Fix/x86arch

src/numberparse.rs

@@ -5,12 +5,16 @@ use crate::StaticNode;
 use crate::{mem, static_cast_i64, Deserializer, ErrorType, Result};
 
 #[cfg(all(target_arch = "x86", feature = "swar-number-parsing"))]
-use std::arch::x86::{
-    __m128i, _mm_cvtsi128_si32, _mm_loadu_si128, _mm_madd_epi16, _mm_maddubs_epi16,
-    _mm_packus_epi32, _mm_set1_epi8, _mm_setr_epi16, _mm_setr_epi8, _mm_sub_epi8,
-};
+use std::arch::x86 as arch;
+
 #[cfg(all(target_arch = "x86_64", feature = "swar-number-parsing"))]
-use std::arch::x86_64::{
+use std::arch::x86_64 as arch;
+
+#[cfg(all(
+    any(target_arch = "x86", target_arch = "x86_64"),
+    feature = "swar-number-parsing"
+))]
+use arch::{
     __m128i, _mm_cvtsi128_si32, _mm_loadu_si128, _mm_madd_epi16, _mm_maddubs_epi16,
     _mm_packus_epi32, _mm_set1_epi8, _mm_setr_epi16, _mm_setr_epi8, _mm_sub_epi8,
 };
@@ -150,7 +154,7 @@ fn parse_eight_digits_unrolled(chars: &[u8]) -> u32 {
                 chars
                     .get_kinda_unchecked(0..16)
                     .as_ptr()
-                    .cast::<std::arch::x86_64::__m128i>(),
+                    .cast::<arch::__m128i>(),
             ),
             ascii0,
         );

src/sse42/deser.rs

@@ -1,9 +1,10 @@
 #[cfg(target_arch = "x86")]
-use std::arch::x86::{
-    __m128i, _mm_cmpeq_epi8, _mm_loadu_si128, _mm_movemask_epi8, _mm_set1_epi8, _mm_storeu_si128,
-};
+use std::arch::x86 as arch;
+
 #[cfg(target_arch = "x86_64")]
-use std::arch::x86_64::{
+use std::arch::x86_64 as arch;
+
+use arch::{
     __m128i, _mm_cmpeq_epi8, _mm_loadu_si128, _mm_movemask_epi8, _mm_set1_epi8, _mm_storeu_si128,
 };
 
@@ -44,9 +45,8 @@ impl<'de> Deserializer<'de> {
         let mut src_i: usize = 0;
         let mut len = src_i;
         loop {
-            let v: __m128i = unsafe {
-                _mm_loadu_si128(src.as_ptr().add(src_i).cast::<std::arch::x86_64::__m128i>())
-            };
+            let v: __m128i =
+                unsafe { _mm_loadu_si128(src.as_ptr().add(src_i).cast::<arch::__m128i>()) };
 
             // store to dest unconditionally - we can overwrite the bits we don't like
             // later
@@ -98,18 +98,11 @@ impl<'de> Deserializer<'de> {
 
         // To be more conform with upstream
         loop {
-            let v: __m128i = unsafe {
-                _mm_loadu_si128(src.as_ptr().add(src_i).cast::<std::arch::x86_64::__m128i>())
-            };
+            let v: __m128i =
+                unsafe { _mm_loadu_si128(src.as_ptr().add(src_i).cast::<arch::__m128i>()) };
 
             unsafe {
-                _mm_storeu_si128(
-                    buffer
-                        .as_mut_ptr()
-                        .add(dst_i)
-                        .cast::<std::arch::x86_64::__m128i>(),
-                    v,
-                );
+                _mm_storeu_si128(buffer.as_mut_ptr().add(dst_i).cast::<arch::__m128i>(), v);
             };
 
             // store to dest unconditionally - we can overwrite the bits we don't like

src/sse42/stage1.rs

@@ -1,12 +1,19 @@
 use crate::{static_cast_i32, static_cast_u32, Stage1Parse};
 #[cfg(target_arch = "x86")]
-use std::arch::x86::{
+use std::arch::x86 as arch;
+
+#[cfg(target_arch = "x86_64")]
+use std::arch::x86_64 as arch;
+
+#[cfg(target_arch = "x86")]
+use arch::{
     __m128i, _mm_add_epi32, _mm_and_si128, _mm_cmpeq_epi8, _mm_cmpgt_epi8, _mm_loadu_si128,
     _mm_max_epu8, _mm_movemask_epi8, _mm_or_si128, _mm_set1_epi8, _mm_set_epi32, _mm_setr_epi8,
     _mm_setzero_si128, _mm_shuffle_epi8, _mm_srli_epi32, _mm_storeu_si128, _mm_testz_si128,
 };
+
 #[cfg(target_arch = "x86_64")]
-use std::arch::x86_64::{
+use arch::{
     __m128i, _mm_add_epi32, _mm_and_si128, _mm_cmpeq_epi8, _mm_loadu_si128, _mm_max_epu8,
     _mm_movemask_epi8, _mm_set1_epi8, _mm_set_epi32, _mm_setr_epi8, _mm_setzero_si128,
     _mm_shuffle_epi8, _mm_srli_epi32, _mm_storeu_si128,
@@ -43,10 +50,10 @@ impl SimdInput {
     pub(crate) fn new(ptr: &[u8]) -> Self {
         unsafe {
             Self {
-                v0: _mm_loadu_si128(ptr.as_ptr().cast::<std::arch::x86_64::__m128i>()),
-                v1: _mm_loadu_si128(ptr.as_ptr().add(16).cast::<std::arch::x86_64::__m128i>()),
-                v2: _mm_loadu_si128(ptr.as_ptr().add(32).cast::<std::arch::x86_64::__m128i>()),
-                v3: _mm_loadu_si128(ptr.as_ptr().add(48).cast::<std::arch::x86_64::__m128i>()),
+                v0: _mm_loadu_si128(ptr.as_ptr().cast::<arch::__m128i>()),
+                v1: _mm_loadu_si128(ptr.as_ptr().add(16).cast::<arch::__m128i>()),
+                v2: _mm_loadu_si128(ptr.as_ptr().add(32).cast::<arch::__m128i>()),
+                v3: _mm_loadu_si128(ptr.as_ptr().add(48).cast::<arch::__m128i>()),
             }
         }
     }
@@ -57,10 +64,7 @@ impl Stage1Parse<__m128i> for SimdInput {
     #[cfg(target_feature = "pclmulqdq")]
     #[allow(clippy::cast_sign_loss)]
     fn compute_quote_mask(quote_bits: u64) -> u64 {
-        #[cfg(target_arch = "x86")]
-        use std::arch::x86::{_mm_clmulepi64_si128, _mm_cvtsi128_si64, _mm_set_epi64x};
-        #[cfg(target_arch = "x86_64")]
-        use std::arch::x86_64::{_mm_clmulepi64_si128, _mm_cvtsi128_si64, _mm_set_epi64x};
+        use arch::{_mm_clmulepi64_si128, _mm_cvtsi128_si64, _mm_set_epi64x};
 
         unsafe {
             _mm_cvtsi128_si64(_mm_clmulepi64_si128(
@@ -275,12 +279,7 @@ impl Stage1Parse<__m128i> for SimdInput {
 
                 let v: __m128i = _mm_set_epi32(v3, v2, v1, v0);
                 let v: __m128i = _mm_add_epi32(idx_64_v, v);
-                _mm_storeu_si128(
-                    base.as_mut_ptr()
-                        .add(l)
-                        .cast::<std::arch::x86_64::__m128i>(),
-                    v,
-                );
+                _mm_storeu_si128(base.as_mut_ptr().add(l).cast::<arch::__m128i>(), v);
             }
             l += 4;
         }