diff options
42 files changed, 2253 insertions, 459 deletions
diff --git a/MAINTAINERS b/MAINTAINERS index 83c127f0d6..d676c73f88 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -127,9 +127,11 @@ F: include/sysemu/cpus.h FPU emulation M: Aurelien Jarno <aurelien@aurel32.net> M: Peter Maydell <peter.maydell@linaro.org> -S: Odd Fixes +M: Alex Bennée <alex.bennee@linaro.org> +S: Maintained F: fpu/ F: include/fpu/ +F: tests/fp/ Alpha M: Richard Henderson <rth@twiddle.net> diff --git a/contrib/gitdm/aliases b/contrib/gitdm/aliases new file mode 100644 index 0000000000..07fd3391a5 --- /dev/null +++ b/contrib/gitdm/aliases @@ -0,0 +1,27 @@ +# +# This is the email aliases file, mapping secondary addresses +# onto a single, canonical address. Duplicates some info from .mailmap +# + +# weird commits +balrog@c046a42c-6fe2-441c-8c8c-71466251a162 balrogg@gmail.com +aliguori@c046a42c-6fe2-441c-8c8c-71466251a162 anthony@codemonkey.ws +aurel32@c046a42c-6fe2-441c-8c8c-71466251a162 aurelien@aurel32.net +blueswir1@c046a42c-6fe2-441c-8c8c-71466251a162 blauwirbel@gmail.com +edgar_igl@c046a42c-6fe2-441c-8c8c-71466251a162 edgar.iglesias@gmail.com +bellard@c046a42c-6fe2-441c-8c8c-71466251a162 fabrice@bellard.org +j_mayer@c046a42c-6fe2-441c-8c8c-71466251a162 l_indien@magic.fr +pbrook@c046a42c-6fe2-441c-8c8c-71466251a162 paul@codesourcery.com +ths@c046a42c-6fe2-441c-8c8c-71466251a162 ths@networkno.de +malc@c046a42c-6fe2-441c-8c8c-71466251a162 av1474@comtv.ru + +# There is also a: +# (no author) <(no author)@c046a42c-6fe2-441c-8c8c-71466251a162> +# for the cvs2svn initialization commit e63c3dc74bf. + +# Next, translate a few commits where mailman rewrote the From: line due +# to strict SPF, although we prefer to avoid adding more entries like that. +"Ed Swierk via Qemu-devel" eswierk@skyportsystems.com +"Ian McKellar via Qemu-devel" ianloic@google.com +"Julia Suvorova via Qemu-devel" jusual@mail.ru +"Justin Terry (VM) via Qemu-devel" juterry@microsoft.com diff --git a/contrib/gitdm/domain-map b/contrib/gitdm/domain-map new file mode 100644 index 0000000000..8cbbcfe93d --- /dev/null +++ b/contrib/gitdm/domain-map @@ -0,0 +1,19 @@ +# +# QEMU gitdm domain-map +# +# This maps email domains to nice easy to read company names +# + +amd.com AMD +greensocs.com GreenSocs +ibm.com IBM +igalia.com Igalia +linaro.org Linaro +oracle.com Oracle +redhat.com Red Hat +siemens.com Siemens +sifive.com SiFive +suse.de SUSE +virtuozzo.com Virtuozzo +wdc.com Western Digital +xilinx.com Xilinx diff --git a/contrib/gitdm/filetypes.txt b/contrib/gitdm/filetypes.txt new file mode 100644 index 0000000000..15d6f803b9 --- /dev/null +++ b/contrib/gitdm/filetypes.txt @@ -0,0 +1,146 @@ +# -*- coding:utf-8 -*- +# Copyright (C) 2006 Libresoft +# +# This program is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 2 of the License, or +# (at your option any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU Library General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program; if not, write to the Free Software +# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. +# +# Authors : Gregorio Robles <grex@gsyc.escet.urjc.es> +# Authors : Germán Póo-Caamaño <gpoo@gnome.org> +# +# This QEMU version is a cut-down version of what originally shipped +# in the gitdm sample-config directory. +# +# This file contains associations parameters regarding filetypes +# (documentation, develompent, multimedia, images...) +# +# format: +# filetype <type> <regex> [<comment>] +# +# Order: +# The list should keep an order, so filetypes can be counted properly. +# ie. we want ltmain.sh -> 'build' instead of 'code'. +# +# If there is an filetype which is not in order but has values, it will +# be added at the end. +# +order build,tests,code,documentation,devel-doc,blobs + +# +# +# Code files (headers and the like included +# (most common languages first +# +filetype code \.c$ # C +filetype code \.inc.c$ # C +filetype code \.C$ # C++ +filetype code \.cpp$ # C++ +filetype code \.c\+\+$ # C++ +filetype code \.cxx$ # C++ +filetype code \.cc$ # C++ +filetype code \.h$ # C or C++ header +filetype code \.hh$ # C++ header +filetype code \.hpp$ # C++ header +filetype code \.hxx$ # C++ header +filetype code \.sh$ # Shell +filetype code \.pl$ # Perl +filetype code \.py$ # Python +filetype code \.s$ # Assembly +filetype code \.S$ # Assembly +filetype code \.asm$ # Assembly +filetype code \.awk$ # awk +filetype code ^common$ # script fragements +filetype code ^common.*$ # script fragements +filetype code (qom|qmp)-\w+$ # python script fragments + +# +# Interface/api files +# +filetype interface \.json$ # json +filetype interface \.hx$ # documented options + +# +# Test related blobs (unfortunately we can't filter out test code) +# +filetype tests \.hex$ +filetype tests \d{2,3}$ # test data 00-999 +filetype tests ^[A-Z]{4}$ # ACPI test data +filetype tests ^[A-Z]{4}\.*$ # ACPI test data +filetype tests \.out$ +filetype tests \.out\.nocache$ +filetype tests \.err$ +filetype tests \.exit$ # bad-if-FOO.exit etc +filetype tests \.decode$ +filetype tests \.yml$ # travis/shippable config + +# +# Development documentation files (for hacking generally) +# +filetype devel-doc ^readme.*$ +filetype devel-doc ^changelog.* +filetype devel-doc ^hacking.*$ +filetype devel-doc ^licen(s|c)e.*$ +filetype devel-doc ^copying.*$ +filetype devel-doc ^MAINTAINERS$ +filetype devel-doc ^BSD-2-Clause$ +filetype devel-doc ^BSD-3-Clause$ +filetype devel-doc ^GPL-2.0$ +filetype devel-doc \.txt$ +filetype devel-doc \.rst$ +filetype devel-doc \.texi$ +filetype devel-doc \.pod$ + +# +# Building, compiling, and configuration admin files +# +filetype build configure.*$ +filetype build Makefile$ +filetype build Makefile\.*$ +filetype build config$ +filetype build conf$ +filetype build \.cfg$ +filetype build \.mk$ +filetype build \.mak$ +filetype build \.docker$ +filetype build \.pre$ +filetype build ^.gitignore$ +filetype build ^.gitmodules$ +filetype build ^.gitpublish$ +filetype build ^.mailmap$ +filetype build ^.dir-locals.el$ +filetype build ^.editorconfig$ +filetype build ^.exrc$ +filetype build ^.gdbinit$ +filetype build \.cocci$ # Coccinelle semantic patches + +# +# Misc blobs +# +filetype blobs \.bin$ +filetype blobs \.dtb$ +filetype blobs \.dts$ +filetype blobs \.rom$ +filetype blobs \.img$ +filetype blobs \.ndrv$ +filetype blobs \.bmp$ +filetype blobs \.svg$ +filetype blobs ^pi_10.com$ + + +# +# Documentation files +# +filetype documentation \.html$ +filetype documentation \.txt$ +filetype documentation \.texi$ +filetype documentation \.po$ # translation files diff --git a/contrib/gitdm/group-map-academics b/contrib/gitdm/group-map-academics new file mode 100644 index 0000000000..08f9d81d13 --- /dev/null +++ b/contrib/gitdm/group-map-academics @@ -0,0 +1,14 @@ +# +# QEMU is quite often used for academic research purposes and we like +# it even better when the work is up-streamed so the project can +# benefit. +# +# We group our academic contributors here +# + +# Institute for System Programming of Russian Academy of Science +ispras.ru + +# Columbia University +cs.columbia.edu +cota@braap.org diff --git a/contrib/gitdm/group-map-cadence b/contrib/gitdm/group-map-cadence new file mode 100644 index 0000000000..ab97dd2fc3 --- /dev/null +++ b/contrib/gitdm/group-map-cadence @@ -0,0 +1,3 @@ +# Cadence Design Systems + +jcmvbkbc@gmail.com diff --git a/contrib/gitdm/group-map-codeweavers b/contrib/gitdm/group-map-codeweavers new file mode 100644 index 0000000000..c4803489e2 --- /dev/null +++ b/contrib/gitdm/group-map-codeweavers @@ -0,0 +1 @@ +sergio.g.delreal@gmail.com diff --git a/contrib/gitdm/group-map-ibm b/contrib/gitdm/group-map-ibm new file mode 100644 index 0000000000..b66db5f4a8 --- /dev/null +++ b/contrib/gitdm/group-map-ibm @@ -0,0 +1,6 @@ +# +# Some IBM contributors submit via another domain +# + +clg@kaod.org +groug@kaod.org diff --git a/contrib/gitdm/group-map-individuals b/contrib/gitdm/group-map-individuals new file mode 100644 index 0000000000..afdbe7d460 --- /dev/null +++ b/contrib/gitdm/group-map-individuals @@ -0,0 +1,10 @@ +# +# Individual and personal contributors +# +# This is simply to allow prolific developers with no company +# affiliations to be grouped together in the summary stats. +# + +f4bug@amsat.org +mjt@tls.msk.ru +mark.cave-ayland@ilande.co.uk diff --git a/contrib/gitdm/group-map-redhat b/contrib/gitdm/group-map-redhat new file mode 100644 index 0000000000..6d05c6b54f --- /dev/null +++ b/contrib/gitdm/group-map-redhat @@ -0,0 +1,7 @@ +# +# Red Hat contributors using non-corporate email +# + +david@gibson.dropbear.id.au +laurent@vivier.eu +pjp@fedoraproject.org diff --git a/contrib/gitdm/group-map-wavecomp b/contrib/gitdm/group-map-wavecomp new file mode 100644 index 0000000000..c571a52c65 --- /dev/null +++ b/contrib/gitdm/group-map-wavecomp @@ -0,0 +1,18 @@ +# +# Wave Computing acquired MIPS in June 2018. Also, from February 2013 +# to October 2017, MIPS was owned by Imagination Technologies. +# + +aleksandar.markovic@imgtec.com +aleksandar.markovic@mips.com +amarkovic@wavecomp.com +arikalo@wavecomp.com +dnikolic@wavecomp.com +james.hogan@mips.com +matthew.fortune@mips.com +paul.burton@imgtec.com +pburton@wavecomp.com +smarkovic@wavecomp.com +yongbok.kim@imgtec.com +yongbok.kim@mips.com +ysu@wavecomp.com diff --git a/fpu/softfloat.c b/fpu/softfloat.c index e1eef954e6..59eac97d10 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -83,6 +83,7 @@ this code that are retained. * target-dependent and needs the TARGET_* macros. */ #include "qemu/osdep.h" +#include <math.h> #include "qemu/bitops.h" #include "fpu/softfloat.h" @@ -95,6 +96,324 @@ this code that are retained. *----------------------------------------------------------------------------*/ #include "fpu/softfloat-macros.h" +/* + * Hardfloat + * + * Fast emulation of guest FP instructions is challenging for two reasons. + * First, FP instruction semantics are similar but not identical, particularly + * when handling NaNs. Second, emulating at reasonable speed the guest FP + * exception flags is not trivial: reading the host's flags register with a + * feclearexcept & fetestexcept pair is slow [slightly slower than soft-fp], + * and trapping on every FP exception is not fast nor pleasant to work with. + * + * We address these challenges by leveraging the host FPU for a subset of the + * operations. To do this we expand on the idea presented in this paper: + * + * Guo, Yu-Chuan, et al. "Translating the ARM Neon and VFP instructions in a + * binary translator." Software: Practice and Experience 46.12 (2016):1591-1615. + * + * The idea is thus to leverage the host FPU to (1) compute FP operations + * and (2) identify whether FP exceptions occurred while avoiding + * expensive exception flag register accesses. + * + * An important optimization shown in the paper is that given that exception + * flags are rarely cleared by the guest, we can avoid recomputing some flags. + * This is particularly useful for the inexact flag, which is very frequently + * raised in floating-point workloads. + * + * We optimize the code further by deferring to soft-fp whenever FP exception + * detection might get hairy. Two examples: (1) when at least one operand is + * denormal/inf/NaN; (2) when operands are not guaranteed to lead to a 0 result + * and the result is < the minimum normal. + */ +#define GEN_INPUT_FLUSH__NOCHECK(name, soft_t) \ + static inline void name(soft_t *a, float_status *s) \ + { \ + if (unlikely(soft_t ## _is_denormal(*a))) { \ + *a = soft_t ## _set_sign(soft_t ## _zero, \ + soft_t ## _is_neg(*a)); \ + s->float_exception_flags |= float_flag_input_denormal; \ + } \ + } + +GEN_INPUT_FLUSH__NOCHECK(float32_input_flush__nocheck, float32) +GEN_INPUT_FLUSH__NOCHECK(float64_input_flush__nocheck, float64) +#undef GEN_INPUT_FLUSH__NOCHECK + +#define GEN_INPUT_FLUSH1(name, soft_t) \ + static inline void name(soft_t *a, float_status *s) \ + { \ + if (likely(!s->flush_inputs_to_zero)) { \ + return; \ + } \ + soft_t ## _input_flush__nocheck(a, s); \ + } + +GEN_INPUT_FLUSH1(float32_input_flush1, float32) +GEN_INPUT_FLUSH1(float64_input_flush1, float64) +#undef GEN_INPUT_FLUSH1 + +#define GEN_INPUT_FLUSH2(name, soft_t) \ + static inline void name(soft_t *a, soft_t *b, float_status *s) \ + { \ + if (likely(!s->flush_inputs_to_zero)) { \ + return; \ + } \ + soft_t ## _input_flush__nocheck(a, s); \ + soft_t ## _input_flush__nocheck(b, s); \ + } + +GEN_INPUT_FLUSH2(float32_input_flush2, float32) +GEN_INPUT_FLUSH2(float64_input_flush2, float64) +#undef GEN_INPUT_FLUSH2 + +#define GEN_INPUT_FLUSH3(name, soft_t) \ + static inline void name(soft_t *a, soft_t *b, soft_t *c, float_status *s) \ + { \ + if (likely(!s->flush_inputs_to_zero)) { \ + return; \ + } \ + soft_t ## _input_flush__nocheck(a, s); \ + soft_t ## _input_flush__nocheck(b, s); \ + soft_t ## _input_flush__nocheck(c, s); \ + } + +GEN_INPUT_FLUSH3(float32_input_flush3, float32) +GEN_INPUT_FLUSH3(float64_input_flush3, float64) +#undef GEN_INPUT_FLUSH3 + +/* + * Choose whether to use fpclassify or float32/64_* primitives in the generated + * hardfloat functions. Each combination of number of inputs and float size + * gets its own value. + */ +#if defined(__x86_64__) +# define QEMU_HARDFLOAT_1F32_USE_FP 0 +# define QEMU_HARDFLOAT_1F64_USE_FP 1 +# define QEMU_HARDFLOAT_2F32_USE_FP 0 +# define QEMU_HARDFLOAT_2F64_USE_FP 1 +# define QEMU_HARDFLOAT_3F32_USE_FP 0 +# define QEMU_HARDFLOAT_3F64_USE_FP 1 +#else +# define QEMU_HARDFLOAT_1F32_USE_FP 0 +# define QEMU_HARDFLOAT_1F64_USE_FP 0 +# define QEMU_HARDFLOAT_2F32_USE_FP 0 +# define QEMU_HARDFLOAT_2F64_USE_FP 0 +# define QEMU_HARDFLOAT_3F32_USE_FP 0 +# define QEMU_HARDFLOAT_3F64_USE_FP 0 +#endif + +/* + * QEMU_HARDFLOAT_USE_ISINF chooses whether to use isinf() over + * float{32,64}_is_infinity when !USE_FP. + * On x86_64/aarch64, using the former over the latter can yield a ~6% speedup. + * On power64 however, using isinf() reduces fp-bench performance by up to 50%. + */ +#if defined(__x86_64__) || defined(__aarch64__) +# define QEMU_HARDFLOAT_USE_ISINF 1 +#else +# define QEMU_HARDFLOAT_USE_ISINF 0 +#endif + +/* + * Some targets clear the FP flags before most FP operations. This prevents + * the use of hardfloat, since hardfloat relies on the inexact flag being + * already set. + */ +#if defined(TARGET_PPC) || defined(__FAST_MATH__) +# if defined(__FAST_MATH__) +# warning disabling hardfloat due to -ffast-math: hardfloat requires an exact \ + IEEE implementation +# endif +# define QEMU_NO_HARDFLOAT 1 +# define QEMU_SOFTFLOAT_ATTR QEMU_FLATTEN +#else +# define QEMU_NO_HARDFLOAT 0 +# define QEMU_SOFTFLOAT_ATTR QEMU_FLATTEN __attribute__((noinline)) +#endif + +static inline bool can_use_fpu(const float_status *s) +{ + if (QEMU_NO_HARDFLOAT) { + return false; + } + return likely(s->float_exception_flags & float_flag_inexact && + s->float_rounding_mode == float_round_nearest_even); +} + +/* + * Hardfloat generation functions. Each operation can have two flavors: + * either using softfloat primitives (e.g. float32_is_zero_or_normal) for + * most condition checks, or native ones (e.g. fpclassify). + * + * The flavor is chosen by the callers. Instead of using macros, we rely on the + * compiler to propagate constants and inline everything into the callers. + * + * We only generate functions for operations with two inputs, since only + * these are common enough to justify consolidating them into common code. + */ + +typedef union { + float32 s; + float h; +} union_float32; + +typedef union { + float64 s; + double h; +} union_float64; + +typedef bool (*f32_check_fn)(union_float32 a, union_float32 b); +typedef bool (*f64_check_fn)(union_float64 a, union_float64 b); + +typedef float32 (*soft_f32_op2_fn)(float32 a, float32 b, float_status *s); +typedef float64 (*soft_f64_op2_fn)(float64 a, float64 b, float_status *s); +typedef float (*hard_f32_op2_fn)(float a, float b); +typedef double (*hard_f64_op2_fn)(double a, double b); + +/* 2-input is-zero-or-normal */ +static inline bool f32_is_zon2(union_float32 a, union_float32 b) +{ + if (QEMU_HARDFLOAT_2F32_USE_FP) { + /* + * Not using a temp variable for consecutive fpclassify calls ends up + * generating faster code. + */ + return (fpclassify(a.h) == FP_NORMAL || fpclassify(a.h) == FP_ZERO) && + (fpclassify(b.h) == FP_NORMAL || fpclassify(b.h) == FP_ZERO); + } + return float32_is_zero_or_normal(a.s) && + float32_is_zero_or_normal(b.s); +} + +static inline bool f64_is_zon2(union_float64 a, union_float64 b) +{ + if (QEMU_HARDFLOAT_2F64_USE_FP) { + return (fpclassify(a.h) == FP_NORMAL || fpclassify(a.h) == FP_ZERO) && + (fpclassify(b.h) == FP_NORMAL || fpclassify(b.h) == FP_ZERO); + } + return float64_is_zero_or_normal(a.s) && + float64_is_zero_or_normal(b.s); +} + +/* 3-input is-zero-or-normal */ +static inline +bool f32_is_zon3(union_float32 a, union_float32 b, union_float32 c) +{ + if (QEMU_HARDFLOAT_3F32_USE_FP) { + return (fpclassify(a.h) == FP_NORMAL || fpclassify(a.h) == FP_ZERO) && + (fpclassify(b.h) == FP_NORMAL || fpclassify(b.h) == FP_ZERO) && + (fpclassify(c.h) == FP_NORMAL || fpclassify(c.h) == FP_ZERO); + } + return float32_is_zero_or_normal(a.s) && + float32_is_zero_or_normal(b.s) && + float32_is_zero_or_normal(c.s); +} + +static inline +bool f64_is_zon3(union_float64 a, union_float64 b, union_float64 c) +{ + if (QEMU_HARDFLOAT_3F64_USE_FP) { + return (fpclassify(a.h) == FP_NORMAL || fpclassify(a.h) == FP_ZERO) && + (fpclassify(b.h) == FP_NORMAL || fpclassify(b.h) == FP_ZERO) && + (fpclassify(c.h) == FP_NORMAL || fpclassify(c.h) == FP_ZERO); + } + return float64_is_zero_or_normal(a.s) && + float64_is_zero_or_normal(b.s) && + float64_is_zero_or_normal(c.s); +} + +static inline bool f32_is_inf(union_float32 a) +{ + if (QEMU_HARDFLOAT_USE_ISINF) { + return isinf(a.h); + } + return float32_is_infinity(a.s); +} + +static inline bool f64_is_inf(union_float64 a) +{ + if (QEMU_HARDFLOAT_USE_ISINF) { + return isinf(a.h); + } + return float64_is_infinity(a.s); +} + +/* Note: @fast_test and @post can be NULL */ +static inline float32 +float32_gen2(float32 xa, float32 xb, float_status *s, + hard_f32_op2_fn hard, soft_f32_op2_fn soft, + f32_check_fn pre, f32_check_fn post, + f32_check_fn fast_test, soft_f32_op2_fn fast_op) +{ + union_float32 ua, ub, ur; + + ua.s = xa; + ub.s = xb; + + if (unlikely(!can_use_fpu(s))) { + goto soft; + } + + float32_input_flush2(&ua.s, &ub.s, s); + if (unlikely(!pre(ua, ub))) { + goto soft; + } + if (fast_test && fast_test(ua, ub)) { + return fast_op(ua.s, ub.s, s); + } + + ur.h = hard(ua.h, ub.h); + if (unlikely(f32_is_inf(ur))) { + s->float_exception_flags |= float_flag_overflow; + } else if (unlikely(fabsf(ur.h) <= FLT_MIN)) { + if (post == NULL || post(ua, ub)) { + goto soft; + } + } + return ur.s; + + soft: + return soft(ua.s, ub.s, s); +} + +static inline float64 +float64_gen2(float64 xa, float64 xb, float_status *s, + hard_f64_op2_fn hard, soft_f64_op2_fn soft, + f64_check_fn pre, f64_check_fn post, + f64_check_fn fast_test, soft_f64_op2_fn fast_op) +{ + union_float64 ua, ub, ur; + + ua.s = xa; + ub.s = xb; + + if (unlikely(!can_use_fpu(s))) { + goto soft; + } + + float64_input_flush2(&ua.s, &ub.s, s); + if (unlikely(!pre(ua, ub))) { + goto soft; + } + if (fast_test && fast_test(ua, ub)) { + return fast_op(ua.s, ub.s, s); + } + + ur.h = hard(ua.h, ub.h); + if (unlikely(f64_is_inf(ur))) { + s->float_exception_flags |= float_flag_overflow; + } else if (unlikely(fabs(ur.h) <= DBL_MIN)) { + if (post == NULL || post(ua, ub)) { + goto soft; + } + } + return ur.s; + + soft: + return soft(ua.s, ub.s, s); +} + /*---------------------------------------------------------------------------- | Returns the fraction bits of the half-precision floating-point value `a'. *----------------------------------------------------------------------------*/ @@ -336,8 +655,8 @@ static inline float64 float64_pack_raw(FloatParts p) #include "softfloat-specialize.h" /* Canonicalize EXP and FRAC, setting CLS. */ -static FloatParts canonicalize(FloatParts part, const FloatFmt *parm, - float_status *status) +static FloatParts sf_canonicalize(FloatParts part, const FloatFmt *parm, + float_status *status) { if (part.exp == parm->exp_max && !parm->arm_althp) { if (part.frac == 0) { @@ -513,7 +832,7 @@ static FloatParts round_canonical(FloatParts p, float_status *s, static FloatParts float16a_unpack_canonical(float16 f, float_status *s, const FloatFmt *params) { - return canonicalize(float16_unpack_raw(f), params, s); + return sf_canonicalize(float16_unpack_raw(f), params, s); } static FloatParts float16_unpack_canonical(float16 f, float_status *s) @@ -534,7 +853,7 @@ static float16 float16_round_pack_canonical(FloatParts p, float_status *s) static FloatParts float32_unpack_canonical(float32 f, float_status *s) { - return canonicalize(float32_unpack_raw(f), &float32_params, s); + return sf_canonicalize(float32_unpack_raw(f), &float32_params, s); } static float32 float32_round_pack_canonical(FloatParts p, float_status *s) @@ -544,7 +863,7 @@ static float32 float32_round_pack_canonical(FloatParts p, float_status *s) static FloatParts float64_unpack_canonical(float64 f, float_status *s) { - return canonicalize(float64_unpack_raw(f), &float64_params, s); + return sf_canonicalize(float64_unpack_raw(f), &float64_params, s); } static float64 float64_round_pack_canonical(FloatParts p, float_status *s) @@ -735,49 +1054,128 @@ float16 QEMU_FLATTEN float16_add(float16 a, float16 b, float_status *status) return float16_round_pack_canonical(pr, status); } -float32 QEMU_FLATTEN float32_add(float32 a, float32 b, float_status *status) +float16 QEMU_FLATTEN float16_sub(float16 a, float16 b, float_status *status) +{ + FloatParts pa = float16_unpack_canonical(a, status); + FloatParts pb = float16_unpack_canonical(b, status); + FloatParts pr = addsub_floats(pa, pb, true, status); + + return float16_round_pack_canonical(pr, status); +} + +static float32 QEMU_SOFTFLOAT_ATTR +soft_f32_addsub(float32 a, float32 b, bool subtract, float_status *status) { FloatParts pa = float32_unpack_canonical(a, status); FloatParts pb = float32_unpack_canonical(b, status); - FloatParts pr = addsub_floats(pa, pb, false, status); + FloatParts pr = addsub_floats(pa, pb, subtract, status); return float32_round_pack_canonical(pr, status); } -float64 QEMU_FLATTEN float64_add(float64 a, float64 b, float_status *status) +static inline float32 soft_f32_add(float32 a, float32 b, float_status *status) +{ + return soft_f32_addsub(a, b, false, status); +} + +static inline float32 soft_f32_sub(float32 a, float32 b, float_status *status) +{ + return soft_f32_addsub(a, b, true, status); +} + +static float64 QEMU_SOFTFLOAT_ATTR +soft_f64_addsub(float64 a, float64 b, bool subtract, float_status *status) { FloatParts pa = float64_unpack_canonical(a, status); FloatParts pb = float64_unpack_canonical(b, status); - FloatParts pr = addsub_floats(pa, pb, false, status); + FloatParts pr = addsub_floats(pa, pb, subtract, status); return float64_round_pack_canonical(pr, status); } -float16 QEMU_FLATTEN float16_sub(float16 a, float16 b, float_status *status) +static inline float64 soft_f64_add(float64 a, float64 b, float_status *status) { - FloatParts pa = float16_unpack_canonical(a, status); - FloatParts pb = float16_unpack_canonical(b, status); - FloatParts pr = addsub_floats(pa, pb, true, status); + return soft_f64_addsub(a, b, false, status); +} - return float16_round_pack_canonical(pr, status); +static inline float64 soft_f64_sub(float64 a, float64 b, float_status *status) +{ + return soft_f64_addsub(a, b, true, status); } -float32 QEMU_FLATTEN float32_sub(float32 a, float32 b, float_status *status) +static float hard_f32_add(float a, float b) { - FloatParts pa = float32_unpack_canonical(a, status); - FloatParts pb = float32_unpack_canonical(b, status); - FloatParts pr = addsub_floats(pa, pb, true, status); + return a + b; +} - return float32_round_pack_canonical(pr, status); +static float hard_f32_sub(float a, float b) +{ + return a - b; } -float64 QEMU_FLATTEN float64_sub(float64 a, float64 b, float_status *status) +static double hard_f64_add(double a, double b) { - FloatParts pa = float64_unpack_canonical(a, status); - FloatParts pb = float64_unpack_canonical(b, status); - FloatParts pr = addsub_floats(pa, pb, true, status); + return a + b; +} - return float64_round_pack_canonical(pr, status); +static double hard_f64_sub(double a, double b) +{ + return a - b; +} + +static bool f32_addsub_post(union_float32 a, union_float32 b) +{ + if (QEMU_HARDFLOAT_2F32_USE_FP) { + return !(fpclassify(a.h) == FP_ZERO && fpclassify(b.h) == FP_ZERO); + } + return !(float32_is_zero(a.s) && float32_is_zero(b.s)); +} + +static bool f64_addsub_post(union_float64 a, union_float64 b) +{ + if (QEMU_HARDFLOAT_2F64_USE_FP) { + return !(fpclassify(a.h) == FP_ZERO && fpclassify(b.h) == FP_ZERO); + } else { + return !(float64_is_zero(a.s) && float64_is_zero(b.s)); + } +} + +static float32 float32_addsub(float32 a, float32 b, float_status *s, + hard_f32_op2_fn hard, soft_f32_op2_fn soft) +{ + return float32_gen2(a, b, s, hard, soft, + f32_is_zon2, f32_addsub_post, NULL, NULL); +} + +static float64 float64_addsub(float64 a, float64 b, float_status *s, + hard_f64_op2_fn hard, soft_f64_op2_fn soft) +{ + return float64_gen2(a, b, s, hard, soft, + f64_is_zon2, f64_addsub_post, NULL, NULL); +} + +float32 QEMU_FLATTEN +float32_add(float32 a, float32 b, float_status *s) +{ + return float32_addsub(a, b, s, hard_f32_add, soft_f32_add); +} + +float32 QEMU_FLATTEN +float32_sub(float32 a, float32 b, float_status *s) +{ + return float32_addsub(a, b, s, hard_f32_sub, soft_f32_sub); +} + +float64 QEMU_FLATTEN +float64_add(float64 a, float64 b, float_status *s) +{ + return float64_addsub(a, b, s, hard_f64_add, soft_f64_add); +} + +float64 QEMU_FLATTEN +float64_sub(float64 a, float64 b, float_status *s) +{ + return float64_addsub(a, b, s, hard_f64_sub, soft_f64_sub); } /* @@ -838,7 +1236,8 @@ float16 QEMU_FLATTEN float16_mul(float16 a, float16 b, float_status *status) return float16_round_pack_canonical(pr, status); } -float32 QEMU_FLATTEN float32_mul(float32 a, float32 b, float_status *status) +static float32 QEMU_SOFTFLOAT_ATTR +soft_f32_mul(float32 a, float32 b, float_status *status) { FloatParts pa = float32_unpack_canonical(a, status); FloatParts pb = float32_unpack_canonical(b, status); @@ -847,7 +1246,8 @@ float32 QEMU_FLATTEN float32_mul(float32 a, float32 b, float_status *status) return float32_round_pack_canonical(pr, status); } -float64 QEMU_FLATTEN float64_mul(float64 a, float64 b, float_status *status) +static float64 QEMU_SOFTFLOAT_ATTR +soft_f64_mul(float64 a, float64 b, float_status *status) { FloatParts pa = float64_unpack_canonical(a, status); FloatParts pb = float64_unpack_canonical(b, status); @@ -856,6 +1256,54 @@ float64 QEMU_FLATTEN float64_mul(float64 a, float64 b, float_status *status) return float64_round_pack_canonical(pr, status); } +static float hard_f32_mul(float a, float b) +{ + return a * b; +} + +static double hard_f64_mul(double a, double b) +{ + return a * b; +} + +static bool f32_mul_fast_test(union_float32 a, union_float32 b) +{ + return float32_is_zero(a.s) || float32_is_zero(b.s); +} + +static bool f64_mul_fast_test(union_float64 a, union_float64 b) +{ + return float64_is_zero(a.s) || float64_is_zero(b.s); +} + +static float32 f32_mul_fast_op(float32 a, float32 b, float_status *s) +{ + bool signbit = float32_is_neg(a) ^ float32_is_neg(b); + + return float32_set_sign(float32_zero, signbit); +} + +static float64 f64_mul_fast_op(float64 a, float64 b, float_status *s) +{ + bool signbit = float64_is_neg(a) ^ float64_is_neg(b); + + return float64_set_sign(float64_zero, signbit); +} + +float32 QEMU_FLATTEN +float32_mul(float32 a, float32 b, float_status *s) +{ + return float32_gen2(a, b, s, hard_f32_mul, soft_f32_mul, + f32_is_zon2, NULL, f32_mul_fast_test, f32_mul_fast_op); +} + +float64 QEMU_FLATTEN +float64_mul(float64 a, float64 b, float_status *s) +{ + return float64_gen2(a, b, s, hard_f64_mul, soft_f64_mul, + f64_is_zon2, NULL, f64_mul_fast_test, f64_mul_fast_op); +} + /* * Returns the result of multiplying the floating-point values `a' and * `b' then adding 'c', with no intermediate rounding step after the @@ -1070,8 +1518,9 @@ float16 QEMU_FLATTEN float16_muladd(float16 a, float16 b, float16 c, return float16_round_pack_canonical(pr, status); } -float32 QEMU_FLATTEN float32_muladd(float32 a, float32 b, float32 c, - int flags, float_status *status) +static float32 QEMU_SOFTFLOAT_ATTR +soft_f32_muladd(float32 a, float32 b, float32 c, int flags, + float_status *status) { FloatParts pa = float32_unpack_canonical(a, status); FloatParts pb = float32_unpack_canonical(b, status); @@ -1081,8 +1530,9 @@ float32 QEMU_FLATTEN float32_muladd(float32 a, float32 b, float32 c, return float32_round_pack_canonical(pr, status); } -float64 QEMU_FLATTEN float64_muladd(float64 a, float64 b, float64 c, - int flags, float_status *status) +static float64 QEMU_SOFTFLOAT_ATTR +soft_f64_muladd(float64 a, float64 b, float64 c, int flags, + float_status *status) { FloatParts pa = float64_unpack_canonical(a, status); FloatParts pb = float64_unpack_canonical(b, status); @@ -1092,6 +1542,128 @@ float64 QEMU_FLATTEN float64_muladd(float64 a, float64 b, float64 c, return float64_round_pack_canonical(pr, status); } +float32 QEMU_FLATTEN +float32_muladd(float32 xa, float32 xb, float32 xc, int flags, float_status *s) +{ + union_float32 ua, ub, uc, ur; + + ua.s = xa; + ub.s = xb; + uc.s = xc; + + if (unlikely(!can_use_fpu(s))) { + goto soft; + } + if (unlikely(flags & float_muladd_halve_result)) { + goto soft; + } + + float32_input_flush3(&ua.s, &ub.s, &uc.s, s); + if (unlikely(!f32_is_zon3(ua, ub, uc))) { + goto soft; + } + /* + * When (a || b) == 0, there's no need to check for under/over flow, + * since we know the addend is (normal || 0) and the product is 0. + */ + if (float32_is_zero(ua.s) || float32_is_zero(ub.s)) { + union_float32 up; + bool prod_sign; + + prod_sign = float32_is_neg(ua.s) ^ float32_is_neg(ub.s); + prod_sign ^= !!(flags & float_muladd_negate_product); + up.s = float32_set_sign(float32_zero, prod_sign); + + if (flags & float_muladd_negate_c) { + uc.h = -uc.h; + } + ur.h = up.h + uc.h; + } else { + if (flags & float_muladd_negate_product) { + ua.h = -ua.h; + } + if (flags & float_muladd_negate_c) { + uc.h = -uc.h; + } + + ur.h = fmaf(ua.h, ub.h, uc.h); + + if (unlikely(f32_is_inf(ur))) { + s->float_exception_flags |= float_flag_overflow; + } else if (unlikely(fabsf(ur.h) <= FLT_MIN)) { + goto soft; + } + } + if (flags & float_muladd_negate_result) { + return float32_chs(ur.s); + } + return ur.s; + + soft: + return soft_f32_muladd(ua.s, ub.s, uc.s, flags, s); +} + +float64 QEMU_FLATTEN +float64_muladd(float64 xa, float64 xb, float64 xc, int flags, float_status *s) +{ + union_float64 ua, ub, uc, ur; + + ua.s = xa; + ub.s = xb; + uc.s = xc; + + if (unlikely(!can_use_fpu(s))) { + goto soft; + } + if (unlikely(flags & float_muladd_halve_result)) { + goto soft; + } + + float64_input_flush3(&ua.s, &ub.s, &uc.s, s); + if (unlikely(!f64_is_zon3(ua, ub, uc))) { + goto soft; + } + /* + * When (a || b) == 0, there's no need to check for under/over flow, + * since we know the addend is (normal || 0) and the product is 0. + */ + if (float64_is_zero(ua.s) || float64_is_zero(ub.s)) { + union_float64 up; + bool prod_sign; + + prod_sign = float64_is_neg(ua.s) ^ float64_is_neg(ub.s); + prod_sign ^= !!(flags & float_muladd_negate_product); + up.s = float64_set_sign(float64_zero, prod_sign); + + if (flags & float_muladd_negate_c) { + uc.h = -uc.h; + } + ur.h = up.h + uc.h; + } else { + if (flags & float_muladd_negate_product) { + ua.h = -ua.h; + } + if (flags & float_muladd_negate_c) { + uc.h = -uc.h; + } + + ur.h = fma(ua.h, ub.h, uc.h); + + if (unlikely(f64_is_inf(ur))) { + s->float_exception_flags |= float_flag_overflow; + } else if (unlikely(fabs(ur.h) <= FLT_MIN)) { + goto soft; + } + } + if (flags & float_muladd_negate_result) { + return float64_chs(ur.s); + } + return ur.s; + + soft: + return soft_f64_muladd(ua.s, ub.s, uc.s, flags, s); +} + /* * Returns the result of dividing the floating-point value `a' by the * corresponding value `b'. The operation is performed according to @@ -1180,7 +1752,8 @@ float16 float16_div(float16 a, float16 b, float_status *status) return float16_round_pack_canonical(pr, status); } -float32 float32_div(float32 a, float32 b, float_status *status) +static float32 QEMU_SOFTFLOAT_ATTR +soft_f32_div(float32 a, float32 b, float_status *status) { FloatParts pa = float32_unpack_canonical(a, status); FloatParts pb = float32_unpack_canonical(b, status); @@ -1189,7 +1762,8 @@ float32 float32_div(float32 a, float32 b, float_status *status) return float32_round_pack_canonical(pr, status); } -float64 float64_div(float64 a, float64 b, float_status *status) +static float64 QEMU_SOFTFLOAT_ATTR +soft_f64_div(float64 a, float64 b, float_status *status) { FloatParts pa = float64_unpack_canonical(a, status); FloatParts pb = float64_unpack_canonical(b, status); @@ -1198,6 +1772,64 @@ float64 float64_div(float64 a, float64 b, float_status *status) return float64_round_pack_canonical(pr, status); } +static float hard_f32_div(float a, float b) +{ + return a / b; +} + +static double hard_f64_div(double a, double b) +{ + return a / b; +} + +static bool f32_div_pre(union_float32 a, union_float32 b) +{ + if (QEMU_HARDFLOAT_2F32_USE_FP) { + return (fpclassify(a.h) == FP_NORMAL || fpclassify(a.h) == FP_ZERO) && + fpclassify(b.h) == FP_NORMAL; + } + return float32_is_zero_or_normal(a.s) && float32_is_normal(b.s); +} + +static bool f64_div_pre(union_float64 a, union_float64 b) +{ + if (QEMU_HARDFLOAT_2F64_USE_FP) { + return (fpclassify(a.h) == FP_NORMAL || fpclassify(a.h) == FP_ZERO) && + fpclassify(b.h) == FP_NORMAL; + } + return float64_is_zero_or_normal(a.s) && float64_is_normal(b.s); +} + +static bool f32_div_post(union_float32 a, union_float32 b) +{ + if (QEMU_HARDFLOAT_2F32_USE_FP) { + return fpclassify(a.h) != FP_ZERO; + } + return !float32_is_zero(a.s); +} + +static bool f64_div_post(union_float64 a, union_float64 b) +{ + if (QEMU_HARDFLOAT_2F64_USE_FP) { + return fpclassify(a.h) != FP_ZERO; + } + return !float64_is_zero(a.s); +} + +float32 QEMU_FLATTEN +float32_div(float32 a, float32 b, float_status *s) +{ + return float32_gen2(a, b, s, hard_f32_div, soft_f32_div, + f32_div_pre, f32_div_post, NULL, NULL); +} + +float64 QEMU_FLATTEN +float64_div(float64 a, float64 b, float_status *s) +{ + return float64_gen2(a, b, s, hard_f64_div, soft_f64_div, + f64_div_pre, f64_div_post, NULL, NULL); +} + /* * Float to Float conversions * @@ -2271,28 +2903,109 @@ static int compare_floats(FloatParts a, FloatParts b, bool is_quiet, } } -#define COMPARE(sz) \ -int float ## sz ## _compare(float ## sz a, float ## sz b, \ - float_status *s) \ +#define COMPARE(name, attr, sz) \ +static int attr \ +name(float ## sz a, float ## sz b, bool is_quiet, float_status *s) \ { \ FloatParts pa = float ## sz ## _unpack_canonical(a, s); \ FloatParts pb = float ## sz ## _unpack_canonical(b, s); \ - return compare_floats(pa, pb, false, s); \ -} \ -int float ## sz ## _compare_quiet(float ## sz a, float ## sz b, \ - float_status *s) \ -{ \ - FloatParts pa = float ## sz ## _unpack_canonical(a, s); \ - FloatParts pb = float ## sz ## _unpack_canonical(b, s); \ - return compare_floats(pa, pb, true, s); \ + return compare_floats(pa, pb, is_quiet, s); \ } -COMPARE(16) -COMPARE(32) -COMPARE(64) +COMPARE(soft_f16_compare, QEMU_FLATTEN, 16) +COMPARE(soft_f32_compare, QEMU_SOFTFLOAT_ATTR, 32) +COMPARE(soft_f64_compare, QEMU_SOFTFLOAT_ATTR, 64) #undef COMPARE +int float16_compare(float16 a, float16 b, float_status *s) +{ + return soft_f16_compare(a, b, false, s); +} + +int float16_compare_quiet(float16 a, float16 b, float_status *s) +{ + return soft_f16_compare(a, b, true, s); +} + +static int QEMU_FLATTEN +f32_compare(float32 xa, float32 xb, bool is_quiet, float_status *s) +{ + union_float32 ua, ub; + + ua.s = xa; + ub.s = xb; + + if (QEMU_NO_HARDFLOAT) { + goto soft; + } + + float32_input_flush2(&ua.s, &ub.s, s); + if (isgreaterequal(ua.h, ub.h)) { + if (isgreater(ua.h, ub.h)) { + return float_relation_greater; + } + return float_relation_equal; + } + if (likely(isless(ua.h, ub.h))) { + return float_relation_less; + } + /* The only condition remaining is unordered. + * Fall through to set flags. + */ + soft: + return soft_f32_compare(ua.s, ub.s, is_quiet, s); +} + +int float32_compare(float32 a, float32 b, float_status *s) +{ + return f32_compare(a, b, false, s); +} + +int float32_compare_quiet(float32 a, float32 b, float_status *s) +{ + return f32_compare(a, b, true, s); +} + +static int QEMU_FLATTEN +f64_compare(float64 xa, float64 xb, bool is_quiet, float_status *s) +{ + union_float64 ua, ub; + + ua.s = xa; + ub.s = xb; + + if (QEMU_NO_HARDFLOAT) { + goto soft; + } + + float64_input_flush2(&ua.s, &ub.s, s); + if (isgreaterequal(ua.h, ub.h)) { + if (isgreater(ua.h, ub.h)) { + return float_relation_greater; + } + return float_relation_equal; + } + if (likely(isless(ua.h, ub.h))) { + return float_relation_less; + } + /* The only condition remaining is unordered. + * Fall through to set flags. + */ + soft: + return soft_f64_compare(ua.s, ub.s, is_quiet, s); +} + +int float64_compare(float64 a, float64 b, float_status *s) +{ + return f64_compare(a, b, false, s); +} + +int float64_compare_quiet(float64 a, float64 b, float_status *s) +{ + return f64_compare(a, b, true, s); +} + /* Multiply A by 2 raised to the power N. */ static FloatParts scalbn_decomposed(FloatParts a, int n, float_status *s) { @@ -2412,20 +3125,76 @@ float16 QEMU_FLATTEN float16_sqrt(float16 a, float_status *status) return float16_round_pack_canonical(pr, status); } -float32 QEMU_FLATTEN float32_sqrt(float32 a, float_status *status) +static float32 QEMU_SOFTFLOAT_ATTR +soft_f32_sqrt(float32 a, float_status *status) { FloatParts pa = float32_unpack_canonical(a, status); FloatParts pr = sqrt_float(pa, status, &float32_params); return float32_round_pack_canonical(pr, status); } -float64 QEMU_FLATTEN float64_sqrt(float64 a, float_status *status) +static float64 QEMU_SOFTFLOAT_ATTR +soft_f64_sqrt(float64 a, float_status *status) { FloatParts pa = float64_unpack_canonical(a, status); FloatParts pr = sqrt_float(pa, status, &float64_params); return float64_round_pack_canonical(pr, status); } +float32 QEMU_FLATTEN float32_sqrt(float32 xa, float_status *s) +{ + union_float32 ua, ur; + + ua.s = xa; + if (unlikely(!can_use_fpu(s))) { + goto soft; + } + + float32_input_flush1(&ua.s, s); + if (QEMU_HARDFLOAT_1F32_USE_FP) { + if (unlikely(!(fpclassify(ua.h) == FP_NORMAL || + fpclassify(ua.h) == FP_ZERO) || + signbit(ua.h))) { + goto soft; + } + } else if (unlikely(!float32_is_zero_or_normal(ua.s) || + float32_is_neg(ua.s))) { + goto soft; + } + ur.h = sqrtf(ua.h); + return ur.s; + + soft: + return soft_f32_sqrt(ua.s, s); +} + +float64 QEMU_FLATTEN float64_sqrt(float64 xa, float_status *s) +{ + union_float64 ua, ur; + + ua.s = xa; + if (unlikely(!can_use_fpu(s))) { + goto soft; + } + + float64_input_flush1(&ua.s, s); + if (QEMU_HARDFLOAT_1F64_USE_FP) { + if (unlikely(!(fpclassify(ua.h) == FP_NORMAL || + fpclassify(ua.h) == FP_ZERO) || + signbit(ua.h))) { + goto soft; + } + } else if (unlikely(!float64_is_zero_or_normal(ua.s) || + float64_is_neg(ua.s))) { + goto soft; + } + ur.h = sqrt(ua.h); + return ur.s; + + soft: + return soft_f64_sqrt(ua.s, s); +} + /*---------------------------------------------------------------------------- | The pattern for a default generated NaN. *----------------------------------------------------------------------------*/ diff --git a/gitdm.config b/gitdm.config new file mode 100644 index 0000000000..7472d4b8be --- /dev/null +++ b/gitdm.config @@ -0,0 +1,50 @@ +# +# This is the gitdm configuration file for QEMU. +# +# It is to be used with LWN's git dataminer tool for generating +# reports about development activity in the QEMU repo. The LWN gitdm +# tool can be found at: +# +# git://git.lwn.net/gitdm.git +# +# A run to generate a report for the last year of activity would be +# +# git log --numstat --since "Last Year" | gitdm -n -l 10 +# + +# EmailAliases lets us cope with developers who use more +# than one address or have changed addresses. This duplicates some of +# the information in the existing .mailmap but in a slightly different +# form. +# +EmailAliases contrib/gitdm/aliases + +# +# EmailMap does the main work of mapping addresses onto +# employers. +# +EmailMap contrib/gitdm/domain-map + +# +# Use GroupMap to map a file full of addresses to the +# same employer. This is used for people that don't post from easily +# identifiable corporate emails. +# + +GroupMap contrib/gitdm/group-map-redhat Red Hat +GroupMap contrib/gitdm/group-map-wavecomp Wave Computing +GroupMap contrib/gitdm/group-map-cadence Cadence Design Systems +GroupMap contrib/gitdm/group-map-codeweavers CodeWeavers +GroupMap contrib/gitdm/group-map-ibm IBM + +# Also group together our prolific individual contributors +# and those working under academic auspices +GroupMap contrib/gitdm/group-map-individuals (None) +GroupMap contrib/gitdm/group-map-academics Academics (various) + +# +# +# Use FileTypeMap to map a file types to file names using regular +# regular expressions. +# +FileTypeMap contrib/gitdm/filetypes.txt diff --git a/include/exec/tb-hash.h b/include/exec/tb-hash.h index 0526c4f678..4f3a37d927 100644 --- a/include/exec/tb-hash.h +++ b/include/exec/tb-hash.h @@ -20,7 +20,7 @@ #ifndef EXEC_TB_HASH_H #define EXEC_TB_HASH_H -#include "exec/tb-hash-xx.h" +#include "qemu/xxhash.h" #ifdef CONFIG_SOFTMMU @@ -61,7 +61,7 @@ static inline uint32_t tb_hash_func(tb_page_addr_t phys_pc, target_ulong pc, uint32_t flags, uint32_t cf_mask, uint32_t trace_vcpu_dstate) { - return tb_hash_func7(phys_pc, pc, flags, cf_mask, trace_vcpu_dstate); + return qemu_xxhash7(phys_pc, pc, flags, cf_mask, trace_vcpu_dstate); } #endif diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h index 8fd9f9bbae..38a5e99cf3 100644 --- a/include/fpu/softfloat.h +++ b/include/fpu/softfloat.h @@ -464,6 +464,21 @@ static inline int float32_is_zero_or_denormal(float32 a) return (float32_val(a) & 0x7f800000) == 0; } +static inline bool float32_is_normal(float32 a) +{ + return ((float32_val(a) + 0x00800000) & 0x7fffffff) >= 0x01000000; +} + +static inline bool float32_is_denormal(float32 a) +{ + return float32_is_zero_or_denormal(a) && !float32_is_zero(a); +} + +static inline bool float32_is_zero_or_normal(float32 a) +{ + return float32_is_normal(a) || float32_is_zero(a); +} + static inline float32 float32_set_sign(float32 a, int sign) { return make_float32((float32_val(a) & 0x7fffffff) | (sign << 31)); @@ -605,6 +620,21 @@ static inline int float64_is_zero_or_denormal(float64 a) return (float64_val(a) & 0x7ff0000000000000LL) == 0; } +static inline bool float64_is_normal(float64 a) +{ + return ((float64_val(a) + (1ULL << 52)) & -1ULL >> 1) >= 1ULL << 53; +} + +static inline bool float64_is_denormal(float64 a) +{ + return float64_is_zero_or_denormal(a) && !float64_is_zero(a); +} + +static inline bool float64_is_zero_or_normal(float64 a) +{ + return float64_is_normal(a) || float64_is_zero(a); +} + static inline float64 float64_set_sign(float64 a, int sign) { return make_float64((float64_val(a) & 0x7fffffffffffffffULL) diff --git a/include/exec/tb-hash-xx.h b/include/qemu/xxhash.h index 747a9a612c..076f1f6054 100644 --- a/include/exec/tb-hash-xx.h +++ b/include/qemu/xxhash.h @@ -31,8 +31,8 @@ * - xxHash source repository : https://github.com/Cyan4973/xxHash */ -#ifndef EXEC_TB_HASH_XX_H -#define EXEC_TB_HASH_XX_H +#ifndef QEMU_XXHASH_H +#define QEMU_XXHASH_H #include "qemu/bitops.h" @@ -42,23 +42,23 @@ #define PRIME32_4 668265263U #define PRIME32_5 374761393U -#define TB_HASH_XX_SEED 1 +#define QEMU_XXHASH_SEED 1 /* * xxhash32, customized for input variables that are not guaranteed to be * contiguous in memory. */ static inline uint32_t -tb_hash_func7(uint64_t a0, uint64_t b0, uint32_t e, uint32_t f, uint32_t g) +qemu_xxhash7(uint64_t ab, uint64_t cd, uint32_t e, uint32_t f, uint32_t g) { - uint32_t v1 = TB_HASH_XX_SEED + PRIME32_1 + PRIME32_2; - uint32_t v2 = TB_HASH_XX_SEED + PRIME32_2; - uint32_t v3 = TB_HASH_XX_SEED + 0; - uint32_t v4 = TB_HASH_XX_SEED - PRIME32_1; - uint32_t a = a0 >> 32; - uint32_t b = a0; - uint32_t c = b0 >> 32; - uint32_t d = b0; + uint32_t v1 = QEMU_XXHASH_SEED + PRIME32_1 + PRIME32_2; + uint32_t v2 = QEMU_XXHASH_SEED + PRIME32_2; + uint32_t v3 = QEMU_XXHASH_SEED + 0; + uint32_t v4 = QEMU_XXHASH_SEED - PRIME32_1; + uint32_t a = ab; + uint32_t b = ab >> 32; + uint32_t c = cd; + uint32_t d = cd >> 32; uint32_t h32; v1 += a * PRIME32_2; @@ -98,4 +98,25 @@ tb_hash_func7(uint64_t a0, uint64_t b0, uint32_t e, uint32_t f, uint32_t g) return h32; } -#endif /* EXEC_TB_HASH_XX_H */ +static inline uint32_t qemu_xxhash2(uint64_t ab) +{ + return qemu_xxhash7(ab, 0, 0, 0, 0); +} + +static inline uint32_t qemu_xxhash4(uint64_t ab, uint64_t cd) +{ + return qemu_xxhash7(ab, cd, 0, 0, 0); +} + +static inline uint32_t qemu_xxhash5(uint64_t ab, uint64_t cd, uint32_t e) +{ + return qemu_xxhash7(ab, cd, e, 0, 0); +} + +static inline uint32_t qemu_xxhash6(uint64_t ab, uint64_t cd, uint32_t e, + uint32_t f) +{ + return qemu_xxhash7(ab, cd, e, f, 0); +} + +#endif /* QEMU_XXHASH_H */ diff --git a/target/tricore/fpu_helper.c b/target/tricore/fpu_helper.c index df162902d6..31df462e4a 100644 --- a/target/tricore/fpu_helper.c +++ b/target/tricore/fpu_helper.c @@ -44,11 +44,6 @@ static inline uint8_t f_get_excp_flags(CPUTriCoreState *env) | float_flag_inexact); } -static inline bool f_is_denormal(float32 arg) -{ - return float32_is_zero_or_denormal(arg) && !float32_is_zero(arg); -} - static inline float32 f_maddsub_nan_result(float32 arg1, float32 arg2, float32 arg3, float32 result, uint32_t muladd_negate_c) @@ -260,8 +255,8 @@ uint32_t helper_fcmp(CPUTriCoreState *env, uint32_t r1, uint32_t r2) set_flush_inputs_to_zero(0, &env->fp_status); result = 1 << (float32_compare_quiet(arg1, arg2, &env->fp_status) + 1); - result |= f_is_denormal(arg1) << 4; - result |= f_is_denormal(arg2) << 5; + result |= float32_is_denormal(arg1) << 4; + result |= float32_is_denormal(arg2) << 5; flags = f_get_excp_flags(env); if (flags) { diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h index 9aea1d1771..f966a4fcb3 100644 --- a/tcg/aarch64/tcg-target.h +++ b/tcg/aarch64/tcg-target.h @@ -137,6 +137,7 @@ typedef enum { #define TCG_TARGET_HAS_mul_vec 1 #define TCG_TARGET_DEFAULT_MO (0) +#define TCG_TARGET_HAS_MEMORY_BSWAP 1 static inline void flush_icache_range(uintptr_t start, uintptr_t stop) { diff --git a/tcg/aarch64/tcg-target.inc.c b/tcg/aarch64/tcg-target.inc.c index 083592a4d7..0562e0aa40 100644 --- a/tcg/aarch64/tcg-target.inc.c +++ b/tcg/aarch64/tcg-target.inc.c @@ -78,48 +78,40 @@ static const int tcg_target_call_oarg_regs[1] = { #define TCG_REG_GUEST_BASE TCG_REG_X28 #endif -static inline void reloc_pc26(tcg_insn_unit *code_ptr, tcg_insn_unit *target) +static inline bool reloc_pc26(tcg_insn_unit *code_ptr, tcg_insn_unit *target) { ptrdiff_t offset = target - code_ptr; - tcg_debug_assert(offset == sextract64(offset, 0, 26)); - /* read instruction, mask away previous PC_REL26 parameter contents, - set the proper offset, then write back the instruction. */ - *code_ptr = deposit32(*code_ptr, 0, 26, offset); -} - -static inline void reloc_pc26_atomic(tcg_insn_unit *code_ptr, - tcg_insn_unit *target) -{ - ptrdiff_t offset = target - code_ptr; - tcg_insn_unit insn; - tcg_debug_assert(offset == sextract64(offset, 0, 26)); - /* read instruction, mask away previous PC_REL26 parameter contents, - set the proper offset, then write back the instruction. */ - insn = atomic_read(code_ptr); - atomic_set(code_ptr, deposit32(insn, 0, 26, offset)); + if (offset == sextract64(offset, 0, 26)) { + /* read instruction, mask away previous PC_REL26 parameter contents, + set the proper offset, then write back the instruction. */ + *code_ptr = deposit32(*code_ptr, 0, 26, offset); + return true; + } + return false; } -static inline void reloc_pc19(tcg_insn_unit *code_ptr, tcg_insn_unit *target) +static inline bool reloc_pc19(tcg_insn_unit *code_ptr, tcg_insn_unit *target) { ptrdiff_t offset = target - code_ptr; - tcg_debug_assert(offset == sextract64(offset, 0, 19)); - *code_ptr = deposit32(*code_ptr, 5, 19, offset); + if (offset == sextract64(offset, 0, 19)) { + *code_ptr = deposit32(*code_ptr, 5, 19, offset); + return true; + } + return false; } -static inline void patch_reloc(tcg_insn_unit *code_ptr, int type, +static inline bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend) { tcg_debug_assert(addend == 0); switch (type) { case R_AARCH64_JUMP26: case R_AARCH64_CALL26: - reloc_pc26(code_ptr, (tcg_insn_unit *)value); - break; + return reloc_pc26(code_ptr, (tcg_insn_unit *)value); case R_AARCH64_CONDBR19: - reloc_pc19(code_ptr, (tcg_insn_unit *)value); - break; + return reloc_pc19(code_ptr, (tcg_insn_unit *)value); default: - tcg_abort(); + g_assert_not_reached(); } } @@ -1141,23 +1133,6 @@ static inline void tcg_out_goto_long(TCGContext *s, tcg_insn_unit *target) } } -static inline void tcg_out_goto_noaddr(TCGContext *s) -{ - /* We pay attention here to not modify the branch target by reading from - the buffer. This ensure that caches and memory are kept coherent during - retranslation. Mask away possible garbage in the high bits for the - first translation, while keeping the offset bits for retranslation. */ - uint32_t old = tcg_in32(s); - tcg_out_insn(s, 3206, B, old); -} - -static inline void tcg_out_goto_cond_noaddr(TCGContext *s, TCGCond c) -{ - /* See comments in tcg_out_goto_noaddr. */ - uint32_t old = tcg_in32(s) >> 5; - tcg_out_insn(s, 3202, B_C, c, old); -} - static inline void tcg_out_callr(TCGContext *s, TCGReg reg) { tcg_out_insn(s, 3207, BLR, reg); @@ -1204,7 +1179,7 @@ static inline void tcg_out_goto_label(TCGContext *s, TCGLabel *l) { if (!l->has_value) { tcg_out_reloc(s, s->code_ptr, R_AARCH64_JUMP26, l, 0); - tcg_out_goto_noaddr(s); + tcg_out_insn(s, 3206, B, 0); } else { tcg_out_goto(s, l->u.value_ptr); } @@ -1415,7 +1390,8 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) TCGMemOp opc = get_memop(oi); TCGMemOp size = opc & MO_SIZE; - reloc_pc19(lb->label_ptr[0], s->code_ptr); + bool ok = reloc_pc19(lb->label_ptr[0], s->code_ptr); + tcg_debug_assert(ok); tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_X0, TCG_AREG0); tcg_out_mov(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg); @@ -1437,7 +1413,8 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) TCGMemOp opc = get_memop(oi); TCGMemOp size = opc & MO_SIZE; - reloc_pc19(lb->label_ptr[0], s->code_ptr); + bool ok = reloc_pc19(lb->label_ptr[0], s->code_ptr); + tcg_debug_assert(ok); tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_X0, TCG_AREG0); tcg_out_mov(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg); @@ -1535,7 +1512,7 @@ static void tcg_out_tlb_read(TCGContext *s, TCGReg addr_reg, TCGMemOp opc, /* If not equal, we jump to the slow path. */ *label_ptr = s->code_ptr; - tcg_out_goto_cond_noaddr(s, TCG_COND_NE); + tcg_out_insn(s, 3202, B_C, TCG_COND_NE, 0); } #endif /* CONFIG_SOFTMMU */ diff --git a/tcg/arm/tcg-target.h b/tcg/arm/tcg-target.h index 94b3578c55..16172f73a3 100644 --- a/tcg/arm/tcg-target.h +++ b/tcg/arm/tcg-target.h @@ -131,6 +131,7 @@ enum { }; #define TCG_TARGET_DEFAULT_MO (0) +#define TCG_TARGET_HAS_MEMORY_BSWAP 1 static inline void flush_icache_range(uintptr_t start, uintptr_t stop) { diff --git a/tcg/arm/tcg-target.inc.c b/tcg/arm/tcg-target.inc.c index e1fbf465cb..49f57d655e 100644 --- a/tcg/arm/tcg-target.inc.c +++ b/tcg/arm/tcg-target.inc.c @@ -187,27 +187,23 @@ static const uint8_t tcg_cond_to_arm_cond[] = { [TCG_COND_GTU] = COND_HI, }; -static inline void reloc_pc24(tcg_insn_unit *code_ptr, tcg_insn_unit *target) +static inline bool reloc_pc24(tcg_insn_unit *code_ptr, tcg_insn_unit *target) { ptrdiff_t offset = (tcg_ptr_byte_diff(target, code_ptr) - 8) >> 2; - *code_ptr = (*code_ptr & ~0xffffff) | (offset & 0xffffff); -} - -static inline void reloc_pc24_atomic(tcg_insn_unit *code_ptr, tcg_insn_unit *target) -{ - ptrdiff_t offset = (tcg_ptr_byte_diff(target, code_ptr) - 8) >> 2; - tcg_insn_unit insn = atomic_read(code_ptr); - tcg_debug_assert(offset == sextract32(offset, 0, 24)); - atomic_set(code_ptr, deposit32(insn, 0, 24, offset)); + if (offset == sextract32(offset, 0, 24)) { + *code_ptr = (*code_ptr & ~0xffffff) | (offset & 0xffffff); + return true; + } + return false; } -static void patch_reloc(tcg_insn_unit *code_ptr, int type, +static bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend) { tcg_debug_assert(addend == 0); if (type == R_ARM_PC24) { - reloc_pc24(code_ptr, (tcg_insn_unit *)value); + return reloc_pc24(code_ptr, (tcg_insn_unit *)value); } else if (type == R_ARM_PC13) { intptr_t diff = value - (uintptr_t)(code_ptr + 2); tcg_insn_unit insn = *code_ptr; @@ -221,7 +217,11 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, } else { int rd = extract32(insn, 12, 4); int rt = rd == TCG_REG_PC ? TCG_REG_TMP : rd; - assert(diff >= 0x1000 && diff < 0x100000); + + if (diff < 0x1000 || diff >= 0x100000) { + return false; + } + /* add rt, pc, #high */ *code_ptr++ = ((insn & 0xf0000000) | (1 << 25) | ARITH_ADD | (TCG_REG_PC << 16) | (rt << 12) @@ -237,6 +237,7 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, } else { g_assert_not_reached(); } + return true; } #define TCG_CT_CONST_ARM 0x100 @@ -374,22 +375,6 @@ static inline void tcg_out_b(TCGContext *s, int cond, int32_t offset) (((offset - 8) >> 2) & 0x00ffffff)); } -static inline void tcg_out_b_noaddr(TCGContext *s, int cond) -{ - /* We pay attention here to not modify the branch target by masking - the corresponding bytes. This ensure that caches and memory are - kept coherent during retranslation. */ - tcg_out32(s, deposit32(*s->code_ptr, 24, 8, (cond << 4) | 0x0a)); -} - -static inline void tcg_out_bl_noaddr(TCGContext *s, int cond) -{ - /* We pay attention here to not modify the branch target by masking - the corresponding bytes. This ensure that caches and memory are - kept coherent during retranslation. */ - tcg_out32(s, deposit32(*s->code_ptr, 24, 8, (cond << 4) | 0x0b)); -} - static inline void tcg_out_bl(TCGContext *s, int cond, int32_t offset) { tcg_out32(s, (cond << 28) | 0x0b000000 | @@ -1090,7 +1075,7 @@ static inline void tcg_out_goto_label(TCGContext *s, int cond, TCGLabel *l) tcg_out_goto(s, cond, l->u.value_ptr); } else { tcg_out_reloc(s, s->code_ptr, R_ARM_PC24, l, 0); - tcg_out_b_noaddr(s, cond); + tcg_out_b(s, cond, 0); } } @@ -1395,7 +1380,8 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) TCGMemOp opc = get_memop(oi); void *func; - reloc_pc24(lb->label_ptr[0], s->code_ptr); + bool ok = reloc_pc24(lb->label_ptr[0], s->code_ptr); + tcg_debug_assert(ok); argreg = tcg_out_arg_reg32(s, TCG_REG_R0, TCG_AREG0); if (TARGET_LONG_BITS == 64) { @@ -1455,7 +1441,8 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) TCGMemOpIdx oi = lb->oi; TCGMemOp opc = get_memop(oi); - reloc_pc24(lb->label_ptr[0], s->code_ptr); + bool ok = reloc_pc24(lb->label_ptr[0], s->code_ptr); + tcg_debug_assert(ok); argreg = TCG_REG_R0; argreg = tcg_out_arg_reg32(s, argreg, TCG_AREG0); @@ -1636,7 +1623,7 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64) /* This a conditional BL only to load a pointer within this opcode into LR for the slow path. We will not be using the value for a tail call. */ label_ptr = s->code_ptr; - tcg_out_bl_noaddr(s, COND_NE); + tcg_out_bl(s, COND_NE, 0); tcg_out_qemu_ld_index(s, opc, datalo, datahi, addrlo, addend); @@ -1768,7 +1755,7 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64) /* The conditional call must come last, as we're going to return here. */ label_ptr = s->code_ptr; - tcg_out_bl_noaddr(s, COND_NE); + tcg_out_bl(s, COND_NE, 0); add_qemu_ldst_label(s, false, oi, datalo, datahi, addrlo, addrhi, s->code_ptr, label_ptr); diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h index 9fdf37f23c..f378d29568 100644 --- a/tcg/i386/tcg-target.h +++ b/tcg/i386/tcg-target.h @@ -84,10 +84,12 @@ typedef enum { TCG_REG_RBP = TCG_REG_EBP, TCG_REG_RSI = TCG_REG_ESI, TCG_REG_RDI = TCG_REG_EDI, + + TCG_AREG0 = TCG_REG_EBP, + TCG_REG_CALL_STACK = TCG_REG_ESP } TCGReg; /* used for function call generation */ -#define TCG_REG_CALL_STACK TCG_REG_ESP #define TCG_TARGET_STACK_ALIGN 16 #if defined(_WIN64) #define TCG_TARGET_CALL_STACK_OFFSET 32 @@ -133,8 +135,9 @@ extern bool have_avx2; #define TCG_TARGET_HAS_direct_jump 1 #if TCG_TARGET_REG_BITS == 64 -#define TCG_TARGET_HAS_extrl_i64_i32 0 -#define TCG_TARGET_HAS_extrh_i64_i32 0 +/* Keep target addresses zero-extended in a register. */ +#define TCG_TARGET_HAS_extrl_i64_i32 (TARGET_LONG_BITS == 32) +#define TCG_TARGET_HAS_extrh_i64_i32 (TARGET_LONG_BITS == 32) #define TCG_TARGET_HAS_div2_i64 1 #define TCG_TARGET_HAS_rot_i64 1 #define TCG_TARGET_HAS_ext8s_i64 1 @@ -194,12 +197,6 @@ extern bool have_avx2; #define TCG_TARGET_extract_i64_valid(ofs, len) \ (((ofs) == 8 && (len) == 8) || ((ofs) + (len)) == 32) -#if TCG_TARGET_REG_BITS == 64 -# define TCG_AREG0 TCG_REG_R14 -#else -# define TCG_AREG0 TCG_REG_EBP -#endif - static inline void flush_icache_range(uintptr_t start, uintptr_t stop) { } @@ -223,6 +220,8 @@ static inline void tb_target_set_jmp_target(uintptr_t tc_ptr, #define TCG_TARGET_DEFAULT_MO (TCG_MO_ALL & ~TCG_MO_ST_LD) +#define TCG_TARGET_HAS_MEMORY_BSWAP 1 + #ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS #endif diff --git a/tcg/i386/tcg-target.inc.c b/tcg/i386/tcg-target.inc.c index 436195894b..c21c3272f2 100644 --- a/tcg/i386/tcg-target.inc.c +++ b/tcg/i386/tcg-target.inc.c @@ -167,7 +167,7 @@ static bool have_lzcnt; static tcg_insn_unit *tb_ret_addr; -static void patch_reloc(tcg_insn_unit *code_ptr, int type, +static bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend) { value += addend; @@ -175,7 +175,7 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, case R_386_PC32: value -= (uintptr_t)code_ptr; if (value != (int32_t)value) { - tcg_abort(); + return false; } /* FALLTHRU */ case R_386_32: @@ -184,13 +184,14 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, case R_386_PC8: value -= (uintptr_t)code_ptr; if (value != (int8_t)value) { - tcg_abort(); + return false; } tcg_patch8(code_ptr, value); break; default: tcg_abort(); } + return true; } #if TCG_TARGET_REG_BITS == 64 @@ -308,13 +309,11 @@ static inline int tcg_target_const_match(tcg_target_long val, TCGType type, #define P_EXT38 0x200 /* 0x0f 0x38 opcode prefix */ #define P_DATA16 0x400 /* 0x66 opcode prefix */ #if TCG_TARGET_REG_BITS == 64 -# define P_ADDR32 0x800 /* 0x67 opcode prefix */ # define P_REXW 0x1000 /* Set REX.W = 1 */ # define P_REXB_R 0x2000 /* REG field as byte register */ # define P_REXB_RM 0x4000 /* R/M field as byte register */ # define P_GS 0x8000 /* gs segment override */ #else -# define P_ADDR32 0 # define P_REXW 0 # define P_REXB_R 0 # define P_REXB_RM 0 @@ -527,9 +526,6 @@ static void tcg_out_opc(TCGContext *s, int opc, int r, int rm, int x) tcg_debug_assert((opc & P_REXW) == 0); tcg_out8(s, 0x66); } - if (opc & P_ADDR32) { - tcg_out8(s, 0x67); - } if (opc & P_SIMDF3) { tcg_out8(s, 0xf3); } else if (opc & P_SIMDF2) { @@ -1658,11 +1654,7 @@ static inline void tcg_out_tlb_load(TCGContext *s, TCGReg addrlo, TCGReg addrhi, tcg_out_modrm_offset(s, OPC_CMP_GvEv + trexw, r1, r0, 0); /* Prepare for both the fast path add of the tlb addend, and the slow - path function argument setup. There are two cases worth note: - For 32-bit guest and x86_64 host, MOVL zero-extends the guest address - before the fastpath ADDQ below. For 64-bit guest and x32 host, MOVQ - copies the entire guest address for the slow path, while truncation - for the 32-bit host happens with the fastpath ADDL below. */ + path function argument setup. */ tcg_out_mov(s, ttype, r1, addrlo); /* jne slow_path */ @@ -1691,7 +1683,8 @@ static inline void tcg_out_tlb_load(TCGContext *s, TCGReg addrlo, TCGReg addrhi, * Record the context of a call to the out of line helper code for the slow path * for a load or store, so that we can later generate the correct helper code */ -static void add_qemu_ldst_label(TCGContext *s, bool is_ld, TCGMemOpIdx oi, +static void add_qemu_ldst_label(TCGContext *s, bool is_ld, bool is_64, + TCGMemOpIdx oi, TCGReg datalo, TCGReg datahi, TCGReg addrlo, TCGReg addrhi, tcg_insn_unit *raddr, @@ -1701,6 +1694,7 @@ static void add_qemu_ldst_label(TCGContext *s, bool is_ld, TCGMemOpIdx oi, label->is_ld = is_ld; label->oi = oi; + label->type = is_64 ? TCG_TYPE_I64 : TCG_TYPE_I32; label->datalo_reg = datalo; label->datahi_reg = datahi; label->addrlo_reg = addrlo; @@ -1721,6 +1715,7 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) TCGMemOp opc = get_memop(oi); TCGReg data_reg; tcg_insn_unit **label_ptr = &l->label_ptr[0]; + int rexw = (l->type == TCG_TYPE_I64 ? P_REXW : 0); /* resolve label address */ tcg_patch32(label_ptr[0], s->code_ptr - label_ptr[0] - 4); @@ -1759,10 +1754,10 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) data_reg = l->datalo_reg; switch (opc & MO_SSIZE) { case MO_SB: - tcg_out_ext8s(s, data_reg, TCG_REG_EAX, P_REXW); + tcg_out_ext8s(s, data_reg, TCG_REG_EAX, rexw); break; case MO_SW: - tcg_out_ext16s(s, data_reg, TCG_REG_EAX, P_REXW); + tcg_out_ext16s(s, data_reg, TCG_REG_EAX, rexw); break; #if TCG_TARGET_REG_BITS == 64 case MO_SL: @@ -1862,30 +1857,49 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) tcg_out_push(s, retaddr); tcg_out_jmp(s, qemu_st_helpers[opc & (MO_BSWAP | MO_SIZE)]); } -#elif defined(__x86_64__) && defined(__linux__) -# include <asm/prctl.h> -# include <sys/prctl.h> - +#elif TCG_TARGET_REG_BITS == 32 +# define x86_guest_base_seg 0 +# define x86_guest_base_index -1 +# define x86_guest_base_offset guest_base +#else +static int x86_guest_base_seg; +static int x86_guest_base_index = -1; +static int32_t x86_guest_base_offset; +# if defined(__x86_64__) && defined(__linux__) +# include <asm/prctl.h> +# include <sys/prctl.h> int arch_prctl(int code, unsigned long addr); - -static int guest_base_flags; -static inline void setup_guest_base_seg(void) +static inline int setup_guest_base_seg(void) { if (arch_prctl(ARCH_SET_GS, guest_base) == 0) { - guest_base_flags = P_GS; + return P_GS; } + return 0; } -#else -# define guest_base_flags 0 -static inline void setup_guest_base_seg(void) { } +# elif defined (__FreeBSD__) || defined (__FreeBSD_kernel__) +# include <machine/sysarch.h> +static inline int setup_guest_base_seg(void) +{ + if (sysarch(AMD64_SET_GSBASE, &guest_base) == 0) { + return P_GS; + } + return 0; +} +# else +static inline int setup_guest_base_seg(void) +{ + return 0; +} +# endif #endif /* SOFTMMU */ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, TCGReg base, int index, intptr_t ofs, - int seg, TCGMemOp memop) + int seg, bool is64, TCGMemOp memop) { const TCGMemOp real_bswap = memop & MO_BSWAP; TCGMemOp bswap = real_bswap; + int rexw = is64 * P_REXW; int movop = OPC_MOVL_GvEv; if (have_movbe && real_bswap) { @@ -1899,7 +1913,7 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, base, index, 0, ofs); break; case MO_SB: - tcg_out_modrm_sib_offset(s, OPC_MOVSBL + P_REXW + seg, datalo, + tcg_out_modrm_sib_offset(s, OPC_MOVSBL + rexw + seg, datalo, base, index, 0, ofs); break; case MO_UW: @@ -1919,9 +1933,9 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, base, index, 0, ofs); tcg_out_rolw_8(s, datalo); } - tcg_out_modrm(s, OPC_MOVSWL + P_REXW, datalo, datalo); + tcg_out_modrm(s, OPC_MOVSWL + rexw, datalo, datalo); } else { - tcg_out_modrm_sib_offset(s, OPC_MOVSWL + P_REXW + seg, + tcg_out_modrm_sib_offset(s, OPC_MOVSWL + rexw + seg, datalo, base, index, 0, ofs); } break; @@ -2009,49 +2023,21 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64) label_ptr, offsetof(CPUTLBEntry, addr_read)); /* TLB Hit. */ - tcg_out_qemu_ld_direct(s, datalo, datahi, TCG_REG_L1, -1, 0, 0, opc); + tcg_out_qemu_ld_direct(s, datalo, datahi, TCG_REG_L1, -1, 0, 0, is64, opc); /* Record the current context of a load into ldst label */ - add_qemu_ldst_label(s, true, oi, datalo, datahi, addrlo, addrhi, + add_qemu_ldst_label(s, true, is64, oi, datalo, datahi, addrlo, addrhi, s->code_ptr, label_ptr); #else - { - int32_t offset = guest_base; - TCGReg base = addrlo; - int index = -1; - int seg = 0; - - /* For a 32-bit guest, the high 32 bits may contain garbage. - We can do this with the ADDR32 prefix if we're not using - a guest base, or when using segmentation. Otherwise we - need to zero-extend manually. */ - if (guest_base == 0 || guest_base_flags) { - seg = guest_base_flags; - offset = 0; - if (TCG_TARGET_REG_BITS > TARGET_LONG_BITS) { - seg |= P_ADDR32; - } - } else if (TCG_TARGET_REG_BITS == 64) { - if (TARGET_LONG_BITS == 32) { - tcg_out_ext32u(s, TCG_REG_L0, base); - base = TCG_REG_L0; - } - if (offset != guest_base) { - tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_L1, guest_base); - index = TCG_REG_L1; - offset = 0; - } - } - - tcg_out_qemu_ld_direct(s, datalo, datahi, - base, index, offset, seg, opc); - } + tcg_out_qemu_ld_direct(s, datalo, datahi, addrlo, x86_guest_base_index, + x86_guest_base_offset, x86_guest_base_seg, + is64, opc); #endif } static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, - TCGReg base, intptr_t ofs, int seg, - TCGMemOp memop) + TCGReg base, int index, intptr_t ofs, + int seg, TCGMemOp memop) { /* ??? Ideally we wouldn't need a scratch register. For user-only, we could perform the bswap twice to restore the original value @@ -2075,8 +2061,8 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, tcg_out_mov(s, TCG_TYPE_I32, scratch, datalo); datalo = scratch; } - tcg_out_modrm_offset(s, OPC_MOVB_EvGv + P_REXB_R + seg, - datalo, base, ofs); + tcg_out_modrm_sib_offset(s, OPC_MOVB_EvGv + P_REXB_R + seg, + datalo, base, index, 0, ofs); break; case MO_16: if (bswap) { @@ -2084,7 +2070,8 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, tcg_out_rolw_8(s, scratch); datalo = scratch; } - tcg_out_modrm_offset(s, movop + P_DATA16 + seg, datalo, base, ofs); + tcg_out_modrm_sib_offset(s, movop + P_DATA16 + seg, datalo, + base, index, 0, ofs); break; case MO_32: if (bswap) { @@ -2092,7 +2079,7 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, tcg_out_bswap32(s, scratch); datalo = scratch; } - tcg_out_modrm_offset(s, movop + seg, datalo, base, ofs); + tcg_out_modrm_sib_offset(s, movop + seg, datalo, base, index, 0, ofs); break; case MO_64: if (TCG_TARGET_REG_BITS == 64) { @@ -2101,22 +2088,27 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, tcg_out_bswap64(s, scratch); datalo = scratch; } - tcg_out_modrm_offset(s, movop + P_REXW + seg, datalo, base, ofs); + tcg_out_modrm_sib_offset(s, movop + P_REXW + seg, datalo, + base, index, 0, ofs); } else if (bswap) { tcg_out_mov(s, TCG_TYPE_I32, scratch, datahi); tcg_out_bswap32(s, scratch); - tcg_out_modrm_offset(s, OPC_MOVL_EvGv + seg, scratch, base, ofs); + tcg_out_modrm_sib_offset(s, OPC_MOVL_EvGv + seg, scratch, + base, index, 0, ofs); tcg_out_mov(s, TCG_TYPE_I32, scratch, datalo); tcg_out_bswap32(s, scratch); - tcg_out_modrm_offset(s, OPC_MOVL_EvGv + seg, scratch, base, ofs+4); + tcg_out_modrm_sib_offset(s, OPC_MOVL_EvGv + seg, scratch, + base, index, 0, ofs + 4); } else { if (real_bswap) { int t = datalo; datalo = datahi; datahi = t; } - tcg_out_modrm_offset(s, movop + seg, datalo, base, ofs); - tcg_out_modrm_offset(s, movop + seg, datahi, base, ofs+4); + tcg_out_modrm_sib_offset(s, movop + seg, datalo, + base, index, 0, ofs); + tcg_out_modrm_sib_offset(s, movop + seg, datahi, + base, index, 0, ofs + 4); } break; default: @@ -2149,44 +2141,14 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64) label_ptr, offsetof(CPUTLBEntry, addr_write)); /* TLB Hit. */ - tcg_out_qemu_st_direct(s, datalo, datahi, TCG_REG_L1, 0, 0, opc); + tcg_out_qemu_st_direct(s, datalo, datahi, TCG_REG_L1, -1, 0, 0, opc); /* Record the current context of a store into ldst label */ - add_qemu_ldst_label(s, false, oi, datalo, datahi, addrlo, addrhi, + add_qemu_ldst_label(s, false, is64, oi, datalo, datahi, addrlo, addrhi, s->code_ptr, label_ptr); #else - { - int32_t offset = guest_base; - TCGReg base = addrlo; - int seg = 0; - - /* See comment in tcg_out_qemu_ld re zero-extension of addrlo. */ - if (guest_base == 0 || guest_base_flags) { - seg = guest_base_flags; - offset = 0; - if (TCG_TARGET_REG_BITS > TARGET_LONG_BITS) { - seg |= P_ADDR32; - } - } else if (TCG_TARGET_REG_BITS == 64) { - /* ??? Note that we can't use the same SIB addressing scheme - as for loads, since we require L0 free for bswap. */ - if (offset != guest_base) { - if (TARGET_LONG_BITS == 32) { - tcg_out_ext32u(s, TCG_REG_L0, base); - base = TCG_REG_L0; - } - tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_L1, guest_base); - tgen_arithr(s, ARITH_ADD + P_REXW, TCG_REG_L1, base); - base = TCG_REG_L1; - offset = 0; - } else if (TARGET_LONG_BITS == 32) { - tcg_out_ext32u(s, TCG_REG_L1, base); - base = TCG_REG_L1; - } - } - - tcg_out_qemu_st_direct(s, datalo, datahi, base, offset, seg, opc); - } + tcg_out_qemu_st_direct(s, datalo, datahi, addrlo, x86_guest_base_index, + x86_guest_base_offset, x86_guest_base_seg, opc); #endif } @@ -2544,12 +2506,16 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, break; case INDEX_op_extu_i32_i64: case INDEX_op_ext32u_i64: + case INDEX_op_extrl_i64_i32: tcg_out_ext32u(s, a0, a1); break; case INDEX_op_ext_i32_i64: case INDEX_op_ext32s_i64: tcg_out_ext32s(s, a0, a1); break; + case INDEX_op_extrh_i64_i32: + tcg_out_shifti(s, SHIFT_SHR + P_REXW, a0, 32); + break; #endif OP_32_64(deposit): @@ -2913,6 +2879,7 @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op) case INDEX_op_neg_i64: case INDEX_op_not_i32: case INDEX_op_not_i64: + case INDEX_op_extrh_i64_i32: return &r_0; case INDEX_op_ext8s_i32: @@ -2928,6 +2895,7 @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op) case INDEX_op_ext32u_i64: case INDEX_op_ext_i32_i64: case INDEX_op_extu_i32_i64: + case INDEX_op_extrl_i64_i32: case INDEX_op_extract_i32: case INDEX_op_extract_i64: case INDEX_op_sextract_i32: @@ -3427,6 +3395,21 @@ static void tcg_target_qemu_prologue(TCGContext *s) (ARRAY_SIZE(tcg_target_callee_save_regs) + 2) * 4 + stack_addend); #else +# if !defined(CONFIG_SOFTMMU) && TCG_TARGET_REG_BITS == 64 + if (guest_base) { + int seg = setup_guest_base_seg(); + if (seg != 0) { + x86_guest_base_seg = seg; + } else if (guest_base == (int32_t)guest_base) { + x86_guest_base_offset = guest_base; + } else { + /* Choose R12 because, as a base, it requires a SIB byte. */ + x86_guest_base_index = TCG_REG_R12; + tcg_out_mov(s, TCG_TYPE_PTR, x86_guest_base_index, guest_base); + tcg_regset_set_reg(s->reserved_regs, x86_guest_base_index); + } + } +# endif tcg_out_mov(s, TCG_TYPE_PTR, TCG_AREG0, tcg_target_call_iarg_regs[0]); tcg_out_addi(s, TCG_REG_ESP, -stack_addend); /* jmp *tb. */ @@ -3452,13 +3435,6 @@ static void tcg_target_qemu_prologue(TCGContext *s) tcg_out_pop(s, tcg_target_callee_save_regs[i]); } tcg_out_opc(s, OPC_RET, 0, 0, 0); - -#if !defined(CONFIG_SOFTMMU) - /* Try to set up a segment register to point to guest_base. */ - if (guest_base) { - setup_guest_base_seg(); - } -#endif } static void tcg_out_nop_fill(tcg_insn_unit *p, int count) diff --git a/tcg/mips/tcg-target.h b/tcg/mips/tcg-target.h index a8222476f0..5cb8672470 100644 --- a/tcg/mips/tcg-target.h +++ b/tcg/mips/tcg-target.h @@ -203,6 +203,7 @@ extern bool use_mips32r2_instructions; #endif #define TCG_TARGET_DEFAULT_MO (0) +#define TCG_TARGET_HAS_MEMORY_BSWAP 1 static inline void flush_icache_range(uintptr_t start, uintptr_t stop) { diff --git a/tcg/mips/tcg-target.inc.c b/tcg/mips/tcg-target.inc.c index cff525373b..be0bc92e8e 100644 --- a/tcg/mips/tcg-target.inc.c +++ b/tcg/mips/tcg-target.inc.c @@ -168,12 +168,13 @@ static inline void reloc_26(tcg_insn_unit *pc, tcg_insn_unit *target) *pc = deposit32(*pc, 0, 26, reloc_26_val(pc, target)); } -static void patch_reloc(tcg_insn_unit *code_ptr, int type, +static bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend) { tcg_debug_assert(type == R_MIPS_PC16); tcg_debug_assert(addend == 0); reloc_pc16(code_ptr, (tcg_insn_unit *)value); + return true; } #define TCG_CT_CONST_ZERO 0x100 @@ -483,12 +484,7 @@ static inline void tcg_out_opc_bf64(TCGContext *s, MIPSInsn opc, MIPSInsn opm, static inline void tcg_out_opc_br(TCGContext *s, MIPSInsn opc, TCGReg rt, TCGReg rs) { - /* We pay attention here to not modify the branch target by reading - the existing value and using it again. This ensure that caches and - memory are kept coherent during retranslation. */ - uint16_t offset = (uint16_t)*s->code_ptr; - - tcg_out_opc_imm(s, opc, rt, rs, offset); + tcg_out_opc_imm(s, opc, rt, rs, 0); } /* @@ -796,7 +792,7 @@ static void tcg_out_addsub2(TCGContext *s, TCGReg rl, TCGReg rh, TCGReg al, tcg_out_opc_imm(s, OPC_ADDIU, rl, al, bl); tcg_out_opc_imm(s, OPC_SLTIU, TCG_TMP0, rl, bl); } else if (rl == al && rl == bl) { - tcg_out_opc_sa(s, OPC_SRL, TCG_TMP0, al, 31); + tcg_out_opc_sa(s, OPC_SRL, TCG_TMP0, al, TCG_TARGET_REG_BITS - 1); tcg_out_opc_reg(s, OPC_ADDU, rl, al, bl); } else { tcg_out_opc_reg(s, OPC_ADDU, rl, al, bl); diff --git a/tcg/optimize.c b/tcg/optimize.c index 5dbe11c3c8..01e80c3e46 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -353,6 +353,15 @@ static TCGArg do_constant_folding_2(TCGOpcode op, TCGArg x, TCGArg y) CASE_OP_32_64(ext16u): return (uint16_t)x; + CASE_OP_32_64(bswap16): + return bswap16(x); + + CASE_OP_32_64(bswap32): + return bswap32(x); + + case INDEX_op_bswap64_i64: + return bswap64(x); + case INDEX_op_ext_i32_i64: case INDEX_op_ext32s_i64: return (int32_t)x; @@ -1105,6 +1114,9 @@ void tcg_optimize(TCGContext *s) CASE_OP_32_64(ext16s): CASE_OP_32_64(ext16u): CASE_OP_32_64(ctpop): + CASE_OP_32_64(bswap16): + CASE_OP_32_64(bswap32): + case INDEX_op_bswap64_i64: case INDEX_op_ext32s_i64: case INDEX_op_ext32u_i64: case INDEX_op_ext_i32_i64: @@ -1249,7 +1261,7 @@ void tcg_optimize(TCGContext *s) uint64_t a = ((uint64_t)ah << 32) | al; uint64_t b = ((uint64_t)bh << 32) | bl; TCGArg rl, rh; - TCGOp *op2 = tcg_op_insert_before(s, op, INDEX_op_movi_i32, 2); + TCGOp *op2 = tcg_op_insert_before(s, op, INDEX_op_movi_i32); if (opc == INDEX_op_add2_i32) { a += b; @@ -1271,7 +1283,7 @@ void tcg_optimize(TCGContext *s) uint32_t b = arg_info(op->args[3])->val; uint64_t r = (uint64_t)a * b; TCGArg rl, rh; - TCGOp *op2 = tcg_op_insert_before(s, op, INDEX_op_movi_i32, 2); + TCGOp *op2 = tcg_op_insert_before(s, op, INDEX_op_movi_i32); rl = op->args[0]; rh = op->args[1]; diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h index be52ad1d2e..52c1bb04b1 100644 --- a/tcg/ppc/tcg-target.h +++ b/tcg/ppc/tcg-target.h @@ -128,6 +128,7 @@ void flush_icache_range(uintptr_t start, uintptr_t stop); void tb_target_set_jmp_target(uintptr_t, uintptr_t, uintptr_t); #define TCG_TARGET_DEFAULT_MO (0) +#define TCG_TARGET_HAS_MEMORY_BSWAP 1 #ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS diff --git a/tcg/ppc/tcg-target.inc.c b/tcg/ppc/tcg-target.inc.c index c2f729ee8f..8c1cfdd7ac 100644 --- a/tcg/ppc/tcg-target.inc.c +++ b/tcg/ppc/tcg-target.inc.c @@ -193,9 +193,14 @@ static uint32_t reloc_pc24_val(tcg_insn_unit *pc, tcg_insn_unit *target) return disp & 0x3fffffc; } -static void reloc_pc24(tcg_insn_unit *pc, tcg_insn_unit *target) +static bool reloc_pc24(tcg_insn_unit *pc, tcg_insn_unit *target) { - *pc = (*pc & ~0x3fffffc) | reloc_pc24_val(pc, target); + ptrdiff_t disp = tcg_ptr_byte_diff(target, pc); + if (in_range_b(disp)) { + *pc = (*pc & ~0x3fffffc) | (disp & 0x3fffffc); + return true; + } + return false; } static uint16_t reloc_pc14_val(tcg_insn_unit *pc, tcg_insn_unit *target) @@ -205,21 +210,14 @@ static uint16_t reloc_pc14_val(tcg_insn_unit *pc, tcg_insn_unit *target) return disp & 0xfffc; } -static void reloc_pc14(tcg_insn_unit *pc, tcg_insn_unit *target) +static bool reloc_pc14(tcg_insn_unit *pc, tcg_insn_unit *target) { - *pc = (*pc & ~0xfffc) | reloc_pc14_val(pc, target); -} - -static inline void tcg_out_b_noaddr(TCGContext *s, int insn) -{ - unsigned retrans = *s->code_ptr & 0x3fffffc; - tcg_out32(s, insn | retrans); -} - -static inline void tcg_out_bc_noaddr(TCGContext *s, int insn) -{ - unsigned retrans = *s->code_ptr & 0xfffc; - tcg_out32(s, insn | retrans); + ptrdiff_t disp = tcg_ptr_byte_diff(target, pc); + if (disp == (int16_t) disp) { + *pc = (*pc & ~0xfffc) | (disp & 0xfffc); + return true; + } + return false; } /* parse target specific constraints */ @@ -525,7 +523,7 @@ static const uint32_t tcg_to_isel[] = { [TCG_COND_GTU] = ISEL | BC_(7, CR_GT), }; -static void patch_reloc(tcg_insn_unit *code_ptr, int type, +static bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend) { tcg_insn_unit *target; @@ -536,11 +534,9 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, switch (type) { case R_PPC_REL14: - reloc_pc14(code_ptr, target); - break; + return reloc_pc14(code_ptr, target); case R_PPC_REL24: - reloc_pc24(code_ptr, target); - break; + return reloc_pc24(code_ptr, target); case R_PPC_ADDR16: /* We are abusing this relocation type. This points to a pair of insns, addis + load. If the displacement is small, we @@ -552,7 +548,9 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, } else { int16_t lo = value; int hi = value - lo; - assert(hi + lo == value); + if (hi + lo != value) { + return false; + } code_ptr[0] = deposit32(code_ptr[0], 0, 16, hi >> 16); code_ptr[1] = deposit32(code_ptr[1], 0, 16, lo); } @@ -560,6 +558,7 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, default: g_assert_not_reached(); } + return true; } static void tcg_out_mem_long(TCGContext *s, int opi, int opx, TCGReg rt, @@ -1179,11 +1178,11 @@ static void tcg_out_setcond(TCGContext *s, TCGType type, TCGCond cond, static void tcg_out_bc(TCGContext *s, int bc, TCGLabel *l) { if (l->has_value) { - tcg_out32(s, bc | reloc_pc14_val(s->code_ptr, l->u.value_ptr)); + bc |= reloc_pc14_val(s->code_ptr, l->u.value_ptr); } else { tcg_out_reloc(s, s->code_ptr, R_PPC_REL14, l, 0); - tcg_out_bc_noaddr(s, bc); } + tcg_out32(s, bc); } static void tcg_out_brcond(TCGContext *s, TCGCond cond, @@ -1649,7 +1648,7 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) TCGMemOp opc = get_memop(oi); TCGReg hi, lo, arg = TCG_REG_R3; - reloc_pc14(lb->label_ptr[0], s->code_ptr); + **lb->label_ptr |= reloc_pc14_val(*lb->label_ptr, s->code_ptr); tcg_out_mov(s, TCG_TYPE_PTR, arg++, TCG_AREG0); @@ -1694,7 +1693,7 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) TCGMemOp s_bits = opc & MO_SIZE; TCGReg hi, lo, arg = TCG_REG_R3; - reloc_pc14(lb->label_ptr[0], s->code_ptr); + **lb->label_ptr |= reloc_pc14_val(*lb->label_ptr, s->code_ptr); tcg_out_mov(s, TCG_TYPE_PTR, arg++, TCG_AREG0); @@ -1771,7 +1770,7 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) /* Load a pointer into the current opcode w/conditional branch-link. */ label_ptr = s->code_ptr; - tcg_out_bc_noaddr(s, BC | BI(7, CR_EQ) | BO_COND_FALSE | LK); + tcg_out32(s, BC | BI(7, CR_EQ) | BO_COND_FALSE | LK); rbase = TCG_REG_R3; #else /* !CONFIG_SOFTMMU */ @@ -1846,7 +1845,7 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) /* Load a pointer into the current opcode w/conditional branch-link. */ label_ptr = s->code_ptr; - tcg_out_bc_noaddr(s, BC | BI(7, CR_EQ) | BO_COND_FALSE | LK); + tcg_out32(s, BC | BI(7, CR_EQ) | BO_COND_FALSE | LK); rbase = TCG_REG_R3; #else /* !CONFIG_SOFTMMU */ @@ -2044,13 +2043,14 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, case INDEX_op_br: { TCGLabel *l = arg_label(args[0]); + uint32_t insn = B; if (l->has_value) { - tcg_out_b(s, 0, l->u.value_ptr); + insn |= reloc_pc24_val(s->code_ptr, l->u.value_ptr); } else { tcg_out_reloc(s, s->code_ptr, R_PPC_REL24, l, 0); - tcg_out_b_noaddr(s, B); } + tcg_out32(s, insn); } break; case INDEX_op_ld8u_i32: diff --git a/tcg/s390/tcg-target.h b/tcg/s390/tcg-target.h index 6f2b06a7d1..853ed6e7aa 100644 --- a/tcg/s390/tcg-target.h +++ b/tcg/s390/tcg-target.h @@ -135,6 +135,7 @@ extern uint64_t s390_facilities; #define TCG_TARGET_CALL_STACK_OFFSET 160 #define TCG_TARGET_EXTEND_ARGS 1 +#define TCG_TARGET_HAS_MEMORY_BSWAP 1 #define TCG_TARGET_DEFAULT_MO (TCG_MO_ALL & ~TCG_MO_ST_LD) diff --git a/tcg/s390/tcg-target.inc.c b/tcg/s390/tcg-target.inc.c index 17c435ade5..39ecf609a1 100644 --- a/tcg/s390/tcg-target.inc.c +++ b/tcg/s390/tcg-target.inc.c @@ -366,7 +366,7 @@ static void * const qemu_st_helpers[16] = { static tcg_insn_unit *tb_ret_addr; uint64_t s390_facilities; -static void patch_reloc(tcg_insn_unit *code_ptr, int type, +static bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend) { intptr_t pcrel2; @@ -377,22 +377,29 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, switch (type) { case R_390_PC16DBL: - assert(pcrel2 == (int16_t)pcrel2); - tcg_patch16(code_ptr, pcrel2); + if (pcrel2 == (int16_t)pcrel2) { + tcg_patch16(code_ptr, pcrel2); + return true; + } break; case R_390_PC32DBL: - assert(pcrel2 == (int32_t)pcrel2); - tcg_patch32(code_ptr, pcrel2); + if (pcrel2 == (int32_t)pcrel2) { + tcg_patch32(code_ptr, pcrel2); + return true; + } break; case R_390_20: - assert(value == sextract64(value, 0, 20)); - old = *(uint32_t *)code_ptr & 0xf00000ff; - old |= ((value & 0xfff) << 16) | ((value & 0xff000) >> 4); - tcg_patch32(code_ptr, old); + if (value == sextract64(value, 0, 20)) { + old = *(uint32_t *)code_ptr & 0xf00000ff; + old |= ((value & 0xfff) << 16) | ((value & 0xff000) >> 4); + tcg_patch32(code_ptr, old); + return true; + } break; default: g_assert_not_reached(); } + return false; } /* parse target specific constraints */ @@ -1329,13 +1336,12 @@ static void tgen_branch(TCGContext *s, int cc, TCGLabel *l) static void tgen_compare_branch(TCGContext *s, S390Opcode opc, int cc, TCGReg r1, TCGReg r2, TCGLabel *l) { - intptr_t off; + intptr_t off = 0; if (l->has_value) { off = l->u.value_ptr - s->code_ptr; + tcg_debug_assert(off == (int16_t)off); } else { - /* We need to keep the offset unchanged for retranslation. */ - off = s->code_ptr[1]; tcg_out_reloc(s, s->code_ptr + 1, R_390_PC16DBL, l, 2); } @@ -1347,13 +1353,12 @@ static void tgen_compare_branch(TCGContext *s, S390Opcode opc, int cc, static void tgen_compare_imm_branch(TCGContext *s, S390Opcode opc, int cc, TCGReg r1, int i2, TCGLabel *l) { - tcg_target_long off; + tcg_target_long off = 0; if (l->has_value) { off = l->u.value_ptr - s->code_ptr; + tcg_debug_assert(off == (int16_t)off); } else { - /* We need to keep the offset unchanged for retranslation. */ - off = s->code_ptr[1]; tcg_out_reloc(s, s->code_ptr + 1, R_390_PC16DBL, l, 2); } @@ -1618,7 +1623,9 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) TCGMemOpIdx oi = lb->oi; TCGMemOp opc = get_memop(oi); - patch_reloc(lb->label_ptr[0], R_390_PC16DBL, (intptr_t)s->code_ptr, 2); + bool ok = patch_reloc(lb->label_ptr[0], R_390_PC16DBL, + (intptr_t)s->code_ptr, 2); + tcg_debug_assert(ok); tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_R2, TCG_AREG0); if (TARGET_LONG_BITS == 64) { @@ -1639,7 +1646,9 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) TCGMemOpIdx oi = lb->oi; TCGMemOp opc = get_memop(oi); - patch_reloc(lb->label_ptr[0], R_390_PC16DBL, (intptr_t)s->code_ptr, 2); + bool ok = patch_reloc(lb->label_ptr[0], R_390_PC16DBL, + (intptr_t)s->code_ptr, 2); + tcg_debug_assert(ok); tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_R2, TCG_AREG0); if (TARGET_LONG_BITS == 64) { @@ -1696,7 +1705,6 @@ static void tcg_out_qemu_ld(TCGContext* s, TCGReg data_reg, TCGReg addr_reg, base_reg = tcg_out_tlb_read(s, addr_reg, opc, mem_index, 1); - /* We need to keep the offset unchanged for retranslation. */ tcg_out16(s, RI_BRC | (S390_CC_NE << 4)); label_ptr = s->code_ptr; s->code_ptr += 1; @@ -1724,7 +1732,6 @@ static void tcg_out_qemu_st(TCGContext* s, TCGReg data_reg, TCGReg addr_reg, base_reg = tcg_out_tlb_read(s, addr_reg, opc, mem_index, 0); - /* We need to keep the offset unchanged for retranslation. */ tcg_out16(s, RI_BRC | (S390_CC_NE << 4)); label_ptr = s->code_ptr; s->code_ptr += 1; diff --git a/tcg/sparc/tcg-target.h b/tcg/sparc/tcg-target.h index d8339bf010..a0ed2a3342 100644 --- a/tcg/sparc/tcg-target.h +++ b/tcg/sparc/tcg-target.h @@ -164,6 +164,7 @@ extern bool use_vis3_instructions; #define TCG_AREG0 TCG_REG_I0 #define TCG_TARGET_DEFAULT_MO (0) +#define TCG_TARGET_HAS_MEMORY_BSWAP 1 static inline void flush_icache_range(uintptr_t start, uintptr_t stop) { diff --git a/tcg/sparc/tcg-target.inc.c b/tcg/sparc/tcg-target.inc.c index 04bdc3df5e..55144c437c 100644 --- a/tcg/sparc/tcg-target.inc.c +++ b/tcg/sparc/tcg-target.inc.c @@ -291,7 +291,7 @@ static inline int check_fit_i32(int32_t val, unsigned int bits) # define check_fit_ptr check_fit_i32 #endif -static void patch_reloc(tcg_insn_unit *code_ptr, int type, +static bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend) { uint32_t insn = *code_ptr; @@ -311,29 +311,12 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, insn &= ~INSN_OFF19(-1); insn |= INSN_OFF19(pcrel); break; - case R_SPARC_13: - /* Note that we're abusing this reloc type for our own needs. */ - if (!check_fit_ptr(value, 13)) { - int adj = (value > 0 ? 0xff8 : -0x1000); - value -= adj; - assert(check_fit_ptr(value, 13)); - *code_ptr++ = (ARITH_ADD | INSN_RD(TCG_REG_T2) - | INSN_RS1(TCG_REG_TB) | INSN_IMM13(adj)); - insn ^= INSN_RS1(TCG_REG_TB) ^ INSN_RS1(TCG_REG_T2); - } - insn &= ~INSN_IMM13(-1); - insn |= INSN_IMM13(value); - break; - case R_SPARC_32: - /* Note that we're abusing this reloc type for our own needs. */ - code_ptr[0] = deposit32(code_ptr[0], 0, 22, value >> 10); - code_ptr[1] = deposit32(code_ptr[1], 0, 10, value); - return; default: g_assert_not_reached(); } *code_ptr = insn; + return true; } /* parse target specific constraints */ @@ -459,6 +442,15 @@ static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret, return; } + /* A 13-bit constant relative to the TB. */ + if (!in_prologue && USE_REG_TB) { + test = arg - (uintptr_t)s->code_gen_ptr; + if (check_fit_ptr(test, 13)) { + tcg_out_arithi(s, ret, TCG_REG_TB, test, ARITH_ADD); + return; + } + } + /* A 32-bit constant, or 32-bit zero-extended to 64-bits. */ if (type == TCG_TYPE_I32 || arg == (uint32_t)arg) { tcg_out_sethi(s, ret, arg); @@ -488,26 +480,6 @@ static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret, return; } - if (!in_prologue) { - if (USE_REG_TB) { - intptr_t diff = arg - (uintptr_t)s->code_gen_ptr; - if (check_fit_ptr(diff, 13)) { - tcg_out_arithi(s, ret, TCG_REG_TB, diff, ARITH_ADD); - } else { - new_pool_label(s, arg, R_SPARC_13, s->code_ptr, - -(intptr_t)s->code_gen_ptr); - tcg_out32(s, LDX | INSN_RD(ret) | INSN_RS1(TCG_REG_TB)); - /* May be used to extend the 13-bit range in patch_reloc. */ - tcg_out32(s, NOP); - } - } else { - new_pool_label(s, arg, R_SPARC_32, s->code_ptr, 0); - tcg_out_sethi(s, ret, 0); - tcg_out32(s, LDX | INSN_RD(ret) | INSN_RS1(ret) | INSN_IMM13(0)); - } - return; - } - /* A 64-bit constant decomposed into 2 32-bit pieces. */ if (check_fit_i32(lo, 13)) { hi = (arg - lo) >> 32; @@ -639,13 +611,11 @@ static void tcg_out_bpcc0(TCGContext *s, int scond, int flags, int off19) static void tcg_out_bpcc(TCGContext *s, int scond, int flags, TCGLabel *l) { - int off19; + int off19 = 0; if (l->has_value) { off19 = INSN_OFF19(tcg_pcrel_diff(s, l->u.value_ptr)); } else { - /* Make sure to preserve destinations during retranslation. */ - off19 = *s->code_ptr & INSN_OFF19(-1); tcg_out_reloc(s, s->code_ptr, R_SPARC_WDISP19, l, 0); } tcg_out_bpcc0(s, scond, flags, off19); @@ -685,13 +655,11 @@ static void tcg_out_brcond_i64(TCGContext *s, TCGCond cond, TCGReg arg1, { /* For 64-bit signed comparisons vs zero, we can avoid the compare. */ if (arg2 == 0 && !is_unsigned_cond(cond)) { - int off16; + int off16 = 0; if (l->has_value) { off16 = INSN_OFF16(tcg_pcrel_diff(s, l->u.value_ptr)); } else { - /* Make sure to preserve destinations during retranslation. */ - off16 = *s->code_ptr & INSN_OFF16(-1); tcg_out_reloc(s, s->code_ptr, R_SPARC_WDISP16, l, 0); } tcg_out32(s, INSN_OP(0) | INSN_OP2(3) | BPR_PT | INSN_RS1(arg1) diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c index 7a8015c5a9..38652db32c 100644 --- a/tcg/tcg-op.c +++ b/tcg/tcg-op.c @@ -1012,24 +1012,24 @@ void tcg_gen_bswap32_i32(TCGv_i32 ret, TCGv_i32 arg) if (TCG_TARGET_HAS_bswap32_i32) { tcg_gen_op2_i32(INDEX_op_bswap32_i32, ret, arg); } else { - TCGv_i32 t0, t1; - t0 = tcg_temp_new_i32(); - t1 = tcg_temp_new_i32(); - - tcg_gen_shli_i32(t0, arg, 24); + TCGv_i32 t0 = tcg_temp_new_i32(); + TCGv_i32 t1 = tcg_temp_new_i32(); + TCGv_i32 t2 = tcg_const_i32(0x00ff00ff); - tcg_gen_andi_i32(t1, arg, 0x0000ff00); - tcg_gen_shli_i32(t1, t1, 8); - tcg_gen_or_i32(t0, t0, t1); + /* arg = abcd */ + tcg_gen_shri_i32(t0, arg, 8); /* t0 = .abc */ + tcg_gen_and_i32(t1, arg, t2); /* t1 = .b.d */ + tcg_gen_and_i32(t0, t0, t2); /* t0 = .a.c */ + tcg_gen_shli_i32(t1, t1, 8); /* t1 = b.d. */ + tcg_gen_or_i32(ret, t0, t1); /* ret = badc */ - tcg_gen_shri_i32(t1, arg, 8); - tcg_gen_andi_i32(t1, t1, 0x0000ff00); - tcg_gen_or_i32(t0, t0, t1); + tcg_gen_shri_i32(t0, ret, 16); /* t0 = ..ba */ + tcg_gen_shli_i32(t1, ret, 16); /* t1 = dc.. */ + tcg_gen_or_i32(ret, t0, t1); /* ret = dcba */ - tcg_gen_shri_i32(t1, arg, 24); - tcg_gen_or_i32(ret, t0, t1); tcg_temp_free_i32(t0); tcg_temp_free_i32(t1); + tcg_temp_free_i32(t2); } } @@ -1638,25 +1638,25 @@ void tcg_gen_bswap32_i64(TCGv_i64 ret, TCGv_i64 arg) } else if (TCG_TARGET_HAS_bswap32_i64) { tcg_gen_op2_i64(INDEX_op_bswap32_i64, ret, arg); } else { - TCGv_i64 t0, t1; - t0 = tcg_temp_new_i64(); - t1 = tcg_temp_new_i64(); - - tcg_gen_shli_i64(t0, arg, 24); - tcg_gen_ext32u_i64(t0, t0); + TCGv_i64 t0 = tcg_temp_new_i64(); + TCGv_i64 t1 = tcg_temp_new_i64(); + TCGv_i64 t2 = tcg_const_i64(0x00ff00ff); - tcg_gen_andi_i64(t1, arg, 0x0000ff00); - tcg_gen_shli_i64(t1, t1, 8); - tcg_gen_or_i64(t0, t0, t1); + /* arg = ....abcd */ + tcg_gen_shri_i64(t0, arg, 8); /* t0 = .....abc */ + tcg_gen_and_i64(t1, arg, t2); /* t1 = .....b.d */ + tcg_gen_and_i64(t0, t0, t2); /* t0 = .....a.c */ + tcg_gen_shli_i64(t1, t1, 8); /* t1 = ....b.d. */ + tcg_gen_or_i64(ret, t0, t1); /* ret = ....badc */ - tcg_gen_shri_i64(t1, arg, 8); - tcg_gen_andi_i64(t1, t1, 0x0000ff00); - tcg_gen_or_i64(t0, t0, t1); + tcg_gen_shli_i64(t1, ret, 48); /* t1 = dc...... */ + tcg_gen_shri_i64(t0, ret, 16); /* t0 = ......ba */ + tcg_gen_shri_i64(t1, t1, 32); /* t1 = ....dc.. */ + tcg_gen_or_i64(ret, t0, t1); /* ret = ....dcba */ - tcg_gen_shri_i64(t1, arg, 24); - tcg_gen_or_i64(ret, t0, t1); tcg_temp_free_i64(t0); tcg_temp_free_i64(t1); + tcg_temp_free_i64(t2); } } @@ -1678,37 +1678,30 @@ void tcg_gen_bswap64_i64(TCGv_i64 ret, TCGv_i64 arg) } else { TCGv_i64 t0 = tcg_temp_new_i64(); TCGv_i64 t1 = tcg_temp_new_i64(); + TCGv_i64 t2 = tcg_temp_new_i64(); - tcg_gen_shli_i64(t0, arg, 56); - - tcg_gen_andi_i64(t1, arg, 0x0000ff00); - tcg_gen_shli_i64(t1, t1, 40); - tcg_gen_or_i64(t0, t0, t1); - - tcg_gen_andi_i64(t1, arg, 0x00ff0000); - tcg_gen_shli_i64(t1, t1, 24); - tcg_gen_or_i64(t0, t0, t1); - - tcg_gen_andi_i64(t1, arg, 0xff000000); - tcg_gen_shli_i64(t1, t1, 8); - tcg_gen_or_i64(t0, t0, t1); - - tcg_gen_shri_i64(t1, arg, 8); - tcg_gen_andi_i64(t1, t1, 0xff000000); - tcg_gen_or_i64(t0, t0, t1); - - tcg_gen_shri_i64(t1, arg, 24); - tcg_gen_andi_i64(t1, t1, 0x00ff0000); - tcg_gen_or_i64(t0, t0, t1); - - tcg_gen_shri_i64(t1, arg, 40); - tcg_gen_andi_i64(t1, t1, 0x0000ff00); - tcg_gen_or_i64(t0, t0, t1); + /* arg = abcdefgh */ + tcg_gen_movi_i64(t2, 0x00ff00ff00ff00ffull); + tcg_gen_shri_i64(t0, arg, 8); /* t0 = .abcdefg */ + tcg_gen_and_i64(t1, arg, t2); /* t1 = .b.d.f.h */ + tcg_gen_and_i64(t0, t0, t2); /* t0 = .a.c.e.g */ + tcg_gen_shli_i64(t1, t1, 8); /* t1 = b.d.f.h. */ + tcg_gen_or_i64(ret, t0, t1); /* ret = badcfehg */ + + tcg_gen_movi_i64(t2, 0x0000ffff0000ffffull); + tcg_gen_shri_i64(t0, ret, 16); /* t0 = ..badcfe */ + tcg_gen_and_i64(t1, ret, t2); /* t1 = ..dc..hg */ + tcg_gen_and_i64(t0, t0, t2); /* t0 = ..ba..fe */ + tcg_gen_shli_i64(t1, t1, 16); /* t1 = dc..hg.. */ + tcg_gen_or_i64(ret, t0, t1); /* ret = dcbahgfe */ + + tcg_gen_shri_i64(t0, ret, 32); /* t0 = ....dcba */ + tcg_gen_shli_i64(t1, ret, 32); /* t1 = hgfe.... */ + tcg_gen_or_i64(ret, t0, t1); /* ret = hgfedcba */ - tcg_gen_shri_i64(t1, arg, 56); - tcg_gen_or_i64(ret, t0, t1); tcg_temp_free_i64(t0); tcg_temp_free_i64(t1); + tcg_temp_free_i64(t2); } } @@ -2701,25 +2694,78 @@ static void tcg_gen_req_mo(TCGBar type) void tcg_gen_qemu_ld_i32(TCGv_i32 val, TCGv addr, TCGArg idx, TCGMemOp memop) { + TCGMemOp orig_memop; + tcg_gen_req_mo(TCG_MO_LD_LD | TCG_MO_ST_LD); memop = tcg_canonicalize_memop(memop, 0, 0); trace_guest_mem_before_tcg(tcg_ctx->cpu, cpu_env, addr, trace_mem_get_info(memop, 0)); + + orig_memop = memop; + if (!TCG_TARGET_HAS_MEMORY_BSWAP && (memop & MO_BSWAP)) { + memop &= ~MO_BSWAP; + /* The bswap primitive requires zero-extended input. */ + if ((memop & MO_SSIZE) == MO_SW) { + memop &= ~MO_SIGN; + } + } + gen_ldst_i32(INDEX_op_qemu_ld_i32, val, addr, memop, idx); + + if ((orig_memop ^ memop) & MO_BSWAP) { + switch (orig_memop & MO_SIZE) { + case MO_16: + tcg_gen_bswap16_i32(val, val); + if (orig_memop & MO_SIGN) { + tcg_gen_ext16s_i32(val, val); + } + break; + case MO_32: + tcg_gen_bswap32_i32(val, val); + break; + default: + g_assert_not_reached(); + } + } } void tcg_gen_qemu_st_i32(TCGv_i32 val, TCGv addr, TCGArg idx, TCGMemOp memop) { + TCGv_i32 swap = NULL; + tcg_gen_req_mo(TCG_MO_LD_ST | TCG_MO_ST_ST); memop = tcg_canonicalize_memop(memop, 0, 1); trace_guest_mem_before_tcg(tcg_ctx->cpu, cpu_env, addr, trace_mem_get_info(memop, 1)); + + if (!TCG_TARGET_HAS_MEMORY_BSWAP && (memop & MO_BSWAP)) { + swap = tcg_temp_new_i32(); + switch (memop & MO_SIZE) { + case MO_16: + tcg_gen_ext16u_i32(swap, val); + tcg_gen_bswap16_i32(swap, swap); + break; + case MO_32: + tcg_gen_bswap32_i32(swap, val); + break; + default: + g_assert_not_reached(); + } + val = swap; + memop &= ~MO_BSWAP; + } + gen_ldst_i32(INDEX_op_qemu_st_i32, val, addr, memop, idx); + + if (swap) { + tcg_temp_free_i32(swap); + } } void tcg_gen_qemu_ld_i64(TCGv_i64 val, TCGv addr, TCGArg idx, TCGMemOp memop) { - tcg_gen_req_mo(TCG_MO_LD_LD | TCG_MO_ST_LD); + TCGMemOp orig_memop; + if (TCG_TARGET_REG_BITS == 32 && (memop & MO_SIZE) < MO_64) { tcg_gen_qemu_ld_i32(TCGV_LOW(val), addr, idx, memop); if (memop & MO_SIGN) { @@ -2730,24 +2776,85 @@ void tcg_gen_qemu_ld_i64(TCGv_i64 val, TCGv addr, TCGArg idx, TCGMemOp memop) return; } + tcg_gen_req_mo(TCG_MO_LD_LD | TCG_MO_ST_LD); memop = tcg_canonicalize_memop(memop, 1, 0); trace_guest_mem_before_tcg(tcg_ctx->cpu, cpu_env, addr, trace_mem_get_info(memop, 0)); + + orig_memop = memop; + if (!TCG_TARGET_HAS_MEMORY_BSWAP && (memop & MO_BSWAP)) { + memop &= ~MO_BSWAP; + /* The bswap primitive requires zero-extended input. */ + if ((memop & MO_SIGN) && (memop & MO_SIZE) < MO_64) { + memop &= ~MO_SIGN; + } + } + gen_ldst_i64(INDEX_op_qemu_ld_i64, val, addr, memop, idx); + + if ((orig_memop ^ memop) & MO_BSWAP) { + switch (orig_memop & MO_SIZE) { + case MO_16: + tcg_gen_bswap16_i64(val, val); + if (orig_memop & MO_SIGN) { + tcg_gen_ext16s_i64(val, val); + } + break; + case MO_32: + tcg_gen_bswap32_i64(val, val); + if (orig_memop & MO_SIGN) { + tcg_gen_ext32s_i64(val, val); + } + break; + case MO_64: + tcg_gen_bswap64_i64(val, val); + break; + default: + g_assert_not_reached(); + } + } } void tcg_gen_qemu_st_i64(TCGv_i64 val, TCGv addr, TCGArg idx, TCGMemOp memop) { - tcg_gen_req_mo(TCG_MO_LD_ST | TCG_MO_ST_ST); + TCGv_i64 swap = NULL; + if (TCG_TARGET_REG_BITS == 32 && (memop & MO_SIZE) < MO_64) { tcg_gen_qemu_st_i32(TCGV_LOW(val), addr, idx, memop); return; } + tcg_gen_req_mo(TCG_MO_LD_ST | TCG_MO_ST_ST); memop = tcg_canonicalize_memop(memop, 1, 1); trace_guest_mem_before_tcg(tcg_ctx->cpu, cpu_env, addr, trace_mem_get_info(memop, 1)); + + if (!TCG_TARGET_HAS_MEMORY_BSWAP && (memop & MO_BSWAP)) { + swap = tcg_temp_new_i64(); + switch (memop & MO_SIZE) { + case MO_16: + tcg_gen_ext16u_i64(swap, val); + tcg_gen_bswap16_i64(swap, swap); + break; + case MO_32: + tcg_gen_ext32u_i64(swap, val); + tcg_gen_bswap32_i64(swap, swap); + break; + case MO_64: + tcg_gen_bswap64_i64(swap, val); + break; + default: + g_assert_not_reached(); + } + val = swap; + memop &= ~MO_BSWAP; + } + gen_ldst_i64(INDEX_op_qemu_st_i64, val, addr, memop, idx); + + if (swap) { + tcg_temp_free_i64(swap); + } } static void tcg_gen_ext_i32(TCGv_i32 ret, TCGv_i32 val, TCGMemOp opc) @@ -66,7 +66,7 @@ static void tcg_target_init(TCGContext *s); static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode); static void tcg_target_qemu_prologue(TCGContext *s); -static void patch_reloc(tcg_insn_unit *code_ptr, int type, +static bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend); /* The CIE and FDE header definitions will be common to all hosts. */ @@ -268,7 +268,8 @@ static void tcg_out_reloc(TCGContext *s, tcg_insn_unit *code_ptr, int type, /* FIXME: This may break relocations on RISC targets that modify instruction fields in place. The caller may not have written the initial value. */ - patch_reloc(code_ptr, type, l->u.value, addend); + bool ok = patch_reloc(code_ptr, type, l->u.value, addend); + tcg_debug_assert(ok); } else { /* add a new relocation entry */ r = tcg_malloc(sizeof(TCGRelocation)); @@ -288,7 +289,8 @@ static void tcg_out_label(TCGContext *s, TCGLabel *l, tcg_insn_unit *ptr) tcg_debug_assert(!l->has_value); for (r = l->u.first_reloc; r != NULL; r = r->next) { - patch_reloc(r->ptr, r->type, value, r->addend); + bool ok = patch_reloc(r->ptr, r->type, value, r->addend); + tcg_debug_assert(ok); } l->has_value = 1; @@ -2203,16 +2205,14 @@ TCGOp *tcg_emit_op(TCGOpcode opc) return op; } -TCGOp *tcg_op_insert_before(TCGContext *s, TCGOp *old_op, - TCGOpcode opc, int nargs) +TCGOp *tcg_op_insert_before(TCGContext *s, TCGOp *old_op, TCGOpcode opc) { TCGOp *new_op = tcg_op_alloc(opc); QTAILQ_INSERT_BEFORE(old_op, new_op, link); return new_op; } -TCGOp *tcg_op_insert_after(TCGContext *s, TCGOp *old_op, - TCGOpcode opc, int nargs) +TCGOp *tcg_op_insert_after(TCGContext *s, TCGOp *old_op, TCGOpcode opc) { TCGOp *new_op = tcg_op_alloc(opc); QTAILQ_INSERT_AFTER(&s->ops, old_op, new_op, link); @@ -2550,7 +2550,7 @@ static bool liveness_pass_2(TCGContext *s) TCGOpcode lopc = (arg_ts->type == TCG_TYPE_I32 ? INDEX_op_ld_i32 : INDEX_op_ld_i64); - TCGOp *lop = tcg_op_insert_before(s, op, lopc, 3); + TCGOp *lop = tcg_op_insert_before(s, op, lopc); lop->args[0] = temp_arg(dir_ts); lop->args[1] = temp_arg(arg_ts->mem_base); @@ -2619,7 +2619,7 @@ static bool liveness_pass_2(TCGContext *s) TCGOpcode sopc = (arg_ts->type == TCG_TYPE_I32 ? INDEX_op_st_i32 : INDEX_op_st_i64); - TCGOp *sop = tcg_op_insert_after(s, op, sopc, 3); + TCGOp *sop = tcg_op_insert_after(s, op, sopc); sop->args[0] = temp_arg(dir_ts); sop->args[1] = temp_arg(arg_ts->mem_base); @@ -1071,8 +1071,8 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args); TCGOp *tcg_emit_op(TCGOpcode opc); void tcg_op_remove(TCGContext *s, TCGOp *op); -TCGOp *tcg_op_insert_before(TCGContext *s, TCGOp *op, TCGOpcode opc, int narg); -TCGOp *tcg_op_insert_after(TCGContext *s, TCGOp *op, TCGOpcode opc, int narg); +TCGOp *tcg_op_insert_before(TCGContext *s, TCGOp *op, TCGOpcode opc); +TCGOp *tcg_op_insert_after(TCGContext *s, TCGOp *op, TCGOpcode opc); void tcg_optimize(TCGContext *s); diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h index 26140d78cb..086f34e69a 100644 --- a/tcg/tci/tcg-target.h +++ b/tcg/tci/tcg-target.h @@ -198,6 +198,8 @@ static inline void flush_icache_range(uintptr_t start, uintptr_t stop) We prefer consistency across hosts on this. */ #define TCG_TARGET_DEFAULT_MO (0) +#define TCG_TARGET_HAS_MEMORY_BSWAP 1 + static inline void tb_target_set_jmp_target(uintptr_t tc_ptr, uintptr_t jmp_addr, uintptr_t addr) { diff --git a/tcg/tci/tcg-target.inc.c b/tcg/tci/tcg-target.inc.c index 62ed097254..0015a98485 100644 --- a/tcg/tci/tcg-target.inc.c +++ b/tcg/tci/tcg-target.inc.c @@ -369,7 +369,7 @@ static const char *const tcg_target_reg_names[TCG_TARGET_NB_REGS] = { }; #endif -static void patch_reloc(tcg_insn_unit *code_ptr, int type, +static bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend) { /* tcg_out_reloc always uses the same type, addend. */ @@ -381,6 +381,7 @@ static void patch_reloc(tcg_insn_unit *code_ptr, int type, } else { tcg_patch64(code_ptr, value); } + return true; } /* Parse target specific constraints. */ diff --git a/tests/fp/.gitignore b/tests/fp/.gitignore index 8d45d18ac4..704fd42992 100644 --- a/tests/fp/.gitignore +++ b/tests/fp/.gitignore @@ -1 +1,2 @@ fp-test +fp-bench diff --git a/tests/fp/Makefile b/tests/fp/Makefile index d649a5a1db..5019dcdca0 100644 --- a/tests/fp/Makefile +++ b/tests/fp/Makefile @@ -29,6 +29,9 @@ QEMU_INCLUDES += -I$(TF_SOURCE_DIR) # work around TARGET_* poisoning QEMU_CFLAGS += -DHW_POISON_H +# define a target to match testfloat's implementation-defined choices, such as +# whether to raise the invalid flag when dealing with NaNs in muladd. +QEMU_CFLAGS += -DTARGET_ARM # capstone has a platform.h file that clashes with softfloat's QEMU_CFLAGS := $(filter-out %capstone, $(QEMU_CFLAGS)) @@ -550,7 +553,7 @@ TF_OBJS_LIB += $(TF_OBJS_WRITECASE) TF_OBJS_LIB += testLoops_common.o TF_OBJS_LIB += $(TF_OBJS_TEST) -BINARIES := fp-test$(EXESUF) +BINARIES := fp-test$(EXESUF) fp-bench$(EXESUF) # everything depends on config-host.h because platform.h includes it all: $(BUILD_DIR)/config-host.h @@ -587,10 +590,13 @@ $(TF_OBJS_LIB) slowfloat.o: %.o: $(TF_SOURCE_DIR)/%.c libtestfloat.a: $(TF_OBJS_LIB) +fp-bench$(EXESUF): fp-bench.o $(QEMU_SOFTFLOAT_OBJ) $(LIBQEMUUTIL) + clean: rm -f *.o *.d $(BINARIES) rm -f *.gcno *.gcda *.gcov rm -f fp-test$(EXESUF) + rm -f fp-bench$(EXESUF) rm -f libsoftfloat.a rm -f libtestfloat.a diff --git a/tests/fp/fp-bench.c b/tests/fp/fp-bench.c new file mode 100644 index 0000000000..f5bc5edebf --- /dev/null +++ b/tests/fp/fp-bench.c @@ -0,0 +1,630 @@ +/* + * fp-bench.c - A collection of simple floating point microbenchmarks. + * + * Copyright (C) 2018, Emilio G. Cota <cota@braap.org> + * + * License: GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ +#ifndef HW_POISON_H +#error Must define HW_POISON_H to work around TARGET_* poisoning +#endif + +#include "qemu/osdep.h" +#include <math.h> +#include <fenv.h> +#include "qemu/timer.h" +#include "fpu/softfloat.h" + +/* amortize the computation of random inputs */ +#define OPS_PER_ITER 50000 + +#define MAX_OPERANDS 3 + +#define SEED_A 0xdeadfacedeadface +#define SEED_B 0xbadc0feebadc0fee +#define SEED_C 0xbeefdeadbeefdead + +enum op { + OP_ADD, + OP_SUB, + OP_MUL, + OP_DIV, + OP_FMA, + OP_SQRT, + OP_CMP, + OP_MAX_NR, +}; + +static const char * const op_names[] = { + [OP_ADD] = "add", + [OP_SUB] = "sub", + [OP_MUL] = "mul", + [OP_DIV] = "div", + [OP_FMA] = "mulAdd", + [OP_SQRT] = "sqrt", + [OP_CMP] = "cmp", + [OP_MAX_NR] = NULL, +}; + +enum precision { + PREC_SINGLE, + PREC_DOUBLE, + PREC_FLOAT32, + PREC_FLOAT64, + PREC_MAX_NR, +}; + +enum rounding { + ROUND_EVEN, + ROUND_ZERO, + ROUND_DOWN, + ROUND_UP, + ROUND_TIEAWAY, + N_ROUND_MODES, +}; + +static const char * const round_names[] = { + [ROUND_EVEN] = "even", + [ROUND_ZERO] = "zero", + [ROUND_DOWN] = "down", + [ROUND_UP] = "up", + [ROUND_TIEAWAY] = "tieaway", +}; + +enum tester { + TESTER_SOFT, + TESTER_HOST, + TESTER_MAX_NR, +}; + +static const char * const tester_names[] = { + [TESTER_SOFT] = "soft", + [TESTER_HOST] = "host", + [TESTER_MAX_NR] = NULL, +}; + +union fp { + float f; + double d; + float32 f32; + float64 f64; + uint64_t u64; +}; + +struct op_state; + +typedef float (*float_func_t)(const struct op_state *s); +typedef double (*double_func_t)(const struct op_state *s); + +union fp_func { + float_func_t float_func; + double_func_t double_func; +}; + +typedef void (*bench_func_t)(void); + +struct op_desc { + const char * const name; +}; + +#define DEFAULT_DURATION_SECS 1 + +static uint64_t random_ops[MAX_OPERANDS] = { + SEED_A, SEED_B, SEED_C, +}; +static float_status soft_status; +static enum precision precision; +static enum op operation; +static enum tester tester; +static uint64_t n_completed_ops; +static unsigned int duration = DEFAULT_DURATION_SECS; +static int64_t ns_elapsed; +/* disable optimizations with volatile */ +static volatile union fp res; + +/* + * From: https://en.wikipedia.org/wiki/Xorshift + * This is faster than rand_r(), and gives us a wider range (RAND_MAX is only + * guaranteed to be >= INT_MAX). + */ +static uint64_t xorshift64star(uint64_t x) +{ + x ^= x >> 12; /* a */ + x ^= x << 25; /* b */ + x ^= x >> 27; /* c */ + return x * UINT64_C(2685821657736338717); +} + +static void update_random_ops(int n_ops, enum precision prec) +{ + int i; + + for (i = 0; i < n_ops; i++) { + uint64_t r = random_ops[i]; + + if (prec == PREC_SINGLE || PREC_FLOAT32) { + do { + r = xorshift64star(r); + } while (!float32_is_normal(r)); + } else if (prec == PREC_DOUBLE || PREC_FLOAT64) { + do { + r = xorshift64star(r); + } while (!float64_is_normal(r)); + } else { + g_assert_not_reached(); + } + random_ops[i] = r; + } +} + +static void fill_random(union fp *ops, int n_ops, enum precision prec, + bool no_neg) +{ + int i; + + for (i = 0; i < n_ops; i++) { + switch (prec) { + case PREC_SINGLE: + case PREC_FLOAT32: + ops[i].f32 = make_float32(random_ops[i]); + if (no_neg && float32_is_neg(ops[i].f32)) { + ops[i].f32 = float32_chs(ops[i].f32); + } + /* raise the exponent to limit the frequency of denormal results */ + ops[i].f32 |= 0x40000000; + break; + case PREC_DOUBLE: + case PREC_FLOAT64: + ops[i].f64 = make_float64(random_ops[i]); + if (no_neg && float64_is_neg(ops[i].f64)) { + ops[i].f64 = float64_chs(ops[i].f64); + } + /* raise the exponent to limit the frequency of denormal results */ + ops[i].f64 |= LIT64(0x4000000000000000); + break; + default: + g_assert_not_reached(); + } + } +} + +/* + * The main benchmark function. Instead of (ab)using macros, we rely + * on the compiler to unfold this at compile-time. + */ +static void bench(enum precision prec, enum op op, int n_ops, bool no_neg) +{ + int64_t tf = get_clock() + duration * 1000000000LL; + + while (get_clock() < tf) { + union fp ops[MAX_OPERANDS]; + int64_t t0; + int i; + + update_random_ops(n_ops, prec); + switch (prec) { + case PREC_SINGLE: + fill_random(ops, n_ops, prec, no_neg); + t0 = get_clock(); + for (i = 0; i < OPS_PER_ITER; i++) { + float a = ops[0].f; + float b = ops[1].f; + float c = ops[2].f; + + switch (op) { + case OP_ADD: + res.f = a + b; + break; + case OP_SUB: + res.f = a - b; + break; + case OP_MUL: + res.f = a * b; + break; + case OP_DIV: + res.f = a / b; + break; + case OP_FMA: + res.f = fmaf(a, b, c); + break; + case OP_SQRT: + res.f = sqrtf(a); + break; + case OP_CMP: + res.u64 = isgreater(a, b); + break; + default: + g_assert_not_reached(); + } + } + break; + case PREC_DOUBLE: + fill_random(ops, n_ops, prec, no_neg); + t0 = get_clock(); + for (i = 0; i < OPS_PER_ITER; i++) { + double a = ops[0].d; + double b = ops[1].d; + double c = ops[2].d; + + switch (op) { + case OP_ADD: + res.d = a + b; + break; + case OP_SUB: + res.d = a - b; + break; + case OP_MUL: + res.d = a * b; + break; + case OP_DIV: + res.d = a / b; + break; + case OP_FMA: + res.d = fma(a, b, c); + break; + case OP_SQRT: + res.d = sqrt(a); + break; + case OP_CMP: + res.u64 = isgreater(a, b); + break; + default: + g_assert_not_reached(); + } + } + break; + case PREC_FLOAT32: + fill_random(ops, n_ops, prec, no_neg); + t0 = get_clock(); + for (i = 0; i < OPS_PER_ITER; i++) { + float32 a = ops[0].f32; + float32 b = ops[1].f32; + float32 c = ops[2].f32; + + switch (op) { + case OP_ADD: + res.f32 = float32_add(a, b, &soft_status); + break; + case OP_SUB: + res.f32 = float32_sub(a, b, &soft_status); + break; + case OP_MUL: + res.f = float32_mul(a, b, &soft_status); + break; + case OP_DIV: + res.f32 = float32_div(a, b, &soft_status); + break; + case OP_FMA: + res.f32 = float32_muladd(a, b, c, 0, &soft_status); + break; + case OP_SQRT: + res.f32 = float32_sqrt(a, &soft_status); + break; + case OP_CMP: + res.u64 = float32_compare_quiet(a, b, &soft_status); + break; + default: + g_assert_not_reached(); + } + } + break; + case PREC_FLOAT64: + fill_random(ops, n_ops, prec, no_neg); + t0 = get_clock(); + for (i = 0; i < OPS_PER_ITER; i++) { + float64 a = ops[0].f64; + float64 b = ops[1].f64; + float64 c = ops[2].f64; + + switch (op) { + case OP_ADD: + res.f64 = float64_add(a, b, &soft_status); + break; + case OP_SUB: + res.f64 = float64_sub(a, b, &soft_status); + break; + case OP_MUL: + res.f = float64_mul(a, b, &soft_status); + break; + case OP_DIV: + res.f64 = float64_div(a, b, &soft_status); + break; + case OP_FMA: + res.f64 = float64_muladd(a, b, c, 0, &soft_status); + break; + case OP_SQRT: + res.f64 = float64_sqrt(a, &soft_status); + break; + case OP_CMP: + res.u64 = float64_compare_quiet(a, b, &soft_status); + break; + default: + g_assert_not_reached(); + } + } + break; + default: + g_assert_not_reached(); + } + ns_elapsed += get_clock() - t0; + n_completed_ops += OPS_PER_ITER; + } +} + +#define GEN_BENCH(name, type, prec, op, n_ops) \ + static void __attribute__((flatten)) name(void) \ + { \ + bench(prec, op, n_ops, false); \ + } + +#define GEN_BENCH_NO_NEG(name, type, prec, op, n_ops) \ + static void __attribute__((flatten)) name(void) \ + { \ + bench(prec, op, n_ops, true); \ + } + +#define GEN_BENCH_ALL_TYPES(opname, op, n_ops) \ + GEN_BENCH(bench_ ## opname ## _float, float, PREC_SINGLE, op, n_ops) \ + GEN_BENCH(bench_ ## opname ## _double, double, PREC_DOUBLE, op, n_ops) \ + GEN_BENCH(bench_ ## opname ## _float32, float32, PREC_FLOAT32, op, n_ops) \ + GEN_BENCH(bench_ ## opname ## _float64, float64, PREC_FLOAT64, op, n_ops) + +GEN_BENCH_ALL_TYPES(add, OP_ADD, 2) +GEN_BENCH_ALL_TYPES(sub, OP_SUB, 2) +GEN_BENCH_ALL_TYPES(mul, OP_MUL, 2) +GEN_BENCH_ALL_TYPES(div, OP_DIV, 2) +GEN_BENCH_ALL_TYPES(fma, OP_FMA, 3) +GEN_BENCH_ALL_TYPES(cmp, OP_CMP, 2) +#undef GEN_BENCH_ALL_TYPES + +#define GEN_BENCH_ALL_TYPES_NO_NEG(name, op, n) \ + GEN_BENCH_NO_NEG(bench_ ## name ## _float, float, PREC_SINGLE, op, n) \ + GEN_BENCH_NO_NEG(bench_ ## name ## _double, double, PREC_DOUBLE, op, n) \ + GEN_BENCH_NO_NEG(bench_ ## name ## _float32, float32, PREC_FLOAT32, op, n) \ + GEN_BENCH_NO_NEG(bench_ ## name ## _float64, float64, PREC_FLOAT64, op, n) + +GEN_BENCH_ALL_TYPES_NO_NEG(sqrt, OP_SQRT, 1) +#undef GEN_BENCH_ALL_TYPES_NO_NEG + +#undef GEN_BENCH_NO_NEG +#undef GEN_BENCH + +#define GEN_BENCH_FUNCS(opname, op) \ + [op] = { \ + [PREC_SINGLE] = bench_ ## opname ## _float, \ + [PREC_DOUBLE] = bench_ ## opname ## _double, \ + [PREC_FLOAT32] = bench_ ## opname ## _float32, \ + [PREC_FLOAT64] = bench_ ## opname ## _float64, \ + } + +static const bench_func_t bench_funcs[OP_MAX_NR][PREC_MAX_NR] = { + GEN_BENCH_FUNCS(add, OP_ADD), + GEN_BENCH_FUNCS(sub, OP_SUB), + GEN_BENCH_FUNCS(mul, OP_MUL), + GEN_BENCH_FUNCS(div, OP_DIV), + GEN_BENCH_FUNCS(fma, OP_FMA), + GEN_BENCH_FUNCS(sqrt, OP_SQRT), + GEN_BENCH_FUNCS(cmp, OP_CMP), +}; + +#undef GEN_BENCH_FUNCS + +static void run_bench(void) +{ + bench_func_t f; + + f = bench_funcs[operation][precision]; + g_assert(f); + f(); +} + +/* @arr must be NULL-terminated */ +static int find_name(const char * const *arr, const char *name) +{ + int i; + + for (i = 0; arr[i] != NULL; i++) { + if (strcmp(name, arr[i]) == 0) { + return i; + } + } + return -1; +} + +static void usage_complete(int argc, char *argv[]) +{ + gchar *op_list = g_strjoinv(", ", (gchar **)op_names); + gchar *tester_list = g_strjoinv(", ", (gchar **)tester_names); + + fprintf(stderr, "Usage: %s [options]\n", argv[0]); + fprintf(stderr, "options:\n"); + fprintf(stderr, " -d = duration, in seconds. Default: %d\n", + DEFAULT_DURATION_SECS); + fprintf(stderr, " -h = show this help message.\n"); + fprintf(stderr, " -o = floating point operation (%s). Default: %s\n", + op_list, op_names[0]); + fprintf(stderr, " -p = floating point precision (single, double). " + "Default: single\n"); + fprintf(stderr, " -r = rounding mode (even, zero, down, up, tieaway). " + "Default: even\n"); + fprintf(stderr, " -t = tester (%s). Default: %s\n", + tester_list, tester_names[0]); + fprintf(stderr, " -z = flush inputs to zero (soft tester only). " + "Default: disabled\n"); + fprintf(stderr, " -Z = flush output to zero (soft tester only). " + "Default: disabled\n"); + + g_free(tester_list); + g_free(op_list); +} + +static int round_name_to_mode(const char *name) +{ + int i; + + for (i = 0; i < N_ROUND_MODES; i++) { + if (!strcmp(round_names[i], name)) { + return i; + } + } + return -1; +} + +static void QEMU_NORETURN die_host_rounding(enum rounding rounding) +{ + fprintf(stderr, "fatal: '%s' rounding not supported on this host\n", + round_names[rounding]); + exit(EXIT_FAILURE); +} + +static void set_host_precision(enum rounding rounding) +{ + int rhost; + + switch (rounding) { + case ROUND_EVEN: + rhost = FE_TONEAREST; + break; + case ROUND_ZERO: + rhost = FE_TOWARDZERO; + break; + case ROUND_DOWN: + rhost = FE_DOWNWARD; + break; + case ROUND_UP: + rhost = FE_UPWARD; + break; + case ROUND_TIEAWAY: + die_host_rounding(rounding); + return; + default: + g_assert_not_reached(); + } + + if (fesetround(rhost)) { + die_host_rounding(rounding); + } +} + +static void set_soft_precision(enum rounding rounding) +{ + signed char mode; + + switch (rounding) { + case ROUND_EVEN: + mode = float_round_nearest_even; + break; + case ROUND_ZERO: + mode = float_round_to_zero; + break; + case ROUND_DOWN: + mode = float_round_down; + break; + case ROUND_UP: + mode = float_round_up; + break; + case ROUND_TIEAWAY: + mode = float_round_ties_away; + break; + default: + g_assert_not_reached(); + } + soft_status.float_rounding_mode = mode; +} + +static void parse_args(int argc, char *argv[]) +{ + int c; + int val; + int rounding = ROUND_EVEN; + + for (;;) { + c = getopt(argc, argv, "d:ho:p:r:t:zZ"); + if (c < 0) { + break; + } + switch (c) { + case 'd': + duration = atoi(optarg); + break; + case 'h': + usage_complete(argc, argv); + exit(EXIT_SUCCESS); + case 'o': + val = find_name(op_names, optarg); + if (val < 0) { + fprintf(stderr, "Unsupported op '%s'\n", optarg); + exit(EXIT_FAILURE); + } + operation = val; + break; + case 'p': + if (!strcmp(optarg, "single")) { + precision = PREC_SINGLE; + } else if (!strcmp(optarg, "double")) { + precision = PREC_DOUBLE; + } else { + fprintf(stderr, "Unsupported precision '%s'\n", optarg); + exit(EXIT_FAILURE); + } + break; + case 'r': + rounding = round_name_to_mode(optarg); + if (rounding < 0) { + fprintf(stderr, "fatal: invalid rounding mode '%s'\n", optarg); + exit(EXIT_FAILURE); + } + break; + case 't': + val = find_name(tester_names, optarg); + if (val < 0) { + fprintf(stderr, "Unsupported tester '%s'\n", optarg); + exit(EXIT_FAILURE); + } + tester = val; + break; + case 'z': + soft_status.flush_inputs_to_zero = 1; + break; + case 'Z': + soft_status.flush_to_zero = 1; + break; + } + } + + /* set precision and rounding mode based on the tester */ + switch (tester) { + case TESTER_HOST: + set_host_precision(rounding); + break; + case TESTER_SOFT: + set_soft_precision(rounding); + switch (precision) { + case PREC_SINGLE: + precision = PREC_FLOAT32; + break; + case PREC_DOUBLE: + precision = PREC_FLOAT64; + break; + default: + g_assert_not_reached(); + } + break; + default: + g_assert_not_reached(); + } +} + +static void pr_stats(void) +{ + printf("%.2f MFlops\n", (double)n_completed_ops / ns_elapsed * 1e3); +} + +int main(int argc, char *argv[]) +{ + parse_args(argc, argv); + run_bench(); + pr_stats(); + return 0; +} diff --git a/tests/qht-bench.c b/tests/qht-bench.c index 2089e2bed1..ab4e708180 100644 --- a/tests/qht-bench.c +++ b/tests/qht-bench.c @@ -9,7 +9,7 @@ #include "qemu/atomic.h" #include "qemu/qht.h" #include "qemu/rcu.h" -#include "exec/tb-hash-xx.h" +#include "qemu/xxhash.h" struct thread_stats { size_t rd; @@ -72,6 +72,7 @@ static const char commands_string[] = " -n = number of threads\n" "\n" " -o = offset at which keys start\n" + " -p = precompute hashes\n" "\n" " -g = set -s,-k,-K,-l,-r to the same value\n" " -s = initial size hint\n" @@ -104,7 +105,7 @@ static bool is_equal(const void *ap, const void *bp) static uint32_t h(unsigned long v) { - return tb_hash_func7(v, 0, 0, 0, 0); + return qemu_xxhash2(v); } static uint32_t hval(unsigned long v) diff --git a/util/qsp.c b/util/qsp.c index a848b09c6d..410f1ba004 100644 --- a/util/qsp.c +++ b/util/qsp.c @@ -61,7 +61,7 @@ #include "qemu/timer.h" #include "qemu/qht.h" #include "qemu/rcu.h" -#include "exec/tb-hash-xx.h" +#include "qemu/xxhash.h" enum QSPType { QSP_MUTEX, @@ -135,13 +135,13 @@ QemuCondWaitFunc qemu_cond_wait_func = qemu_cond_wait_impl; * without it we still get a pretty unique hash. */ static inline -uint32_t do_qsp_callsite_hash(const QSPCallSite *callsite, uint64_t a) +uint32_t do_qsp_callsite_hash(const QSPCallSite *callsite, uint64_t ab) { - uint64_t b = (uint64_t)(uintptr_t)callsite->obj; + uint64_t cd = (uint64_t)(uintptr_t)callsite->obj; uint32_t e = callsite->line; uint32_t f = callsite->type; - return tb_hash_func7(a, b, e, f, 0); + return qemu_xxhash6(ab, cd, e, f); } static inline @@ -169,11 +169,11 @@ static uint32_t qsp_entry_no_thread_hash(const QSPEntry *entry) static uint32_t qsp_entry_no_thread_obj_hash(const QSPEntry *entry) { const QSPCallSite *callsite = entry->callsite; - uint64_t a = g_str_hash(callsite->file); - uint64_t b = callsite->line; + uint64_t ab = g_str_hash(callsite->file); + uint64_t cd = callsite->line; uint32_t e = callsite->type; - return tb_hash_func7(a, b, e, 0, 0); + return qemu_xxhash5(ab, cd, e); } static bool qsp_callsite_cmp(const void *ap, const void *bp) |