Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(2)

Issues Search
    My Issues | Starred     Open | Closed | All

Unified Diff: simd/jsimd_arm64.c

Issue 434123003: Add ARM64 SIMD support to libjpeg_turbo (Closed) Base URL: http://src.chromium.org/svn/trunk/deps/third_party/libjpeg_turbo
Patch Set: Add neon fixes Created 6 years, 4 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View side-by-side diff with in-line comments
Download patch
« no previous file with comments | « libjpeg.gyp ('k') | simd/jsimd_arm64_neon.S » ('j') | no next file with comments »
Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
Index: simd/jsimd_arm64.c
diff --git a/simd/jsimd_arm.c b/simd/jsimd_arm64.c
similarity index 64%
copy from simd/jsimd_arm.c
copy to simd/jsimd_arm64.c
index bd717a409798c92e447c91e15235298cb5c81293..65724cb9ff94a816ffd2706e140333f28656039f 100644
--- a/simd/jsimd_arm.c
+++ b/simd/jsimd_arm64.c
@@ -1,18 +1,16 @@
/*
- * jsimd_arm.c
+ * jsimd_arm64.c
*
* Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2009-2011 D. R. Commander
+ * Copyright 2009-2011, 2013-2014 D. R. Commander
*
* Based on the x86 SIMD extension for IJG JPEG library,
* Copyright (C) 1999-2006, MIYASAKA Masaru.
* For conditions of distribution and use, see copyright notice in jsimdext.inc
*
* This file contains the interface between the "normal" portions
- * of the library and the SIMD implementations when running on
- * ARM architecture.
- *
- * Based on the stubs from 'jsimd_none.c'
+ * of the library and the SIMD implementations when running on a
+ * 64-bit ARM architecture.
*/
#define JPEG_INTERNALS
@@ -29,104 +27,35 @@
static unsigned int simd_support = ~0;
-#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
-
-#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT (1024 * 1024)
-
-LOCAL(int)
-check_feature (char *buffer, char *feature)
-{
- char *p;
- if (*feature == 0)
- return 0;
- if (strncmp(buffer, "Features", 8) != 0)
- return 0;
- buffer += 8;
- while (isspace(*buffer))
- buffer++;
-
- /* Check if 'feature' is present in the buffer as a separate word */
- while ((p = strstr(buffer, feature))) {
- if (p > buffer && !isspace(*(p - 1))) {
- buffer++;
- continue;
- }
- p += strlen(feature);
- if (*p != 0 && !isspace(*p)) {
- buffer++;
- continue;
- }
- return 1;
- }
- return 0;
-}
-
-LOCAL(int)
-parse_proc_cpuinfo (int bufsize)
-{
- char *buffer = (char *)malloc(bufsize);
- FILE *fd;
- simd_support = 0;
-
- if (!buffer)
- return 0;
-
- fd = fopen("/proc/cpuinfo", "r");
- if (fd) {
- while (fgets(buffer, bufsize, fd)) {
- if (!strchr(buffer, '\n') && !feof(fd)) {
- /* "impossible" happened - insufficient size of the buffer! */
- fclose(fd);
- free(buffer);
- return 0;
- }
- if (check_feature(buffer, "neon"))
- simd_support |= JSIMD_ARM_NEON;
- }
- fclose(fd);
- }
- free(buffer);
- return 1;
-}
-
-#endif
-
/*
* Check what SIMD accelerations are supported.
*
* FIXME: This code is racy under a multi-threaded environment.
*/
+
+/*
+ * ARMv8 architectures support NEON extensions by default.
+ * It is no longer optional as it was with ARMv7.
+ */
+
+
LOCAL(void)
init_simd (void)
{
char *env = NULL;
-#if !defined(__ARM_NEON__) && defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
- int bufsize = 1024; /* an initial guess for the line buffer size limit */
-#endif
if (simd_support != ~0U)
return;
simd_support = 0;
-#if defined(__ARM_NEON__)
simd_support |= JSIMD_ARM_NEON;
-#elif defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
- /* We still have a chance to use NEON regardless of globally used
- * -mcpu/-mfpu options passed to gcc by performing runtime detection via
- * /proc/cpuinfo parsing on linux/android */
- while (!parse_proc_cpuinfo(bufsize)) {
- bufsize *= 2;
- if (bufsize > SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT)
- break;
- }
-#endif
/* Force different settings through environment variables */
- env = getenv("JSIMD_FORCE_ARM_NEON");
+ env = getenv("JSIMD_FORCENEON");
if ((env != NULL) && (strcmp(env, "1") == 0))
simd_support &= JSIMD_ARM_NEON;
- env = getenv("JSIMD_FORCE_NO_SIMD");
+ env = getenv("JSIMD_FORCENONE");
if ((env != NULL) && (strcmp(env, "1") == 0))
simd_support = 0;
}
@@ -136,17 +65,6 @@ jsimd_can_rgb_ycc (void)
{
init_simd();
- /* The code is optimised for these values only */
- if (BITS_IN_JSAMPLE != 8)
- return 0;
- if (sizeof(JDIMENSION) != 4)
- return 0;
- if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
- return 0;
-
- if (simd_support & JSIMD_ARM_NEON)
- return 1;
-
return 0;
}
@@ -170,6 +88,24 @@ jsimd_can_ycc_rgb (void)
return 0;
if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
return 0;
+
+ if (simd_support & JSIMD_ARM_NEON)
+ return 1;
+
+ return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+ init_simd();
+
+ /* The code is optimised for these values only */
+ if (BITS_IN_JSAMPLE != 8)
+ return 0;
+ if (sizeof(JDIMENSION) != 4)
+ return 0;
+
if (simd_support & JSIMD_ARM_NEON)
return 1;
@@ -181,40 +117,6 @@ jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
JDIMENSION output_row, int num_rows)
{
- void (*neonfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
-
- switch(cinfo->in_color_space)
- {
- case JCS_EXT_RGB:
- neonfct=jsimd_extrgb_ycc_convert_neon;
- break;
- case JCS_EXT_RGBX:
- case JCS_EXT_RGBA:
- neonfct=jsimd_extrgbx_ycc_convert_neon;
- break;
- case JCS_EXT_BGR:
- neonfct=jsimd_extbgr_ycc_convert_neon;
- break;
- case JCS_EXT_BGRX:
- case JCS_EXT_BGRA:
- neonfct=jsimd_extbgrx_ycc_convert_neon;
- break;
- case JCS_EXT_XBGR:
- case JCS_EXT_ABGR:
- neonfct=jsimd_extxbgr_ycc_convert_neon;
- break;
- case JCS_EXT_XRGB:
- case JCS_EXT_ARGB:
- neonfct=jsimd_extxrgb_ycc_convert_neon;
- break;
- default:
- neonfct=jsimd_extrgb_ycc_convert_neon;
- break;
- }
-
- if (simd_support & JSIMD_ARM_NEON)
- neonfct(cinfo->image_width, input_buf,
- output_buf, output_row, num_rows);
}
GLOBAL(void)
@@ -231,8 +133,7 @@ jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
{
void (*neonfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
- switch(cinfo->out_color_space)
- {
+ switch(cinfo->out_color_space) {
case JCS_EXT_RGB:
neonfct=jsimd_ycc_extrgb_convert_neon;
break;
@@ -255,14 +156,23 @@ jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
case JCS_EXT_ARGB:
neonfct=jsimd_ycc_extxrgb_convert_neon;
break;
- default:
+ default:
neonfct=jsimd_ycc_extrgb_convert_neon;
break;
}
if (simd_support & JSIMD_ARM_NEON)
- neonfct(cinfo->output_width, input_buf,
- input_row, output_buf, num_rows);
+ neonfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+ JSAMPIMAGE input_buf, JDIMENSION input_row,
+ JSAMPARRAY output_buf, int num_rows)
+{
+ if (simd_support & JSIMD_ARM_NEON)
+ jsimd_ycc_rgb565_convert_neon(cinfo->output_width, input_buf, input_row,
+ output_buf, num_rows);
}
GLOBAL(int)
@@ -311,7 +221,7 @@ jsimd_can_h2v1_upsample (void)
GLOBAL(void)
jsimd_h2v2_upsample (j_decompress_ptr cinfo,
- jpeg_component_info * compptr,
+ jpeg_component_info * compptr,
JSAMPARRAY input_data,
JSAMPARRAY * output_data_ptr)
{
@@ -319,7 +229,7 @@ jsimd_h2v2_upsample (j_decompress_ptr cinfo,
GLOBAL(void)
jsimd_h2v1_upsample (j_decompress_ptr cinfo,
- jpeg_component_info * compptr,
+ jpeg_component_info * compptr,
JSAMPARRAY input_data,
JSAMPARRAY * output_data_ptr)
{
@@ -338,21 +248,12 @@ jsimd_can_h2v1_fancy_upsample (void)
{
init_simd();
- /* The code is optimised for these values only */
- if (BITS_IN_JSAMPLE != 8)
- return 0;
- if (sizeof(JDIMENSION) != 4)
- return 0;
-
- if (simd_support & JSIMD_ARM_NEON)
- return 1;
-
return 0;
}
GLOBAL(void)
jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
- jpeg_component_info * compptr,
+ jpeg_component_info * compptr,
JSAMPARRAY input_data,
JSAMPARRAY * output_data_ptr)
{
@@ -360,13 +261,10 @@ jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
GLOBAL(void)
jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
- jpeg_component_info * compptr,
+ jpeg_component_info * compptr,
JSAMPARRAY input_data,
JSAMPARRAY * output_data_ptr)
{
- if (simd_support & JSIMD_ARM_NEON)
- jsimd_h2v1_fancy_upsample_neon(cinfo->max_v_samp_factor,
- compptr->downsampled_width, input_data, output_data_ptr);
}
GLOBAL(int)
@@ -406,19 +304,6 @@ jsimd_can_convsamp (void)
{
init_simd();
- /* The code is optimised for these values only */
- if (DCTSIZE != 8)
- return 0;
- if (BITS_IN_JSAMPLE != 8)
- return 0;
- if (sizeof(JDIMENSION) != 4)
- return 0;
- if (sizeof(DCTELEM) != 2)
- return 0;
-
- if (simd_support & JSIMD_ARM_NEON)
- return 1;
-
return 0;
}
@@ -434,8 +319,6 @@ GLOBAL(void)
jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
DCTELEM * workspace)
{
- if (simd_support & JSIMD_ARM_NEON)
- jsimd_convsamp_neon(sample_data, start_col, workspace);
}
GLOBAL(void)
@@ -457,15 +340,6 @@ jsimd_can_fdct_ifast (void)
{
init_simd();
- /* The code is optimised for these values only */
- if (DCTSIZE != 8)
- return 0;
- if (sizeof(DCTELEM) != 2)
- return 0;
-
- if (simd_support & JSIMD_ARM_NEON)
- return 1;
-
return 0;
}
@@ -485,8 +359,6 @@ jsimd_fdct_islow (DCTELEM * data)
GLOBAL(void)
jsimd_fdct_ifast (DCTELEM * data)
{
- if (simd_support & JSIMD_ARM_NEON)
- jsimd_fdct_ifast_neon(data);
}
GLOBAL(void)
@@ -499,17 +371,6 @@ jsimd_can_quantize (void)
{
init_simd();
- /* The code is optimised for these values only */
- if (DCTSIZE != 8)
- return 0;
- if (sizeof(JCOEF) != 2)
- return 0;
- if (sizeof(DCTELEM) != 2)
- return 0;
-
- if (simd_support & JSIMD_ARM_NEON)
- return 1;
-
return 0;
}
@@ -525,8 +386,6 @@ GLOBAL(void)
jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors,
DCTELEM * workspace)
{
- if (simd_support & JSIMD_ARM_NEON)
- jsimd_quantize_neon(coef_block, divisors, workspace);
}
GLOBAL(void)
@@ -552,7 +411,7 @@ jsimd_can_idct_2x2 (void)
if (sizeof(ISLOW_MULT_TYPE) != 2)
return 0;
- if ((simd_support & JSIMD_ARM_NEON))
+ if (simd_support & JSIMD_ARM_NEON)
return 1;
return 0;
@@ -575,7 +434,7 @@ jsimd_can_idct_4x4 (void)
if (sizeof(ISLOW_MULT_TYPE) != 2)
return 0;
- if ((simd_support & JSIMD_ARM_NEON))
+ if (simd_support & JSIMD_ARM_NEON)
return 1;
return 0;
@@ -586,8 +445,9 @@ jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
JCOEFPTR coef_block, JSAMPARRAY output_buf,
JDIMENSION output_col)
{
- if ((simd_support & JSIMD_ARM_NEON))
- jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf, output_col);
+ if (simd_support & JSIMD_ARM_NEON)
+ jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf,
+ output_col);
}
GLOBAL(void)
@@ -595,8 +455,9 @@ jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
JCOEFPTR coef_block, JSAMPARRAY output_buf,
JDIMENSION output_col)
{
- if ((simd_support & JSIMD_ARM_NEON))
- jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf, output_col);
+ if (simd_support & JSIMD_ARM_NEON)
+ jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf,
+ output_col);
}
GLOBAL(int)
@@ -641,7 +502,7 @@ jsimd_can_idct_ifast (void)
if (IFAST_SCALE_BITS != 2)
return 0;
- if ((simd_support & JSIMD_ARM_NEON))
+ if (simd_support & JSIMD_ARM_NEON)
return 1;
return 0;
@@ -657,26 +518,27 @@ jsimd_can_idct_float (void)
GLOBAL(void)
jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
- JCOEFPTR coef_block, JSAMPARRAY output_buf,
- JDIMENSION output_col)
+ JCOEFPTR coef_block, JSAMPARRAY output_buf,
+ JDIMENSION output_col)
{
- if ((simd_support & JSIMD_ARM_NEON))
- jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf, output_col);
+ if (simd_support & JSIMD_ARM_NEON)
+ jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf,
+ output_col);
}
GLOBAL(void)
jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr,
- JCOEFPTR coef_block, JSAMPARRAY output_buf,
- JDIMENSION output_col)
+ JCOEFPTR coef_block, JSAMPARRAY output_buf,
+ JDIMENSION output_col)
{
- if ((simd_support & JSIMD_ARM_NEON))
- jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf, output_col);
+ if (simd_support & JSIMD_ARM_NEON)
+ jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf,
+ output_col);
}
GLOBAL(void)
jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr,
- JCOEFPTR coef_block, JSAMPARRAY output_buf,
- JDIMENSION output_col)
+ JCOEFPTR coef_block, JSAMPARRAY output_buf,
+ JDIMENSION output_col)
{
}
-
« no previous file with comments | « libjpeg.gyp ('k') | simd/jsimd_arm64_neon.S » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698