mirror of
https://github.com/ddidderr/wuffs.git
synced 2026-11-21 00:47:28 +01:00
Rename some std etc_sse42 functions
This commit is contained in:
1 parent
ce97a7ccd9
commit
8c4348b8f6
5 files changed
+38
-38
No files matched your search
@@ -27,7 +27,7 @@ pub func hasher.update_u32!(x: slice base.u8) base.u32 {
|
||||
this.state = 1
|
||||
choose up = [
|
||||
up_arm_neon,
|
||||
up_sse42]
|
||||
up_x86_sse42]
|
||||
}
|
||||
this.up!(x: args.x)
|
||||
return this.state
|
||||
|
||||
@@ -12,7 +12,7 @@
|
||||
// See the License for the specific language governing permissions and
|
||||
// limitations under the License.
|
||||
|
||||
pri func hasher.up_sse42!(x: slice base.u8),
|
||||
pri func hasher.up_x86_sse42!(x: slice base.u8),
|
||||
choose cpu_arch >= x86_sse42,
|
||||
{
|
||||
// These variables are the same as the non-SIMD version.
|
||||
|
||||
@@ -19,7 +19,7 @@
|
||||
// This (filter = 1, distance = 3) implementation doesn't actually bench faster
|
||||
// than the non-SIMD one.
|
||||
//
|
||||
// pri func decoder.filter_1_distance_3_sse42!(curr: slice base.u8),
|
||||
// pri func decoder.filter_1_distance_3_x86_sse42!(curr: slice base.u8),
|
||||
// choose cpu_arch >= x86_sse42,
|
||||
// {
|
||||
// var c : slice base.u8
|
||||
@@ -42,7 +42,7 @@
|
||||
// https://github.com/google/wuffs/commit/1660f9268621ed4415b3b363f0a0e1026d4aa83d
|
||||
// "Have std/png filter_1_distance_? use more SIMD" for a pessimizing example.
|
||||
|
||||
pri func decoder.filter_1_distance_4_sse42!(curr: slice base.u8),
|
||||
pri func decoder.filter_1_distance_4_x86_sse42!(curr: slice base.u8),
|
||||
choose cpu_arch >= x86_sse42,
|
||||
{
|
||||
var c : slice base.u8
|
||||
@@ -61,16 +61,16 @@ pri func decoder.filter_1_distance_4_sse42!(curr: slice base.u8),
|
||||
|
||||
// Filter 3: Average.
|
||||
|
||||
// Similar to filter_1_distance_3_sse42, the SIMD implementation for (filter =
|
||||
// 3, distance = 3) doesn't actually bench faster than the non-SIMD one.
|
||||
// Similar to filter_1_distance_3_x86_sse42, the SIMD implementation for
|
||||
// (filter = 3, distance = 3) doesn't actually bench faster than non-SIMD.
|
||||
//
|
||||
// pri func decoder.filter_3_distance_3_sse42!(curr: slice base.u8, prev: slice base.u8),
|
||||
// pri func decoder.filter_3_distance_3_x86_sse42!(curr: slice base.u8, prev: slice base.u8),
|
||||
// choose cpu_arch >= x86_sse42,
|
||||
// {
|
||||
// etc
|
||||
// }
|
||||
|
||||
pri func decoder.filter_3_distance_4_sse42!(curr: slice base.u8, prev: slice base.u8),
|
||||
pri func decoder.filter_3_distance_4_x86_sse42!(curr: slice base.u8, prev: slice base.u8),
|
||||
choose cpu_arch >= x86_sse42,
|
||||
{
|
||||
var c : slice base.u8
|
||||
@@ -127,13 +127,13 @@ pri func decoder.filter_3_distance_4_sse42!(curr: slice base.u8, prev: slice bas
|
||||
|
||||
// Filter 4: Paeth.
|
||||
|
||||
pri func decoder.filter_4_distance_3_sse42!(curr: slice base.u8, prev: slice base.u8),
|
||||
pri func decoder.filter_4_distance_3_x86_sse42!(curr: slice base.u8, prev: slice base.u8),
|
||||
choose cpu_arch >= x86_sse42,
|
||||
{
|
||||
// See the comments in filter_4_distance_4_sse42 for an explanation of how
|
||||
// this works. That function's single loop is copied twice here, once with
|
||||
// "length: 4" and once with "length: 3". It's generally faster to load 4
|
||||
// bytes at a time instead of 3.
|
||||
// See the comments in filter_4_distance_4_x86_sse42 for an explanation of
|
||||
// how this works. That function's single loop is copied twice here, once
|
||||
// with "length: 4" and once with "length: 3". It's generally faster to
|
||||
// load 4 bytes at a time instead of 3.
|
||||
//
|
||||
// Differences between that function and this one are marked with a §.
|
||||
|
||||
@@ -205,7 +205,7 @@ pri func decoder.filter_4_distance_3_sse42!(curr: slice base.u8, prev: slice bas
|
||||
}
|
||||
}
|
||||
|
||||
pri func decoder.filter_4_distance_4_sse42!(curr: slice base.u8, prev: slice base.u8),
|
||||
pri func decoder.filter_4_distance_4_x86_sse42!(curr: slice base.u8, prev: slice base.u8),
|
||||
choose cpu_arch >= x86_sse42,
|
||||
{
|
||||
var c : slice base.u8
|
||||
|
||||
@@ -377,20 +377,20 @@ pri func decoder.choose_filter_implementations!() {
|
||||
choose filter_3 = [filter_3_distance_3_fallback]
|
||||
choose filter_4 = [
|
||||
filter_4_distance_3_arm_neon,
|
||||
filter_4_distance_3_sse42,
|
||||
filter_4_distance_3_x86_sse42,
|
||||
filter_4_distance_3_fallback]
|
||||
} else if this.filter_distance == 4 {
|
||||
choose filter_1 = [
|
||||
filter_1_distance_4_arm_neon,
|
||||
filter_1_distance_4_sse42,
|
||||
filter_1_distance_4_x86_sse42,
|
||||
filter_1_distance_4_fallback]
|
||||
choose filter_3 = [
|
||||
filter_3_distance_4_arm_neon,
|
||||
filter_3_distance_4_sse42,
|
||||
filter_3_distance_4_x86_sse42,
|
||||
filter_3_distance_4_fallback]
|
||||
choose filter_4 = [
|
||||
filter_4_distance_4_arm_neon,
|
||||
filter_4_distance_4_sse42,
|
||||
filter_4_distance_4_x86_sse42,
|
||||
filter_4_distance_4_fallback]
|
||||
}
|
||||
}
|
||||
|
||||
Reference in new issue
Block a user