External/cglm
1
0
Fork 0
mirror of https://github.com/recp/cglm.git synced 2025-12-24 20:34:58 +00:00
Code Issues Packages Projects Releases Wiki Activity

simd128: code style

Browse Source
This commit is contained in:
myfreeer
2023-04-01 19:38:28 +08:00
parent 73adfe08c0
commit 51ce4db82e
8 changed files with 36 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
include/cglm/quat.h
View File

@@ -262,7 +262,7 @@ glm_quat_normalize_to(versor q, versor dest) {
x0 = glmm_load(q);
xdot = glmm_vdot(x0, x0);
// dot = _mm_cvtss_f32(xdot);
/* dot = _mm_cvtss_f32(xdot); */
dot = wasm_f32x4_extract_lane(xdot, 0);
if (dot <= 0.0f) {

4
include/cglm/simd/wasm.h
View File

@@ -41,7 +41,7 @@ glmm_vhadds(glmm_128 v) {
glmm_128 shuf, sums;
shuf = glmm_shuff1(v, 2, 3, 0, 1);
sums = wasm_f32x4_add(v, shuf);
// shuf = _mm_movehl_ps(shuf, sums);
/* shuf = _mm_movehl_ps(shuf, sums); */
shuf = wasm_i32x4_shuffle(shuf, sums, 6, 7, 2, 3);
sums = wasm_i32x4_shuffle(sums, wasm_f32x4_add(sums, shuf), 4, 1, 2, 3);
return sums;
@@ -76,7 +76,7 @@ glmm_vhmax(glmm_128 v) {
x0 = glmm_shuff1(v, 2, 3, 2, 3); /* [2, 3, 2, 3] */
x1 = wasm_f32x4_pmax(x0, v); /* [0|2, 1|3, 2|2, 3|3] */
x2 = glmm_splat(x1, 1); /* [1|3, 1|3, 1|3, 1|3] */
// _mm_max_ss
/* _mm_max_ss */
return wasm_i32x4_shuffle(x1, wasm_f32x4_pmax(x1, x2), 4, 1, 2, 3);
}

10
include/cglm/simd/wasm/affine.h
View File

@@ -93,18 +93,18 @@ glm_inv_tr_wasm(mat4 mat) {
r3 = glmm_load(mat[3]);
x1 = wasm_f32x4_const(0.0f, 0.0f, 0.0f, 1.0f);
// _MM_TRANSPOSE4_PS(r0, r1, r2, x1);
/* _MM_TRANSPOSE4_PS(r0, r1, r2, x1); */
x2 = wasm_i32x4_shuffle(r0, r1, 0, 4, 1, 5);
x3 = wasm_i32x4_shuffle(r0, r1, 2, 6, 3, 7);
x4 = wasm_i32x4_shuffle(r2, x1, 0, 4, 1, 5);
x5 = wasm_i32x4_shuffle(r2, x1, 2, 6, 3, 7);
// r0 = _mm_movelh_ps(x2, x4);
/* r0 = _mm_movelh_ps(x2, x4); */
r0 = wasm_i32x4_shuffle(x2, x4, 0, 1, 4, 5);
// r1 = _mm_movehl_ps(x4, x2);
/* r1 = _mm_movehl_ps(x4, x2); */
r1 = wasm_i32x4_shuffle(x4, x2, 6, 7, 2, 3);
// r2 = _mm_movelh_ps(x3, x5);
/* r2 = _mm_movelh_ps(x3, x5); */
r2 = wasm_i32x4_shuffle(x3, x5, 0, 1, 4, 5);
// x1 = _mm_movehl_ps(x5, x3);
/* x1 = _mm_movehl_ps(x5, x3); */
x1 = wasm_i32x4_shuffle(x5, x3, 6, 7, 2, 3);
x2 = glmm_shuff1(r3, 0, 0, 0, 0);

4
include/cglm/simd/wasm/mat2.h
View File

@@ -22,9 +22,9 @@ glm_mat2_mul_wasm(mat2 m1, mat2 m2, mat2 dest) {
x3 = glmm_shuff1(x2, 2, 2, 0, 0);
x4 = glmm_shuff1(x2, 3, 3, 1, 1);
// x0 = _mm_movelh_ps(x1, x1);
/* x0 = _mm_movelh_ps(x1, x1); */
x0 = wasm_i32x4_shuffle(x1, x1, 0, 1, 4, 5);
// x2 = _mm_movehl_ps(x1, x1);
/* x2 = _mm_movehl_ps(x1, x1); */
x2 = wasm_i32x4_shuffle(x1, x1, 6, 7, 2, 3);
/*

18
include/cglm/simd/wasm/mat3.h
View File

@@ -42,22 +42,22 @@ glm_mat3_mul_wasm(mat3 m1, mat3 m2, mat3 dest) {
x6 = glmm_shuff1(x3, 2, 0, 0, 0); /* b11 b01 b01 b01 */
x2 = glmm_shuff1(r1, 3, 3, 0, 0); /* b21 b21 b11 b11 */
// x8 = _mm_unpackhi_ps(x8, x4);
// x9 = _mm_unpackhi_ps(x7, x2);
/* x8 = _mm_unpackhi_ps(x8, x4); */
/* x9 = _mm_unpackhi_ps(x7, x2); */
x8 = wasm_i32x4_shuffle(x8, x4, 2, 6, 3, 7); /* a10 a00 a12 a02 */
x9 = wasm_i32x4_shuffle(x7, x2, 2, 6, 3, 7); /* b21 b20 b21 b20 */
x0 = glmm_fmadd(x4, x6, x0);
x1 = glmm_fmadd(x5, x2, x1);
// x2 = _mm_movehl_ps(l2, l1);
/* x2 = _mm_movehl_ps(l2, l1); */
x2 = wasm_i32x4_shuffle(l2, l1, 6, 7, 2, 3); /* a22 a22 a21 a20 */
x3 = glmm_shuff1(x2, 0, 2, 1, 0); /* a20 a22 a21 a20 */
x2 = glmm_shuff1(x2, 1, 0, 2, 1); /* a21 a20 a22 a21 */
x4 = wasm_i32x4_shuffle(r0, r1, 2, 2, 5, 5); /* b12 b12 b02 b02 */
x5 = glmm_shuff1(x4, 3, 0, 0, 0); /* b12 b02 b02 b02 */
// x4 = _mm_movehl_ps(r2, x4);
/* x4 = _mm_movehl_ps(r2, x4); */
x4 = wasm_i32x4_shuffle(r2, x4, 6, 7, 2, 3); /* b22 b22 b12 b12 */
x0 = glmm_fmadd(x3, x5, x0);
x1 = glmm_fmadd(x2, x4, x1);
@@ -67,17 +67,17 @@ glm_mat3_mul_wasm(mat3 m1, mat3 m2, mat3 dest) {
a12 * b21 +
a22 * b22 +
0 * 00 */
// x2 = _mm_movelh_ps(x8, l2);
// x3 = _mm_movelh_ps(x9, r2);
/* x2 = _mm_movelh_ps(x8, l2); */
/* x3 = _mm_movelh_ps(x9, r2); */
x2 = wasm_i32x4_shuffle(x8, l2, 0, 1, 4, 5); /* 0.f a22 a12 a02 */
x3 = wasm_i32x4_shuffle(x9, r2, 0, 1, 4, 5); /* 0.f b22 b21 b20 */
x2 = glmm_vdots(x2, x3);
// _mm_storeu_ps(&dest[0][0], x0);
/* _mm_storeu_ps(&dest[0][0], x0); */
wasm_v128_store(&dest[0][0], x0);
// _mm_storeu_ps(&dest[1][1], x1);
/* _mm_storeu_ps(&dest[1][1], x1); */
wasm_v128_store(&dest[1][1], x1);
// _mm_store_ss (&dest[2][2], x2);
/* _mm_store_ss (&dest[2][2], x2); */
wasm_v128_store32_lane(&dest[2][2], x2, 0);
}

28
include/cglm/simd/wasm/mat4.h
View File

@@ -36,18 +36,18 @@ glm_mat4_transp_wasm(mat4 m, mat4 dest) {
r2 = glmm_load(m[2]);
r3 = glmm_load(m[3]);
// _MM_TRANSPOSE4_PS(r0, r1, r2, r3);
/* _MM_TRANSPOSE4_PS(r0, r1, r2, r3); */
tmp0 = wasm_i32x4_shuffle(r0, r1, 0, 4, 1, 5);
tmp1 = wasm_i32x4_shuffle(r0, r1, 2, 6, 3, 7);
tmp2 = wasm_i32x4_shuffle(r2, r3, 0, 4, 1, 5);
tmp3 = wasm_i32x4_shuffle(r2, r3, 2, 6, 3, 7);
// r0 = _mm_movelh_ps(tmp0, tmp2);
/* r0 = _mm_movelh_ps(tmp0, tmp2); */
r0 = wasm_i32x4_shuffle(tmp0, tmp2, 0, 1, 4, 5);
// r1 = _mm_movehl_ps(tmp2, tmp0);
/* r1 = _mm_movehl_ps(tmp2, tmp0); */
r1 = wasm_i32x4_shuffle(tmp2, tmp0, 6, 7, 2, 3);
// r2 = _mm_movelh_ps(tmp1, tmp3);
/* r2 = _mm_movelh_ps(tmp1, tmp3); */
r2 = wasm_i32x4_shuffle(tmp1, tmp3, 0, 1, 4, 5);
// r3 = _mm_movehl_ps(tmp3, tmp1);
/* r3 = _mm_movehl_ps(tmp3, tmp1); */
r3 = wasm_i32x4_shuffle(tmp3, tmp1, 6, 7, 2, 3);
glmm_store(dest[0], r0);
@@ -186,9 +186,9 @@ glm_mat4_inv_fast_wasm(mat4 mat, mat4 dest) {
r1 = glmm_load(mat[1]); /* h g f e */
r2 = glmm_load(mat[2]); /* l k j i */
r3 = glmm_load(mat[3]); /* p o n m */
// x0 = _mm_movehl_ps(r3, r2);
/* x0 = _mm_movehl_ps(r3, r2); */
x0 = wasm_i32x4_shuffle(r3, r2, 6, 7, 2, 3); /* p o l k */
// x3 = _mm_movelh_ps(r2, r3);
/* x3 = _mm_movelh_ps(r2, r3); */
x3 = wasm_i32x4_shuffle(r2, r3, 0, 1, 4, 5); /* n m j i */
x1 = glmm_shuff1(x0, 1, 3, 3 ,3); /* l p p p */
x2 = glmm_shuff1(x0, 0, 2, 2, 2); /* k o o o */
@@ -242,9 +242,9 @@ glm_mat4_inv_fast_wasm(mat4 mat, mat4 dest) {
t2[5] = e * n - m * f;
t3[5] = e * j - i * f; */
t5 = glmm_fnmadd(x7, x5, t5);
// x4 = _mm_movelh_ps(r0, r1);
/* x4 = _mm_movelh_ps(r0, r1); */
x4 = wasm_i32x4_shuffle(r0, r1, 0, 1, 4, 5); /* f e b a */
// x5 = _mm_movehl_ps(r1, r0);
/* x5 = _mm_movehl_ps(r1, r0); */
x5 = wasm_i32x4_shuffle(r1, r0, 6, 7, 2, 3); /* h g d c */
x0 = glmm_shuff1(x4, 0, 0, 0, 2); /* a a a e */
@@ -300,7 +300,7 @@ glm_mat4_inv_fast_wasm(mat4 mat, mat4 dest) {
x1 = wasm_i32x4_shuffle(v2, v3, 0, 0, 4, 4);
x0 = wasm_i32x4_shuffle(x0, x1, 0, 2, 4, 6);
// x0 = _mm_rcp_ps(glmm_vhadd(wasm_f32x4_mul(x0, r0)));
/* x0 = _mm_rcp_ps(glmm_vhadd(wasm_f32x4_mul(x0, r0))); */
x0 = wasm_f32x4_div(wasm_f32x4_const_splat(1.0f),
glmm_vhadd(wasm_f32x4_mul(x0, r0)));
@@ -326,9 +326,9 @@ glm_mat4_inv_wasm(mat4 mat, mat4 dest) {
r1 = glmm_load(mat[1]); /* h g f e */
r2 = glmm_load(mat[2]); /* l k j i */
r3 = glmm_load(mat[3]); /* p o n m */
// x0 = _mm_movehl_ps(r3, r2);
/* x0 = _mm_movehl_ps(r3, r2); */
x0 = wasm_i32x4_shuffle(r3, r2, 6, 7, 2, 3); /* p o l k */
// x3 = _mm_movelh_ps(r2, r3);
/* x3 = _mm_movelh_ps(r2, r3); */
x3 = wasm_i32x4_shuffle(r2, r3, 0, 1, 4, 5); /* n m j i */
x1 = glmm_shuff1(x0, 1, 3, 3 ,3); /* l p p p */
x2 = glmm_shuff1(x0, 0, 2, 2, 2); /* k o o o */
@@ -382,9 +382,9 @@ glm_mat4_inv_wasm(mat4 mat, mat4 dest) {
t2[5] = e * n - m * f;
t3[5] = e * j - i * f; */
t5 = glmm_fnmadd(x7, x5, t5);
// x4 = _mm_movelh_ps(r0, r1);
/* x4 = _mm_movelh_ps(r0, r1); */
x4 = wasm_i32x4_shuffle(r0, r1, 0, 1, 4, 5); /* f e b a */
// x5 = _mm_movehl_ps(r1, r0);
/* x5 = _mm_movehl_ps(r1, r0); */
x5 = wasm_i32x4_shuffle(r1, r0, 6, 7, 2, 3); /* h g d c */
x0 = glmm_shuff1(x4, 0, 0, 0, 2); /* a a a e */

2
include/cglm/simd/wasm/quat.h
View File

@@ -28,7 +28,7 @@ glm_quat_mul_wasm(versor p, versor q, versor dest) {
xq = glmm_load(q);
x1 = wasm_f32x4_const(0.f, -0.f, 0.f, -0.f); /* TODO: _mm_set1_ss() + shuff ? */
r = wasm_f32x4_mul(glmm_splat_w(xp), xq);
// x2 = _mm_unpackhi_ps(x1, x1);
/* x2 = _mm_unpackhi_ps(x1, x1); */
x2 = wasm_i32x4_shuffle(x1, x1, 2, 6, 3, 7);
x3 = glmm_shuff1(x1, 3, 2, 0, 1);
x = glmm_splat_x(xp);

4
include/cglm/vec4.h
View File

@@ -163,7 +163,7 @@ CGLM_INLINE
void
glm_vec4_ucopy(vec4 v, vec4 dest) {
#if defined(__wasm__) && defined(__wasm_simd128__)
// note here wasm v128.load/v128.store support unaligned loads and stores
/* note here wasm v128.load/v128.store support unaligned loads and stores */
wasm_v128_store(dest, wasm_v128_load(v));
#else
dest[0] = v[0];
@@ -761,7 +761,7 @@ glm_vec4_normalize_to(vec4 v, vec4 dest) {
x0 = glmm_load(v);
xdot = glmm_vdot(x0, x0);
// dot = _mm_cvtss_f32(xdot);
/* dot = _mm_cvtss_f32(xdot); */
dot = wasm_f32x4_extract_lane(xdot, 0);
if (dot == 0.0f) {