Skip to content

Commit

Permalink
Separate SIMD variant of f32 types into unique types.
Browse files Browse the repository at this point in the history 
Adds Mat2A, Mat4A, QuatA, Vec4A.
  • Loading branch information
@bitshifter
bitshifter committed Feb 24, 2024
1 parent 1ea8163 commit cba3284
Show file tree
Hide file tree
Showing 97 changed files with 15,836 additions and 9,431 deletions.
125 changes: 86 additions & 39 deletions codegen/src/outputs.rs
View file
Expand Up @@ -25,6 +25,7 @@ impl ContextBuilder {
.with_key_val("vec4_t", &format!("{prefix}Vec4"))
.with_self_t(&format!("{prefix}Vec{dim}"))
.with_dimension(dim)
.with_is_align(false)
}

pub fn new_vec2_swizzle_impl() -> Self {
Expand All @@ -36,15 +37,25 @@ impl ContextBuilder {
}

pub fn new_vec3a_swizzle_impl() -> Self {
Self::new_tvecn_swizzle_impl(3, "")
Self::new_vec3_swizzle_impl()
.with_key_val("vec3_t", "Vec3A")
.with_key_val("vec4_t", "Vec4A")
.with_self_t("Vec3A")
.with_is_align(true)
}

pub fn new_vec4_swizzle_impl() -> Self {
Self::new_tvecn_swizzle_impl(4, "")
}

pub fn new_vec4a_swizzle_impl() -> Self {
Self::new_vec4_swizzle_impl()
.with_key_val("vec3_t", "Vec3A")
.with_key_val("vec4_t", "Vec4A")
.with_self_t("Vec4A")
.with_is_align(true)
}

pub fn new_dvec2_swizzle_impl() -> Self {
Self::new_tvecn_swizzle_impl(2, "D")
}
Expand Down Expand Up @@ -205,11 +216,15 @@ impl ContextBuilder {
}

pub fn new_vec3a() -> Self {
Self::new_vecn(3).with_scalar_t("f32").with_is_align(true)
Self::new_vec3().with_is_align(true)
}

pub fn new_vec4() -> Self {
Self::new_vecn(4).with_scalar_t("f32").with_is_align(true)
Self::new_vecn(4).with_scalar_t("f32")
}

pub fn new_vec4a() -> Self {
Self::new_vec4().with_is_align(true)
}

pub fn new_dvec2() -> Self {
Expand Down Expand Up @@ -303,6 +318,10 @@ impl ContextBuilder {
.with_scalar_t("f32")
}

pub fn new_quata() -> Self {
Self::new_quat().with_is_align(true)
}

pub fn new_dquat() -> Self {
Self::new_quat().with_scalar_t("f64")
}
Expand All @@ -319,6 +338,10 @@ impl ContextBuilder {
Self::new_tmatn(2, "f32")
}

pub fn new_mat2a() -> Self {
Self::new_mat2().with_is_align(true)
}

pub fn new_dmat2() -> Self {
Self::new_tmatn(2, "f64")
}
Expand All @@ -328,7 +351,7 @@ impl ContextBuilder {
}

pub fn new_mat3a() -> Self {
Self::new_tmatn(3, "f32").with_is_align(true)
Self::new_mat3().with_is_align(true)
}

pub fn new_dmat3() -> Self {
Expand All @@ -339,6 +362,10 @@ impl ContextBuilder {
Self::new_tmatn(4, "f32")
}

pub fn new_mat4a() -> Self {
Self::new_mat4().with_is_align(true)
}

pub fn new_dmat4() -> Self {
Self::new_tmatn(4, "f64")
}
Expand Down Expand Up @@ -441,24 +468,28 @@ pub fn build_output_pairs() -> HashMap<&'static str, tera::Context> {
.build(),
),
(
"src/swizzles/scalar/vec4_impl.rs",
"src/swizzles/vec4_impl.rs",
ContextBuilder::new_vec4_swizzle_impl().build(),
),
(
"src/swizzles/sse2/vec4_impl.rs",
ContextBuilder::new_vec4_swizzle_impl()
"src/swizzles/scalar/vec4a_impl.rs",
ContextBuilder::new_vec4a_swizzle_impl().build(),
),
(
"src/swizzles/sse2/vec4a_impl.rs",
ContextBuilder::new_vec4a_swizzle_impl()
.target_sse2()
.build(),
),
(
"src/swizzles/wasm32/vec4_impl.rs",
ContextBuilder::new_vec4_swizzle_impl()
"src/swizzles/wasm32/vec4a_impl.rs",
ContextBuilder::new_vec4a_swizzle_impl()
.target_wasm32()
.build(),
),
(
"src/swizzles/coresimd/vec4_impl.rs",
ContextBuilder::new_vec4_swizzle_impl()
"src/swizzles/coresimd/vec4a_impl.rs",
ContextBuilder::new_vec4a_swizzle_impl()
.target_coresimd()
.build(),
),
Expand Down Expand Up @@ -612,18 +643,22 @@ pub fn build_output_pairs() -> HashMap<&'static str, tera::Context> {
"src/f32/coresimd/vec3a.rs",
ContextBuilder::new_vec3a().target_coresimd().build(),
),
("src/f32/scalar/vec4.rs", ContextBuilder::new_vec4().build()),
("src/f32/vec4.rs", ContextBuilder::new_vec4().build()),
(
"src/f32/sse2/vec4.rs",
ContextBuilder::new_vec4().target_sse2().build(),
"src/f32/scalar/vec4a.rs",
ContextBuilder::new_vec4a().build(),
),
(
"src/f32/wasm32/vec4.rs",
ContextBuilder::new_vec4().target_wasm32().build(),
"src/f32/sse2/vec4a.rs",
ContextBuilder::new_vec4a().target_sse2().build(),
),
(
"src/f32/coresimd/vec4.rs",
ContextBuilder::new_vec4().target_coresimd().build(),
"src/f32/wasm32/vec4a.rs",
ContextBuilder::new_vec4a().target_wasm32().build(),
),
(
"src/f32/coresimd/vec4a.rs",
ContextBuilder::new_vec4a().target_coresimd().build(),
),
("src/f64/dvec2.rs", ContextBuilder::new_dvec2().build()),
("src/f64/dvec3.rs", ContextBuilder::new_dvec3().build()),
Expand All @@ -646,32 +681,40 @@ pub fn build_output_pairs() -> HashMap<&'static str, tera::Context> {
("src/u64/u64vec2.rs", ContextBuilder::new_u64vec2().build()),
("src/u64/u64vec3.rs", ContextBuilder::new_u64vec3().build()),
("src/u64/u64vec4.rs", ContextBuilder::new_u64vec4().build()),
("src/f32/scalar/quat.rs", ContextBuilder::new_quat().build()),
("src/f32/quat.rs", ContextBuilder::new_quat().build()),
(
"src/f32/sse2/quat.rs",
ContextBuilder::new_quat().target_sse2().build(),
"src/f32/scalar/quata.rs",
ContextBuilder::new_quata().build(),
),
(
"src/f32/wasm32/quat.rs",
ContextBuilder::new_quat().target_wasm32().build(),
"src/f32/sse2/quata.rs",
ContextBuilder::new_quata().target_sse2().build(),
),
(
"src/f32/coresimd/quat.rs",
ContextBuilder::new_quat().target_coresimd().build(),
"src/f32/wasm32/quata.rs",
ContextBuilder::new_quata().target_wasm32().build(),
),
(
"src/f32/coresimd/quata.rs",
ContextBuilder::new_quata().target_coresimd().build(),
),
("src/f64/dquat.rs", ContextBuilder::new_dquat().build()),
("src/f32/scalar/mat2.rs", ContextBuilder::new_mat2().build()),
("src/f32/mat2.rs", ContextBuilder::new_mat2().build()),
(
"src/f32/scalar/mat2a.rs",
ContextBuilder::new_mat2a().build(),
),
(
"src/f32/sse2/mat2.rs",
ContextBuilder::new_mat2().target_sse2().build(),
"src/f32/sse2/mat2a.rs",
ContextBuilder::new_mat2a().target_sse2().build(),
),
(
"src/f32/wasm32/mat2.rs",
ContextBuilder::new_mat2().target_wasm32().build(),
"src/f32/wasm32/mat2a.rs",
ContextBuilder::new_mat2a().target_wasm32().build(),
),
(
"src/f32/coresimd/mat2.rs",
ContextBuilder::new_mat2().target_coresimd().build(),
"src/f32/coresimd/mat2a.rs",
ContextBuilder::new_mat2a().target_coresimd().build(),
),
("src/f64/dmat2.rs", ContextBuilder::new_dmat2().build()),
("src/f32/mat3.rs", ContextBuilder::new_mat3().build()),
Expand All @@ -691,18 +734,22 @@ pub fn build_output_pairs() -> HashMap<&'static str, tera::Context> {
"src/f32/coresimd/mat3a.rs",
ContextBuilder::new_mat3a().target_coresimd().build(),
),
("src/f32/scalar/mat4.rs", ContextBuilder::new_mat4().build()),
("src/f32/mat4.rs", ContextBuilder::new_mat4().build()),
(
"src/f32/scalar/mat4a.rs",
ContextBuilder::new_mat4a().build(),
),
(
"src/f32/sse2/mat4.rs",
ContextBuilder::new_mat4().target_sse2().build(),
"src/f32/sse2/mat4a.rs",
ContextBuilder::new_mat4a().target_sse2().build(),
),
(
"src/f32/wasm32/mat4.rs",
ContextBuilder::new_mat4().target_wasm32().build(),
"src/f32/wasm32/mat4a.rs",
ContextBuilder::new_mat4a().target_wasm32().build(),
),
(
"src/f32/coresimd/mat4.rs",
ContextBuilder::new_mat4().target_coresimd().build(),
"src/f32/coresimd/mat4a.rs",
ContextBuilder::new_mat4a().target_coresimd().build(),
),
("src/f64/dmat3.rs", ContextBuilder::new_dmat3().build()),
("src/f64/dmat4.rs", ContextBuilder::new_dmat4().build()),
Expand Down
8 changes: 4 additions & 4 deletions codegen/templates/coresimd.rs.tera
View file
Expand Up @@ -243,9 +243,9 @@
let rcp0 = f32x4::splat(dot0.recip());

Self {
x_axis: Vec4(inv0 * rcp0),
y_axis: Vec4(inv1 * rcp0),
z_axis: Vec4(inv2 * rcp0),
w_axis: Vec4(inv3 * rcp0),
x_axis: Vec4A(inv0 * rcp0),
y_axis: Vec4A(inv1 * rcp0),
z_axis: Vec4A(inv2 * rcp0),
w_axis: Vec4A(inv3 * rcp0),
}
{% endmacro impl_mat4_inverse %}

0 comments on commit cba3284

Please sign in to comment.