Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Make assembly compatible with clang #720

Merged
merged 2 commits into from
Apr 1, 2022
Merged

Make assembly compatible with clang #720

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
145341d
Make assembly compatible with clang
Tombana Apr 1, 2022
98a63b5
Merge branch 'main' into arm32_asm_syntax
Tombana Apr 1, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 14 additions & 14 deletions larq_compute_engine/core/bgemm/kernels_arm32.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -324,29 +324,29 @@ void BinaryKernelNeon4x4(BinaryKernelParams<DstScalar, 4, 4>& params) {

// Load the clamp_max bound (in parallel with the shift)
"ldr r1, [%[params], #" RUY_STR(RUY_OFFSET_OUTPUT_TRANSFORM_CLAMP_MIN) "]\n"
"vdup q12.32, r1 \n" // clamp_min
"vdup.32 q12, r1 \n" // clamp_min

// Perform the backtransformation shift (in int32)
"vshl q8.s32, q8.s32, #1\n"
"vshl q9.s32, q9.s32, #1\n"
"vshl q10.s32, q10.s32, #1\n"
"vshl q11.s32, q11.s32, #1\n"
"vshl.s32 q8, q8, #1\n"
"vshl.s32 q9, q9, #1\n"
"vshl.s32 q10, q10, #1\n"
"vshl.s32 q11, q11, #1\n"

// Load the clamp_max bound (in parallel with the clamp_min)
"ldr r2, [%[params], #" RUY_STR(RUY_OFFSET_OUTPUT_TRANSFORM_CLAMP_MAX) "]\n"
"vdup q13.32, r2\n" // clamp_max
"vdup.32 q13, r2\n" // clamp_max

// Perform the activation function, by clamping
// Apply the clamp_min bound
"vmax q8.s32, q8.s32, q12.s32\n"
"vmax q9.s32, q9.s32, q12.s32\n"
"vmax q10.s32, q10.s32, q12.s32\n"
"vmax q11.s32, q11.s32, q12.s32\n"
"vmax.s32 q8, q8, q12\n"
"vmax.s32 q9, q9, q12\n"
"vmax.s32 q10, q10, q12\n"
"vmax.s32 q11, q11, q12\n"
// Apply the clamp_max bound
"vmin q8.s32, q8.s32, q13.s32\n"
"vmin q9.s32, q9.s32, q13.s32\n"
"vmin q10.s32, q10.s32, q13.s32\n"
"vmin q11.s32, q11.s32, q13.s32\n"
"vmin.s32 q8, q8, q13\n"
"vmin.s32 q9, q9, q13\n"
"vmin.s32 q10, q10, q13\n"
"vmin.s32 q11, q11, q13\n"

// Convert to single precision float
"vcvt.f32.s32 q8, q8\n"
Expand Down