[FFmpeg-devel] [PATCH 1/2] tests/checkasm/blockdsp: Fix unaligned stores

[FFmpeg-devel] [PATCH 1/2] tests/checkasm/blockdsp: Fix unaligned stores

[Andreas Rheinhardt]

Fixes this test with UBSan (and maybe also on arches on which
unaligned stores trap).

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
---
 tests/checkasm/blockdsp.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/tests/checkasm/blockdsp.c b/tests/checkasm/blockdsp.c
index 19d69b8687..f6f25f773e 100644
--- a/tests/checkasm/blockdsp.c
+++ b/tests/checkasm/blockdsp.c
@@ -36,8 +36,7 @@ typedef struct {
 
 #define randomize_buffers(size)             \
     do {                                    \
-        int i;                              \
-        for (i = 0; i < size; i++) {        \
+        for (int i = 0; i < size; i += 2) { \
             uint16_t r = rnd();             \
             AV_WN16A(buf0 + i, r);          \
             AV_WN16A(buf1 + i, r);          \
-- 
2.40.1

_______________________________________________
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".

[FFmpeg-devel] [PATCH 2/2] tests/checkasm/blockdsp: Speed up randomizing buffer

[Andreas Rheinhardt]

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
---
 tests/checkasm/blockdsp.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/tests/checkasm/blockdsp.c b/tests/checkasm/blockdsp.c
index f6f25f773e..4cd0fa3c82 100644
--- a/tests/checkasm/blockdsp.c
+++ b/tests/checkasm/blockdsp.c
@@ -36,10 +36,10 @@ typedef struct {
 
 #define randomize_buffers(size)             \
     do {                                    \
-        for (int i = 0; i < size; i += 2) { \
-            uint16_t r = rnd();             \
-            AV_WN16A(buf0 + i, r);          \
-            AV_WN16A(buf1 + i, r);          \
+        for (int i = 0; i < size; i += 4) { \
+            uint32_t r = rnd();             \
+            AV_WN32A(buf0 + i, r);          \
+            AV_WN32A(buf1 + i, r);          \
         }                                   \
     } while (0)
 
-- 
2.40.1

_______________________________________________
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".

Re: [FFmpeg-devel] [PATCH 1/2] tests/checkasm/blockdsp: Fix unaligned stores

[Andreas Rheinhardt]

Andreas Rheinhardt:
> Fixes this test with UBSan (and maybe also on arches on which
> unaligned stores trap).
> 
> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
> ---
>  tests/checkasm/blockdsp.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/tests/checkasm/blockdsp.c b/tests/checkasm/blockdsp.c
> index 19d69b8687..f6f25f773e 100644
> --- a/tests/checkasm/blockdsp.c
> +++ b/tests/checkasm/blockdsp.c
> @@ -36,8 +36,7 @@ typedef struct {
>  
>  #define randomize_buffers(size)             \
>      do {                                    \
> -        int i;                              \
> -        for (i = 0; i < size; i++) {        \
> +        for (int i = 0; i < size; i += 2) { \
>              uint16_t r = rnd();             \
>              AV_WN16A(buf0 + i, r);          \
>              AV_WN16A(buf1 + i, r);          \

On second thought, this test seems even more broken: It only randomizes
size bytes (or size + 1 bytes on master), but the actual block
dimensions are size x size. Had size x size bytes been randomized, there
would have been a stack buffer overflow.

- Andreas

_______________________________________________
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".

Re: [FFmpeg-devel] [PATCH 1/2] tests/checkasm/blockdsp: Fix unaligned stores

[James Almer]

On 5/7/2024 7:29 AM, Andreas Rheinhardt wrote:
> Fixes this test with UBSan (and maybe also on arches on which
> unaligned stores trap).
> 
> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
> ---
>   tests/checkasm/blockdsp.c | 3 +--
>   1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/tests/checkasm/blockdsp.c b/tests/checkasm/blockdsp.c
> index 19d69b8687..f6f25f773e 100644
> --- a/tests/checkasm/blockdsp.c
> +++ b/tests/checkasm/blockdsp.c
> @@ -36,8 +36,7 @@ typedef struct {
>   
>   #define randomize_buffers(size)             \
>       do {                                    \
> -        int i;                              \
> -        for (i = 0; i < size; i++) {        \
> +        for (int i = 0; i < size; i += 2) { \
>               uint16_t r = rnd();             \
>               AV_WN16A(buf0 + i, r);          \
>               AV_WN16A(buf1 + i, r);          \

This is incorrect when buf0 and buf1 are uint16_t arrays, as is the case 
for the clear_block tests.
This function doesn't need to be called for fill_block_tab, hence my fix 
using memset.
_______________________________________________
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".