From: Pranav Kant via ffmpeg-devel <ffmpeg-devel@ffmpeg.org>
To: ffmpeg-devel@ffmpeg.org
Cc: Pranav Kant <prka@google.com>
Subject: Re: [FFmpeg-devel] [PATCH v5] Mark C globals with small code model
Date: Mon, 28 Apr 2025 18:19:19 -0700
Message-ID: <CAPvNhsKE2ZfXjq=tRagXGUehAN1WE=itsfsSNdyzgv-YUN74_A@mail.gmail.com> (raw)
In-Reply-To: <CAPvNhsK=TbjBtN2am+WHJ=7kcpVd0KE6Jb=qVjjcCrg+Qv2Qmw@mail.gmail.com>
Ping.
On Tue, Apr 15, 2025 at 4:22 PM Pranav Kant <prka@google.com> wrote:
> Hello again. Is there anything else I can do here?
>
> On Fri, Apr 4, 2025 at 11:40 AM Pranav Kant <prka@google.com> wrote:
>
>> Any thoughts on this?
>>
>> On Thu, Mar 20, 2025 at 5:30 PM Pranav Kant <prka@google.com> wrote:
>>
>>> Patch version v5:
>>> - Uses two new macros DECLARE_ASM_VAR (used for both external and inline
>>> asm) and DECLARE_EXTERNAL_ASM_VAR (used only for external asm)
>>> - I intend to remove explicit existing use of
>>> attribute_visibility_hidden in follow-up patch and instead use
>>> DECLARE_EXTERNAL_ASM_VAR for those variables
>>> - Other variables will be marked with these two new macros in a
>>> follow-up patch. I want to settle down on the infrastructure first with a
>>> handful of variables.
>>>
>>> Let me know if you have any questions.
>>>
>>> On Thu, Mar 20, 2025 at 5:28 PM Pranav Kant <prka@google.com> wrote:
>>>
>>>> By default, all globals in C/C++ compiled by clang are allocated
>>>> in non-large data sections. See [1] for background on code models.
>>>> For PIC (Position independent code), this is fine as long as binary is
>>>> small but as binary size increases, users maybe want to use medium/large
>>>> code models (-mcmodel=medium) which moves data in to large sections.
>>>> As data in these large sections cannot be accessed using PIC code
>>>> anymore (as it may be too far away), compiler ends up using a different
>>>> instruction sequence when building C/C++ code -- using GOT to access
>>>> these globals (which can be relaxed by linker at link time if binary
>>>> ends up being smaller). However, external assembly and inline assembly
>>>> continue to access these globals using older PC-relative addressing
>>>> which may not work because globals may be placed too far away.
>>>>
>>>> Introduce new macros for such variables that mark them with small code
>>>> model attribute. This ensures that these variables are never allocated
>>>> in large data sections, and continue to be validly accessed from
>>>> assembly
>>>> code.
>>>>
>>>> This patch should not have any affect on builds that use small code
>>>> model, which is the default mode.
>>>>
>>>> [1]
>>>> https://eli.thegreenplace.net/2012/01/03/understanding-the-x64-code-models
>>>>
>>>> Signed-off-by: Pranav Kant <prka@google.com>
>>>> ---
>>>> libavcodec/ac3dsp.h | 4 +++-
>>>> libavcodec/cabac.h | 4 +++-
>>>> libavcodec/x86/constants.h | 12 +++++++-----
>>>> libavutil/attributes.h | 6 ++++++
>>>> libavutil/attributes_internal.h | 16 ++++++++++++++++
>>>> libavutil/mem_internal.h | 13 +++++++++++++
>>>> 6 files changed, 48 insertions(+), 7 deletions(-)
>>>>
>>>> diff --git a/libavcodec/ac3dsp.h b/libavcodec/ac3dsp.h
>>>> index b1b2bced8f..a3c55a833b 100644
>>>> --- a/libavcodec/ac3dsp.h
>>>> +++ b/libavcodec/ac3dsp.h
>>>> @@ -25,11 +25,13 @@
>>>> #include <stddef.h>
>>>> #include <stdint.h>
>>>>
>>>> +#include "libavutil/mem_internal.h"
>>>> +
>>>> /**
>>>> * Number of mantissa bits written for each bap value.
>>>> * bap values with fractional bits are set to 0 and are calculated
>>>> separately.
>>>> */
>>>> -extern const uint16_t ff_ac3_bap_bits[16];
>>>> +extern DECLARE_EXTERNAL_ASM_VAR(16, const uint16_t,
>>>> ff_ac3_bap_bits)[16];
>>>>
>>>> typedef struct AC3DSPContext {
>>>> /**
>>>> diff --git a/libavcodec/cabac.h b/libavcodec/cabac.h
>>>> index 38d06b2842..df352258c6 100644
>>>> --- a/libavcodec/cabac.h
>>>> +++ b/libavcodec/cabac.h
>>>> @@ -29,7 +29,9 @@
>>>>
>>>> #include <stdint.h>
>>>>
>>>> -extern const uint8_t ff_h264_cabac_tables[512 + 4*2*64 + 4*64 + 63];
>>>> +#include "libavutil/mem_internal.h"
>>>> +
>>>> +extern DECLARE_ASM_VAR(1, const uint8_t, ff_h264_cabac_tables)[512 +
>>>> 4*2*64 + 4*64 + 63];
>>>> #define H264_NORM_SHIFT_OFFSET 0
>>>> #define H264_LPS_RANGE_OFFSET 512
>>>> #define H264_MLPS_STATE_OFFSET 1024
>>>> diff --git a/libavcodec/x86/constants.h b/libavcodec/x86/constants.h
>>>> index 0c6bf41fa0..2561302604 100644
>>>> --- a/libavcodec/x86/constants.h
>>>> +++ b/libavcodec/x86/constants.h
>>>> @@ -23,13 +23,14 @@
>>>>
>>>> #include <stdint.h>
>>>>
>>>> +#include "libavutil/mem_internal.h"
>>>> #include "libavutil/x86/asm.h"
>>>>
>>>> -extern const ymm_reg ff_pw_1;
>>>> +extern DECLARE_EXTERNAL_ASM_VAR(32, const ymm_reg, ff_pw_1);
>>>> extern const ymm_reg ff_pw_2;
>>>> extern const xmm_reg ff_pw_3;
>>>> -extern const ymm_reg ff_pw_4;
>>>> -extern const xmm_reg ff_pw_5;
>>>> +extern DECLARE_ASM_VAR(32, const ymm_reg, ff_pw_4);
>>>> +extern DECLARE_ASM_VAR(16, const xmm_reg, ff_pw_5);
>>>> extern const xmm_reg ff_pw_8;
>>>> extern const xmm_reg ff_pw_9;
>>>> extern const uint64_t ff_pw_15;
>>>> @@ -43,7 +44,7 @@ extern const uint64_t ff_pw_128;
>>>> extern const ymm_reg ff_pw_255;
>>>> extern const ymm_reg ff_pw_256;
>>>> extern const ymm_reg ff_pw_512;
>>>> -extern const ymm_reg ff_pw_1023;
>>>> +extern DECLARE_EXTERNAL_ASM_VAR(32, const ymm_reg, ff_pw_1023);
>>>> extern const ymm_reg ff_pw_1024;
>>>> extern const ymm_reg ff_pw_2048;
>>>> extern const ymm_reg ff_pw_4095;
>>>> @@ -52,9 +53,10 @@ extern const ymm_reg ff_pw_8192;
>>>> extern const ymm_reg ff_pw_m1;
>>>>
>>>> extern const ymm_reg ff_pb_0;
>>>> -extern const ymm_reg ff_pb_1;
>>>> +extern DECLARE_EXTERNAL_ASM_VAR(32, const ymm_reg, ff_pb_1);
>>>> extern const ymm_reg ff_pb_2;
>>>> extern const ymm_reg ff_pb_3;
>>>> +extern DECLARE_ASM_VAR(32, const xmm_reg, ff_pb_15);
>>>> extern const ymm_reg ff_pb_80;
>>>> extern const ymm_reg ff_pb_FE;
>>>> extern const uint64_t ff_pb_FC;
>>>> diff --git a/libavutil/attributes.h b/libavutil/attributes.h
>>>> index 04c615c952..dfc35fa31e 100644
>>>> --- a/libavutil/attributes.h
>>>> +++ b/libavutil/attributes.h
>>>> @@ -40,6 +40,12 @@
>>>> # define AV_HAS_BUILTIN(x) 0
>>>> #endif
>>>>
>>>> +#ifdef __has_attribute
>>>> +# define AV_HAS_ATTRIBUTE(x) __has_attribute(x)
>>>> +#else
>>>> +# define AV_HAS_ATTRIBUTE(x) 0
>>>> +#endif
>>>> +
>>>> #ifndef av_always_inline
>>>> #if AV_GCC_VERSION_AT_LEAST(3,1)
>>>> # define av_always_inline __attribute__((always_inline)) inline
>>>> diff --git a/libavutil/attributes_internal.h
>>>> b/libavutil/attributes_internal.h
>>>> index bc85ce77ff..c557fa0af0 100644
>>>> --- a/libavutil/attributes_internal.h
>>>> +++ b/libavutil/attributes_internal.h
>>>> @@ -19,6 +19,7 @@
>>>> #ifndef AVUTIL_ATTRIBUTES_INTERNAL_H
>>>> #define AVUTIL_ATTRIBUTES_INTERNAL_H
>>>>
>>>> +#include "config.h"
>>>> #include "attributes.h"
>>>>
>>>> #if (AV_GCC_VERSION_AT_LEAST(4,0) || defined(__clang__)) &&
>>>> (defined(__ELF__) || defined(__MACH__))
>>>> @@ -33,4 +34,19 @@
>>>>
>>>> #define EXTERN extern attribute_visibility_hidden
>>>>
>>>> +/**
>>>> + * Some globals defined in C files are used from hardcoded asm that
>>>> assumes small
>>>> + * code model (that is, accessing these globals without GOT). This is
>>>> a problem
>>>> + * when FFMpeg is built with medium code model (-mcmodel=medium) which
>>>> allocates
>>>> + * all globals in a data section that's unreachable with PC relative
>>>> instructions
>>>> + * (small code model instruction sequence). We mark all such globals
>>>> with this
>>>> + * attribute_mcmodel_small to ensure assembly accessible globals
>>>> continue to be
>>>> + * allocated in sections reachable from PC relative instructions.
>>>> + */
>>>> +#if ARCH_X86_64 && defined(__ELF__) && AV_HAS_ATTRIBUTE(model)
>>>> +# define attribute_mcmodel_small __attribute__((model("small")))
>>>> +#else
>>>> +# define attribute_mcmodel_small
>>>> +#endif
>>>> +
>>>> #endif /* AVUTIL_ATTRIBUTES_INTERNAL_H */
>>>> diff --git a/libavutil/mem_internal.h b/libavutil/mem_internal.h
>>>> index c027fa51c3..efb4c89c39 100644
>>>> --- a/libavutil/mem_internal.h
>>>> +++ b/libavutil/mem_internal.h
>>>> @@ -29,6 +29,7 @@
>>>> #endif
>>>>
>>>> #include "attributes.h"
>>>> +#include "attributes_internal.h"
>>>> #include "macros.h"
>>>>
>>>> /**
>>>> @@ -113,6 +114,18 @@
>>>> #define DECLARE_ALIGNED_32(t,v) DECLARE_ALIGNED_T(ALIGN_32, t, v)
>>>> #define DECLARE_ALIGNED_64(t,v) DECLARE_ALIGNED_T(ALIGN_64, t, v)
>>>>
>>>> +// DECLARE_ASM_VAR used for variables accessed by inline asm
>>>> +// and external assembly
>>>> +#define DECLARE_ASM_VAR(n,t,v) \
>>>> + attribute_mcmodel_small \
>>>> + alignas(n) t v
>>>> +// DECLARE_EXTERNAL_ASM_VAR used for variables exclusively
>>>> +// accessed by external assembly
>>>> +#define DECLARE_EXTERNAL_ASM_VAR(n,t,v) \
>>>> + attribute_visibility_hidden \
>>>> + attribute_mcmodel_small \
>>>> + alignas(n) t v
>>>> +
>>>> // Some broken preprocessors need a second expansion
>>>> // to be forced to tokenize __VA_ARGS__
>>>> #define E1(x) x
>>>> --
>>>> 2.49.0.395.g12beb8f557-goog
>>>>
>>>>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next prev parent reply other threads:[~2025-04-29 1:19 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-02-25 21:37 [FFmpeg-devel] [PATCH] " Pranav Kant via ffmpeg-devel
2025-02-25 23:03 ` Andreas Rheinhardt
2025-02-26 19:44 ` Pranav Kant via ffmpeg-devel
2025-02-26 19:45 ` Pranav Kant via ffmpeg-devel
2025-02-27 3:36 ` James Zern via ffmpeg-devel
2025-02-28 1:14 ` Michael Niedermayer
2025-03-07 2:13 ` Pranav Kant via ffmpeg-devel
2025-03-03 20:56 ` [FFmpeg-devel] [PATCH v3] " Pranav Kant via ffmpeg-devel
2025-03-11 19:17 ` [FFmpeg-devel] [PATCH v4] " Pranav Kant via ffmpeg-devel
2025-03-11 19:18 ` Pranav Kant via ffmpeg-devel
2025-03-11 23:45 ` Andreas Rheinhardt
2025-03-12 7:05 ` Martin Storsjö
2025-03-12 7:17 ` Andreas Rheinhardt
2025-03-12 7:47 ` Martin Storsjö
2025-03-14 0:16 ` Pranav Kant via ffmpeg-devel
2025-03-21 0:28 ` [FFmpeg-devel] [PATCH v5] " Pranav Kant via ffmpeg-devel
2025-03-21 0:30 ` Pranav Kant via ffmpeg-devel
2025-04-04 18:40 ` Pranav Kant via ffmpeg-devel
2025-04-15 23:22 ` Pranav Kant via ffmpeg-devel
2025-04-29 1:19 ` Pranav Kant via ffmpeg-devel [this message]
2025-05-06 23:24 ` [FFmpeg-devel] [PATCH v6] " Pranav Kant via ffmpeg-devel
2025-05-06 23:27 ` Pranav Kant via ffmpeg-devel
2025-05-07 12:15 ` [FFmpeg-devel] [PATCH v5] " Rémi Denis-Courmont
2025-02-28 2:31 ` [FFmpeg-devel] [PATCH] " Lynne
2025-03-07 2:14 ` Pranav Kant via ffmpeg-devel
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAPvNhsKE2ZfXjq=tRagXGUehAN1WE=itsfsSNdyzgv-YUN74_A@mail.gmail.com' \
--to=ffmpeg-devel@ffmpeg.org \
--cc=prka@google.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git