[Mesa-dev] [PATCH 2/2] i965: add runtime check for SSSE3 rgba8_copy
Timothy Arceri
t_arceri at yahoo.com.au
Thu Nov 6 13:22:47 PST 2014
On Thu, 2014-11-06 at 10:03 -0800, Matt Turner wrote:
> On Thu, Nov 6, 2014 at 4:20 AM, Timothy Arceri <t_arceri at yahoo.com.au> wrote:
> > Also cleans up some if statements in the *faster functions.
> >
> > Callgrind cpu usage results from pts benchmarks:
> >
> > For ytile_copy_faster()
> >
> > Nexuiz 1.6.1: 2.16% -> 1.20%
> >
> > Signed-off-by: Timothy Arceri <t_arceri at yahoo.com.au>
> > ---
> > src/mesa/Makefile.am | 8 +++
> > src/mesa/drivers/dri/i965/intel_tex_subimage.c | 82 ++++++--------------------
> > src/mesa/main/fast_rgba8_copy.c | 78 ++++++++++++++++++++++++
> > src/mesa/main/fast_rgba8_copy.h | 37 ++++++++++++
> > 4 files changed, 141 insertions(+), 64 deletions(-)
> > create mode 100644 src/mesa/main/fast_rgba8_copy.c
> > create mode 100644 src/mesa/main/fast_rgba8_copy.h
> >
> > diff --git a/src/mesa/Makefile.am b/src/mesa/Makefile.am
> > index e71bccb..2402096 100644
> > --- a/src/mesa/Makefile.am
> > +++ b/src/mesa/Makefile.am
> > @@ -107,6 +107,10 @@ AM_CXXFLAGS = $(LLVM_CFLAGS) $(VISIBILITY_CXXFLAGS)
> >
> > ARCH_LIBS =
> >
> > +if SSSE3_SUPPORTED
> > +ARCH_LIBS += libmesa_ssse3.la
> > +endif
> > +
> > if SSE41_SUPPORTED
> > ARCH_LIBS += libmesa_sse41.la
> > endif
> > @@ -154,6 +158,10 @@ libmesa_sse41_la_SOURCES = \
> > main/streaming-load-memcpy.c
> > libmesa_sse41_la_CFLAGS = $(AM_CFLAGS) -msse4.1
> >
> > +libmesa_ssse3_la_SOURCES = \
> > + main/fast_rgba8_copy.c
> > +libmesa_ssse3_la_CFLAGS = $(AM_CFLAGS) -mssse3
> > +
> > pkgconfigdir = $(libdir)/pkgconfig
> > pkgconfig_DATA = gl.pc
> >
> > diff --git a/src/mesa/drivers/dri/i965/intel_tex_subimage.c b/src/mesa/drivers/dri/i965/intel_tex_subimage.c
> > index cb5738a..0deeb75 100644
> > --- a/src/mesa/drivers/dri/i965/intel_tex_subimage.c
> > +++ b/src/mesa/drivers/dri/i965/intel_tex_subimage.c
> > @@ -27,6 +27,7 @@
> > **************************************************************************/
> >
> > #include "main/bufferobj.h"
> > +#include "main/fast_rgba8_copy.h"
> > #include "main/image.h"
> > #include "main/macros.h"
> > #include "main/mtypes.h"
> > @@ -42,9 +43,7 @@
> > #include "intel_mipmap_tree.h"
> > #include "intel_blit.h"
> >
> > -#ifdef __SSSE3__
> > -#include <tmmintrin.h>
> > -#endif
> > +#include "x86/common_x86_asm.h"
> >
> > #define FILE_DEBUG_FLAG DEBUG_TEXTURE
> >
> > @@ -175,18 +174,6 @@ err:
> > return false;
> > }
> >
> > -#ifdef __SSSE3__
> > -static const uint8_t rgba8_permutation[16] =
> > - { 2,1,0,3, 6,5,4,7, 10,9,8,11, 14,13,12,15 };
> > -
> > -/* NOTE: dst must be 16 byte aligned */
> > -#define rgba8_copy_16(dst, src) \
> > - *(__m128i *)(dst) = _mm_shuffle_epi8( \
> > - (__m128i) _mm_loadu_ps((float *)(src)), \
> > - *(__m128i *) rgba8_permutation \
> > - )
> > -#endif
> > -
> > /**
> > * Copy RGBA to BGRA - swap R and B.
> > */
> > @@ -196,29 +183,6 @@ rgba8_copy(void *dst, const void *src, size_t bytes)
> > uint8_t *d = dst;
> > uint8_t const *s = src;
> >
> > -#ifdef __SSSE3__
> > - /* Fast copying for tile spans.
> > - *
> > - * As long as the destination texture is 16 aligned,
> > - * any 16 or 64 spans we get here should also be 16 aligned.
> > - */
> > -
> > - if (bytes == 16) {
> > - assert(!(((uintptr_t)dst) & 0xf));
> > - rgba8_copy_16(d+ 0, s+ 0);
> > - return dst;
> > - }
> > -
> > - if (bytes == 64) {
> > - assert(!(((uintptr_t)dst) & 0xf));
> > - rgba8_copy_16(d+ 0, s+ 0);
> > - rgba8_copy_16(d+16, s+16);
> > - rgba8_copy_16(d+32, s+32);
> > - rgba8_copy_16(d+48, s+48);
> > - return dst;
> > - }
> > -#endif
> > -
> > while (bytes >= 4) {
> > d[0] = s[2];
> > d[1] = s[1];
> > @@ -352,19 +316,8 @@ xtile_copy_faster(uint32_t x0, uint32_t x1, uint32_t x2, uint32_t x3,
> > mem_copy_fn mem_copy)
> > {
> > if (x0 == 0 && x3 == xtile_width && y0 == 0 && y1 == xtile_height) {
> > - if (mem_copy == memcpy)
> > - return xtile_copy(0, 0, xtile_width, xtile_width, 0, xtile_height,
> > - dst, src, src_pitch, swizzle_bit, memcpy);
> > - else if (mem_copy == rgba8_copy)
> > - return xtile_copy(0, 0, xtile_width, xtile_width, 0, xtile_height,
> > - dst, src, src_pitch, swizzle_bit, rgba8_copy);
> > - } else {
> > - if (mem_copy == memcpy)
> > - return xtile_copy(x0, x1, x2, x3, y0, y1,
> > - dst, src, src_pitch, swizzle_bit, memcpy);
> > - else if (mem_copy == rgba8_copy)
> > - return xtile_copy(x0, x1, x2, x3, y0, y1,
> > - dst, src, src_pitch, swizzle_bit, rgba8_copy);
> > + return xtile_copy(0, 0, xtile_width, xtile_width, 0, xtile_height,
> > + dst, src, src_pitch, swizzle_bit, mem_copy);
> > }
> > xtile_copy(x0, x1, x2, x3, y0, y1,
> > dst, src, src_pitch, swizzle_bit, mem_copy);
> > @@ -388,19 +341,8 @@ ytile_copy_faster(uint32_t x0, uint32_t x1, uint32_t x2, uint32_t x3,
> > mem_copy_fn mem_copy)
> > {
> > if (x0 == 0 && x3 == ytile_width && y0 == 0 && y1 == ytile_height) {
> > - if (mem_copy == memcpy)
> > - return ytile_copy(0, 0, ytile_width, ytile_width, 0, ytile_height,
> > - dst, src, src_pitch, swizzle_bit, memcpy);
> > - else if (mem_copy == rgba8_copy)
> > - return ytile_copy(0, 0, ytile_width, ytile_width, 0, ytile_height,
> > - dst, src, src_pitch, swizzle_bit, rgba8_copy);
> > - } else {
> > - if (mem_copy == memcpy)
> > - return ytile_copy(x0, x1, x2, x3, y0, y1,
> > - dst, src, src_pitch, swizzle_bit, memcpy);
> > - else if (mem_copy == rgba8_copy)
> > - return ytile_copy(x0, x1, x2, x3, y0, y1,
> > - dst, src, src_pitch, swizzle_bit, rgba8_copy);
>
> The purpose of this bizarre nested if mess is so that ytile_copy() can
> be inlined with memcpy/rgba8_copy inlined as well. This patch is going
> to prevent that from happening, and I don't think that's okay.
I'll have a closer look at this, my test didn't show a difference but
ytile_copy_faster was only called 40,000 times in my test so maybe
that's why I didn't see a difference.
>
> > + return ytile_copy(0, 0, ytile_width, ytile_width, 0, ytile_height,
> > + dst, src, src_pitch, swizzle_bit, mem_copy);
> > }
> > ytile_copy(x0, x1, x2, x3, y0, y1,
> > dst, src, src_pitch, swizzle_bit, mem_copy);
> > @@ -582,6 +524,12 @@ intel_texsubimage_tiled_memcpy(struct gl_context * ctx,
> > if (format == GL_BGRA) {
> > mem_copy = memcpy;
> > } else if (format == GL_RGBA) {
> > + #if defined(USE_SSSE3)
> > + if (cpu_has_ssse3) {
> > + mem_copy = _mesa_fast_rgba8_copy;
> > + }
> > + else
> > + #endif
> > mem_copy = rgba8_copy;
> > }
> > } else if ((texImage->TexFormat == MESA_FORMAT_R8G8B8A8_UNORM) ||
> > @@ -591,6 +539,12 @@ intel_texsubimage_tiled_memcpy(struct gl_context * ctx,
> > /* Copying from RGBA to BGRA is the same as BGRA to RGBA so we can
> > * use the same function.
> > */
> > + #if defined(USE_SSSE3)
>
> Don't indent the preprocessor tests.
>
> > + if (cpu_has_ssse3) {
> > + mem_copy = _mesa_fast_rgba8_copy;
> > + }
> > + else
> > + #endif
> > mem_copy = rgba8_copy;
> > } else if (format == GL_RGBA) {
> > mem_copy = memcpy;
> > diff --git a/src/mesa/main/fast_rgba8_copy.c b/src/mesa/main/fast_rgba8_copy.c
> > new file mode 100644
> > index 0000000..7ac3f3b
> > --- /dev/null
> > +++ b/src/mesa/main/fast_rgba8_copy.c
> > @@ -0,0 +1,78 @@
> > +/*
> > + * Copyright 2003 VMware, Inc.
> > + * All Rights Reserved.
>
> This was copied from the Intel driver. This copyright is not what you want.
Actually that is the copyright in intel_tex_subimage.c
>
> > + *
> > + * Permission is hereby granted, free of charge, to any person obtaining a
> > + * copy of this software and associated documentation files (the
> > + * "Software"), to deal in the Software without restriction, including
> > + * without limitation the rights to use, copy, modify, merge, publish,
> > + * distribute, sub license, and/or sell copies of the Software, and to
> > + * permit persons to whom the Software is furnished to do so, subject to
> > + * the following conditions:
> > + *
> > + * The above copyright notice and this permission notice (including the
> > + * next paragraph) shall be included in all copies or substantial portions
> > + * of the Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
> > + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> > + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
> > + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
> > + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
> > + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
> > + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
> > + *
> > + */
> > +
> > +#ifdef __SSSE3__
> > +#include "main/fast_rgba8_copy.h"
> > +#include <tmmintrin.h>
> > +
> > +static const uint8_t rgba8_permutation[16] =
> > + { 2,1,0,3, 6,5,4,7, 10,9,8,11, 14,13,12,15 };
> > +
> > +/* NOTE: dst must be 16 byte aligned */
> > +#define rgba8_copy_16(dst, src) \
> > + *(__m128i *)(dst) = _mm_shuffle_epi8( \
> > + (__m128i) _mm_loadu_ps((float *)(src)), \
> > + *(__m128i *) rgba8_permutation \
> > + )
> > +
> > +/* Fast copying for tile spans.
> > + *
> > + * As long as the destination texture is 16 aligned,
> > + * any 16 or 64 spans we get here should also be 16 aligned.
> > + */
> > +void *
> > +_mesa_fast_rgba8_copy(void *dst, const void *src, size_t bytes)
> > +{
> > + uint8_t *d = dst;
> > + uint8_t const *s = src;
> > +
> > + if (bytes == 16) {
> > + assert(!(((uintptr_t)dst) & 0xf));
> > + rgba8_copy_16(d+ 0, s+ 0);
> > + return dst;
> > + }
> > +
> > + if (bytes == 64) {
> > + assert(!(((uintptr_t)dst) & 0xf));
> > + rgba8_copy_16(d+ 0, s+ 0);
> > + rgba8_copy_16(d+16, s+16);
> > + rgba8_copy_16(d+32, s+32);
> > + rgba8_copy_16(d+48, s+48);
> > + return dst;
> > + }
> > +
> > + while (bytes >= 4) {
> > + d[0] = s[2];
> > + d[1] = s[1];
> > + d[2] = s[0];
> > + d[3] = s[3];
> > + d += 4;
> > + s += 4;
> > + bytes -= 4;
> > + }
> > + return dst;
> > +}
> > +#endif
> > diff --git a/src/mesa/main/fast_rgba8_copy.h b/src/mesa/main/fast_rgba8_copy.h
> > new file mode 100644
> > index 0000000..2154daa
> > --- /dev/null
> > +++ b/src/mesa/main/fast_rgba8_copy.h
> > @@ -0,0 +1,37 @@
> > +/*
> > + * Copyright 2003 VMware, Inc.
> > + * All Rights Reserved.
>
> Same.
>
> > + *
> > + * Permission is hereby granted, free of charge, to any person obtaining a
> > + * copy of this software and associated documentation files (the
> > + * "Software"), to deal in the Software without restriction, including
> > + * without limitation the rights to use, copy, modify, merge, publish,
> > + * distribute, sub license, and/or sell copies of the Software, and to
> > + * permit persons to whom the Software is furnished to do so, subject to
> > + * the following conditions:
> > + *
> > + * The above copyright notice and this permission notice (including the
> > + * next paragraph) shall be included in all copies or substantial portions
> > + * of the Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
> > + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> > + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
> > + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
> > + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
> > + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
> > + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
> > + *
> > + */
> > +
> > +#include <assert.h>
> > +#include <stdint.h>
> > +#include <stddef.h>
>
> I don't think you need these three includes for a single prototype.
Right I can move assert to the .c
>
> > +
> > +/* Fast copying for tile spans.
> > + *
> > + * As long as the destination texture is 16 aligned,
> > + * any 16 or 64 spans we get here should also be 16 aligned.
> > + */
> > +void *
> > +_mesa_fast_rgba8_copy(void *dst, const void *src, size_t n);
> > --
> > 1.9.3
> >
> > _______________________________________________
> > mesa-dev mailing list
> > mesa-dev at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/mesa-dev
More information about the mesa-dev
mailing list