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

Leverage experimental Sign Select WAsm SIMD instructions #1236

Closed
wants to merge 1 commit into from
Closed

Leverage experimental Sign Select WAsm SIMD instructions #1236

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
10 changes: 5 additions & 5 deletions src/f32-sigmoid/gen/wasmsimd-lut64-p2-div-x12.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,9 +118,9 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x12(
vf4567 = wasm_v128_andnot(vf4567, wasm_f32x4_gt(vz4567, vdenorm_cutoff));
vf89AB = wasm_v128_andnot(vf89AB, wasm_f32x4_gt(vz89AB, vdenorm_cutoff));

vf0123 = wasm_v128_bitselect(vf0123, wasm_f32x4_sub(vone, vf0123), wasm_i32x4_shr(vx0123, 31));
vf4567 = wasm_v128_bitselect(vf4567, wasm_f32x4_sub(vone, vf4567), wasm_i32x4_shr(vx4567, 31));
vf89AB = wasm_v128_bitselect(vf89AB, wasm_f32x4_sub(vone, vf89AB), wasm_i32x4_shr(vx89AB, 31));
vf0123 = __builtin_wasm_signselect_i32x4(vf0123, wasm_f32x4_sub(vone, vf0123), vx0123);
vf4567 = __builtin_wasm_signselect_i32x4(vf4567, wasm_f32x4_sub(vone, vf4567), vx4567);
vf89AB = __builtin_wasm_signselect_i32x4(vf89AB, wasm_f32x4_sub(vone, vf89AB), vx89AB);

wasm_v128_store(y, vf0123);
wasm_v128_store(y + 4, vf4567);
Expand Down Expand Up @@ -159,7 +159,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x12(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -195,7 +195,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x12(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
12 changes: 6 additions & 6 deletions src/f32-sigmoid/gen/wasmsimd-lut64-p2-div-x16.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,10 +140,10 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x16(
vf89AB = wasm_v128_andnot(vf89AB, wasm_f32x4_gt(vz89AB, vdenorm_cutoff));
vfCDEF = wasm_v128_andnot(vfCDEF, wasm_f32x4_gt(vzCDEF, vdenorm_cutoff));

vf0123 = wasm_v128_bitselect(vf0123, wasm_f32x4_sub(vone, vf0123), wasm_i32x4_shr(vx0123, 31));
vf4567 = wasm_v128_bitselect(vf4567, wasm_f32x4_sub(vone, vf4567), wasm_i32x4_shr(vx4567, 31));
vf89AB = wasm_v128_bitselect(vf89AB, wasm_f32x4_sub(vone, vf89AB), wasm_i32x4_shr(vx89AB, 31));
vfCDEF = wasm_v128_bitselect(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), wasm_i32x4_shr(vxCDEF, 31));
vf0123 = __builtin_wasm_signselect_i32x4(vf0123, wasm_f32x4_sub(vone, vf0123), vx0123);
vf4567 = __builtin_wasm_signselect_i32x4(vf4567, wasm_f32x4_sub(vone, vf4567), vx4567);
vf89AB = __builtin_wasm_signselect_i32x4(vf89AB, wasm_f32x4_sub(vone, vf89AB), vx89AB);
vfCDEF = __builtin_wasm_signselect_i32x4(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), vxCDEF);

wasm_v128_store(y, vf0123);
wasm_v128_store(y + 4, vf4567);
Expand Down Expand Up @@ -183,7 +183,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x16(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -219,7 +219,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x16(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
14 changes: 7 additions & 7 deletions src/f32-sigmoid/gen/wasmsimd-lut64-p2-div-x20.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -162,11 +162,11 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x20(
vfCDEF = wasm_v128_andnot(vfCDEF, wasm_f32x4_gt(vzCDEF, vdenorm_cutoff));
vfGHIJ = wasm_v128_andnot(vfGHIJ, wasm_f32x4_gt(vzGHIJ, vdenorm_cutoff));

vf0123 = wasm_v128_bitselect(vf0123, wasm_f32x4_sub(vone, vf0123), wasm_i32x4_shr(vx0123, 31));
vf4567 = wasm_v128_bitselect(vf4567, wasm_f32x4_sub(vone, vf4567), wasm_i32x4_shr(vx4567, 31));
vf89AB = wasm_v128_bitselect(vf89AB, wasm_f32x4_sub(vone, vf89AB), wasm_i32x4_shr(vx89AB, 31));
vfCDEF = wasm_v128_bitselect(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), wasm_i32x4_shr(vxCDEF, 31));
vfGHIJ = wasm_v128_bitselect(vfGHIJ, wasm_f32x4_sub(vone, vfGHIJ), wasm_i32x4_shr(vxGHIJ, 31));
vf0123 = __builtin_wasm_signselect_i32x4(vf0123, wasm_f32x4_sub(vone, vf0123), vx0123);
vf4567 = __builtin_wasm_signselect_i32x4(vf4567, wasm_f32x4_sub(vone, vf4567), vx4567);
vf89AB = __builtin_wasm_signselect_i32x4(vf89AB, wasm_f32x4_sub(vone, vf89AB), vx89AB);
vfCDEF = __builtin_wasm_signselect_i32x4(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), vxCDEF);
vfGHIJ = __builtin_wasm_signselect_i32x4(vfGHIJ, wasm_f32x4_sub(vone, vfGHIJ), vxGHIJ);

wasm_v128_store(y, vf0123);
wasm_v128_store(y + 4, vf4567);
Expand Down Expand Up @@ -207,7 +207,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x20(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -243,7 +243,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x20(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
16 changes: 8 additions & 8 deletions src/f32-sigmoid/gen/wasmsimd-lut64-p2-div-x24.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -184,12 +184,12 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x24(
vfGHIJ = wasm_v128_andnot(vfGHIJ, wasm_f32x4_gt(vzGHIJ, vdenorm_cutoff));
vfKLMN = wasm_v128_andnot(vfKLMN, wasm_f32x4_gt(vzKLMN, vdenorm_cutoff));

vf0123 = wasm_v128_bitselect(vf0123, wasm_f32x4_sub(vone, vf0123), wasm_i32x4_shr(vx0123, 31));
vf4567 = wasm_v128_bitselect(vf4567, wasm_f32x4_sub(vone, vf4567), wasm_i32x4_shr(vx4567, 31));
vf89AB = wasm_v128_bitselect(vf89AB, wasm_f32x4_sub(vone, vf89AB), wasm_i32x4_shr(vx89AB, 31));
vfCDEF = wasm_v128_bitselect(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), wasm_i32x4_shr(vxCDEF, 31));
vfGHIJ = wasm_v128_bitselect(vfGHIJ, wasm_f32x4_sub(vone, vfGHIJ), wasm_i32x4_shr(vxGHIJ, 31));
vfKLMN = wasm_v128_bitselect(vfKLMN, wasm_f32x4_sub(vone, vfKLMN), wasm_i32x4_shr(vxKLMN, 31));
vf0123 = __builtin_wasm_signselect_i32x4(vf0123, wasm_f32x4_sub(vone, vf0123), vx0123);
vf4567 = __builtin_wasm_signselect_i32x4(vf4567, wasm_f32x4_sub(vone, vf4567), vx4567);
vf89AB = __builtin_wasm_signselect_i32x4(vf89AB, wasm_f32x4_sub(vone, vf89AB), vx89AB);
vfCDEF = __builtin_wasm_signselect_i32x4(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), vxCDEF);
vfGHIJ = __builtin_wasm_signselect_i32x4(vfGHIJ, wasm_f32x4_sub(vone, vfGHIJ), vxGHIJ);
vfKLMN = __builtin_wasm_signselect_i32x4(vfKLMN, wasm_f32x4_sub(vone, vfKLMN), vxKLMN);

wasm_v128_store(y, vf0123);
wasm_v128_store(y + 4, vf4567);
Expand Down Expand Up @@ -231,7 +231,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x24(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -267,7 +267,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x24(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
4 changes: 2 additions & 2 deletions src/f32-sigmoid/gen/wasmsimd-lut64-p2-div-x4.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,7 +66,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x4(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -102,7 +102,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x4(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
8 changes: 4 additions & 4 deletions src/f32-sigmoid/gen/wasmsimd-lut64-p2-div-x8.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,8 +96,8 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x8(
vf0123 = wasm_v128_andnot(vf0123, wasm_f32x4_gt(vz0123, vdenorm_cutoff));
vf4567 = wasm_v128_andnot(vf4567, wasm_f32x4_gt(vz4567, vdenorm_cutoff));

vf0123 = wasm_v128_bitselect(vf0123, wasm_f32x4_sub(vone, vf0123), wasm_i32x4_shr(vx0123, 31));
vf4567 = wasm_v128_bitselect(vf4567, wasm_f32x4_sub(vone, vf4567), wasm_i32x4_shr(vx4567, 31));
vf0123 = __builtin_wasm_signselect_i32x4(vf0123, wasm_f32x4_sub(vone, vf0123), vx0123);
vf4567 = __builtin_wasm_signselect_i32x4(vf4567, wasm_f32x4_sub(vone, vf4567), vx4567);

wasm_v128_store(y, vf0123);
wasm_v128_store(y + 4, vf4567);
Expand Down Expand Up @@ -135,7 +135,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x8(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -171,7 +171,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x8(

v128_t vf = wasm_f32x4_div(vy, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
10 changes: 5 additions & 5 deletions src/f32-sigmoid/gen/wasmsimd-p5-div-x12.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -101,9 +101,9 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x12(
vf4567 = wasm_v128_andnot(vf4567, wasm_f32x4_gt(vz4567, vdenorm_cutoff));
vf89AB = wasm_v128_andnot(vf89AB, wasm_f32x4_gt(vz89AB, vdenorm_cutoff));

vf0123 = wasm_v128_bitselect(vf0123, wasm_f32x4_sub(vone, vf0123), wasm_i32x4_shr(vx0123, 31));
vf4567 = wasm_v128_bitselect(vf4567, wasm_f32x4_sub(vone, vf4567), wasm_i32x4_shr(vx4567, 31));
vf89AB = wasm_v128_bitselect(vf89AB, wasm_f32x4_sub(vone, vf89AB), wasm_i32x4_shr(vx89AB, 31));
vf0123 = __builtin_wasm_signselect_i32x4(vf0123, wasm_f32x4_sub(vone, vf0123), vx0123);
vf4567 = __builtin_wasm_signselect_i32x4(vf4567, wasm_f32x4_sub(vone, vf4567), vx4567);
vf89AB = __builtin_wasm_signselect_i32x4(vf89AB, wasm_f32x4_sub(vone, vf89AB), vx89AB);

wasm_v128_store(y, vf0123);
wasm_v128_store(y + 4, vf4567);
Expand Down Expand Up @@ -134,7 +134,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x12(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -162,7 +162,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x12(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
12 changes: 6 additions & 6 deletions src/f32-sigmoid/gen/wasmsimd-p5-div-x16.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -117,10 +117,10 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x16(
vf89AB = wasm_v128_andnot(vf89AB, wasm_f32x4_gt(vz89AB, vdenorm_cutoff));
vfCDEF = wasm_v128_andnot(vfCDEF, wasm_f32x4_gt(vzCDEF, vdenorm_cutoff));

vf0123 = wasm_v128_bitselect(vf0123, wasm_f32x4_sub(vone, vf0123), wasm_i32x4_shr(vx0123, 31));
vf4567 = wasm_v128_bitselect(vf4567, wasm_f32x4_sub(vone, vf4567), wasm_i32x4_shr(vx4567, 31));
vf89AB = wasm_v128_bitselect(vf89AB, wasm_f32x4_sub(vone, vf89AB), wasm_i32x4_shr(vx89AB, 31));
vfCDEF = wasm_v128_bitselect(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), wasm_i32x4_shr(vxCDEF, 31));
vf0123 = __builtin_wasm_signselect_i32x4(vf0123, wasm_f32x4_sub(vone, vf0123), vx0123);
vf4567 = __builtin_wasm_signselect_i32x4(vf4567, wasm_f32x4_sub(vone, vf4567), vx4567);
vf89AB = __builtin_wasm_signselect_i32x4(vf89AB, wasm_f32x4_sub(vone, vf89AB), vx89AB);
vfCDEF = __builtin_wasm_signselect_i32x4(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), vxCDEF);

wasm_v128_store(y, vf0123);
wasm_v128_store(y + 4, vf4567);
Expand Down Expand Up @@ -152,7 +152,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x16(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -180,7 +180,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x16(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
14 changes: 7 additions & 7 deletions src/f32-sigmoid/gen/wasmsimd-p5-div-x20.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -133,11 +133,11 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x20(
vfCDEF = wasm_v128_andnot(vfCDEF, wasm_f32x4_gt(vzCDEF, vdenorm_cutoff));
vfGHIJ = wasm_v128_andnot(vfGHIJ, wasm_f32x4_gt(vzGHIJ, vdenorm_cutoff));

vf0123 = wasm_v128_bitselect(vf0123, wasm_f32x4_sub(vone, vf0123), wasm_i32x4_shr(vx0123, 31));
vf4567 = wasm_v128_bitselect(vf4567, wasm_f32x4_sub(vone, vf4567), wasm_i32x4_shr(vx4567, 31));
vf89AB = wasm_v128_bitselect(vf89AB, wasm_f32x4_sub(vone, vf89AB), wasm_i32x4_shr(vx89AB, 31));
vfCDEF = wasm_v128_bitselect(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), wasm_i32x4_shr(vxCDEF, 31));
vfGHIJ = wasm_v128_bitselect(vfGHIJ, wasm_f32x4_sub(vone, vfGHIJ), wasm_i32x4_shr(vxGHIJ, 31));
vf0123 = __builtin_wasm_signselect_i32x4(vf0123, wasm_f32x4_sub(vone, vf0123), vx0123);
vf4567 = __builtin_wasm_signselect_i32x4(vf4567, wasm_f32x4_sub(vone, vf4567), vx4567);
vf89AB = __builtin_wasm_signselect_i32x4(vf89AB, wasm_f32x4_sub(vone, vf89AB), vx89AB);
vfCDEF = __builtin_wasm_signselect_i32x4(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), vxCDEF);
vfGHIJ = __builtin_wasm_signselect_i32x4(vfGHIJ, wasm_f32x4_sub(vone, vfGHIJ), vxGHIJ);

wasm_v128_store(y, vf0123);
wasm_v128_store(y + 4, vf4567);
Expand Down Expand Up @@ -170,7 +170,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x20(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -198,7 +198,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x20(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
16 changes: 8 additions & 8 deletions src/f32-sigmoid/gen/wasmsimd-p5-div-x24.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -149,12 +149,12 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x24(
vfGHIJ = wasm_v128_andnot(vfGHIJ, wasm_f32x4_gt(vzGHIJ, vdenorm_cutoff));
vfKLMN = wasm_v128_andnot(vfKLMN, wasm_f32x4_gt(vzKLMN, vdenorm_cutoff));

vf0123 = wasm_v128_bitselect(vf0123, wasm_f32x4_sub(vone, vf0123), wasm_i32x4_shr(vx0123, 31));
vf4567 = wasm_v128_bitselect(vf4567, wasm_f32x4_sub(vone, vf4567), wasm_i32x4_shr(vx4567, 31));
vf89AB = wasm_v128_bitselect(vf89AB, wasm_f32x4_sub(vone, vf89AB), wasm_i32x4_shr(vx89AB, 31));
vfCDEF = wasm_v128_bitselect(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), wasm_i32x4_shr(vxCDEF, 31));
vfGHIJ = wasm_v128_bitselect(vfGHIJ, wasm_f32x4_sub(vone, vfGHIJ), wasm_i32x4_shr(vxGHIJ, 31));
vfKLMN = wasm_v128_bitselect(vfKLMN, wasm_f32x4_sub(vone, vfKLMN), wasm_i32x4_shr(vxKLMN, 31));
vf0123 = __builtin_wasm_signselect_i32x4(vf0123, wasm_f32x4_sub(vone, vf0123), vx0123);
vf4567 = __builtin_wasm_signselect_i32x4(vf4567, wasm_f32x4_sub(vone, vf4567), vx4567);
vf89AB = __builtin_wasm_signselect_i32x4(vf89AB, wasm_f32x4_sub(vone, vf89AB), vx89AB);
vfCDEF = __builtin_wasm_signselect_i32x4(vfCDEF, wasm_f32x4_sub(vone, vfCDEF), vxCDEF);
vfGHIJ = __builtin_wasm_signselect_i32x4(vfGHIJ, wasm_f32x4_sub(vone, vfGHIJ), vxGHIJ);
vfKLMN = __builtin_wasm_signselect_i32x4(vfKLMN, wasm_f32x4_sub(vone, vfKLMN), vxKLMN);

wasm_v128_store(y, vf0123);
wasm_v128_store(y + 4, vf4567);
Expand Down Expand Up @@ -188,7 +188,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x24(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -216,7 +216,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x24(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
4 changes: 2 additions & 2 deletions src/f32-sigmoid/gen/wasmsimd-p5-div-x4.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,7 +59,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x4(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

wasm_v128_store(y, vf);
y += 4;
Expand Down Expand Up @@ -87,7 +87,7 @@ void xnn_f32_sigmoid_ukernel__wasmsimd_p5_div_x4(

v128_t vf = wasm_f32x4_div(ve, vd);
vf = wasm_v128_andnot(vf, wasm_f32x4_gt(vz, vdenorm_cutoff));
vf = wasm_v128_bitselect(vf, wasm_f32x4_sub(vone, vf), wasm_i32x4_shr(vx, 31));
vf = __builtin_wasm_signselect_i32x4(vf, wasm_f32x4_sub(vone, vf), vx);

if (n & (2 * sizeof(float))) {
*((double*) y) = wasm_f64x2_extract_lane(vf, 0);
Expand Down
Loading