• State Accepted Answer
  • +2 person also asked this people also asked this
  • Locked Locked
  • Replies 1 reply
  • Subscribers 351 subscribers
  • Views 195 views
  • Users 0 members are here
2025 survey

We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.

Options
How was your experience today?
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Unexpected result from svqdmulh_s32 with negative input values

Yevh Prill

Dear ARM Support Team,

I'm having trouble understanding how the svqdmulh_s32 intrinsic behaves when handling negative values.

Here's a specific case:

Fullscreen
1
2
3
int32_t a = 115998720;
int32_t b = -27704;
int32_t result = (a * b) >> 31; // Expected: -1497
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

However, when I use svqdmulh_s32, I get 2147483647 instead of -1497.

From what I understand, this intrinsic should behave similarly to the NEON vqdmulhq_s32, which gives the correct result for this case.

My build configuration:

Fullscreen
1
-g3 --debug -fno-inline -O0 -MD --target=aarch64-arm-none-eabi -march=armv8-a+sve2
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

Is it possible that I need to explicitly specify the target processor to get correct behavior? I haven't defined a specific CPU in the toolchain so far.

Any clarification would be greatly appreciated.

Best regards,
Yevh

0