(ld16u): ctx->z_mask =3D MAKE_64BIT_MASK(0, 16); + ctx->s_mask =3D MAKE_64BIT_MASK(17, 47); + break; + case INDEX_op_ld32s_i64: + ctx->s_mask =3D MAKE_64BIT_MASK(32, 32); break; case INDEX_op_ld32u_i64: ctx->z_mask =3D MAKE_64BIT_MASK(0, 32); + ctx->s_mask =3D MAKE_64BIT_MASK(33, 31); break; default: g_assert_not_reached(); @@ -1889,9 +1966,10 @@ void tcg_optimize(TCGContext *s) ctx.type =3D TCG_TYPE_I32; } =20 - /* Assume all bits affected, and no bits known zero. */ + /* Assume all bits affected, no bits known zero, no sign reps. */ ctx.a_mask =3D -1; ctx.z_mask =3D -1; + ctx.s_mask =3D 0; =20 /* * Process each opcode. @@ -1964,8 +2042,11 @@ void tcg_optimize(TCGContext *s) case INDEX_op_extrh_i64_i32: done =3D fold_extu(&ctx, op); break; + CASE_OP_32_64(ld8s): CASE_OP_32_64(ld8u): + CASE_OP_32_64(ld16s): CASE_OP_32_64(ld16u): + case INDEX_op_ld32s_i64: case INDEX_op_ld32u_i64: done =3D fold_tcg_ld(&ctx, op); break; --=20 2.25.1