e2k, ppc: Make shifts unsigned

simde-common.h

@@ -914,4 +914,14 @@ HEDLEY_DIAGNOSTIC_POP
 #  define SIMDE_BUG_IGNORE_SIGN_CONVERSION(expr) (expr)
 #endif
 
+/* Usually the shift count is signed (for example, NEON or SSE).
+ * OTOH, unsigned is good for PPC (vec_srl uses unsigned), and the only option for E2K.
+ * Further info: https://github.com/simd-everywhere/simde/pull/700
+ */
+#if defined(SIMDE_ARCH_E2K) || defined(SIMDE_ARCH_POWER)
+  #define SIMDE_CAST_VECTOR_SHIFT_COUNT(width, value) HEDLEY_STATIC_CAST(uint##width##_t, (value))
+#else
+  #define SIMDE_CAST_VECTOR_SHIFT_COUNT(width, value) HEDLEY_STATIC_CAST(int##width##_t, (value))
+#endif
+
 #endif /* !defined(SIMDE_COMMON_H) */

x86/avx2.h

@@ -4716,7 +4716,7 @@ simde_mm256_srl_epi16 (simde__m256i a, simde__m128i count) {
       uint64_t shift = HEDLEY_STATIC_CAST(uint64_t , (count_.i64[0] > 16 ? 16 : count_.i64[0]));
 
       #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-        r_.u16 = a_.u16 >> HEDLEY_STATIC_CAST(int16_t, shift);
+        r_.u16 = a_.u16 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(16, shift);
       #else
         SIMDE_VECTORIZE
         for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
@@ -4753,7 +4753,7 @@ simde_mm256_srl_epi32 (simde__m256i a, simde__m128i count) {
       uint64_t shift = HEDLEY_STATIC_CAST(uint64_t , (count_.i64[0] > 32 ? 32 : count_.i64[0]));
 
       #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-        r_.u32 = a_.u32 >> HEDLEY_STATIC_CAST(int32_t, shift);
+        r_.u32 = a_.u32 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(32, shift);
       #else
         SIMDE_VECTORIZE
         for (size_t i = 0 ; i < (sizeof(r_.i32) / sizeof(r_.i32[0])) ; i++) {
@@ -4790,7 +4790,7 @@ simde_mm256_srl_epi64 (simde__m256i a, simde__m128i count) {
       uint64_t shift = HEDLEY_STATIC_CAST(uint64_t , (count_.i64[0] > 64 ? 64 : count_.i64[0]));
 
       #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-        r_.u64 = a_.u64 >> HEDLEY_STATIC_CAST(int64_t, shift);
+        r_.u64 = a_.u64 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(64, shift);
       #else
         SIMDE_VECTORIZE
         for (size_t i = 0 ; i < (sizeof(r_.i64) / sizeof(r_.i64[0])) ; i++) {
@@ -4828,7 +4828,7 @@ simde_mm256_srli_epi16 (simde__m256i a, const int imm8)
       simde_memset(&r_, 0, sizeof(r_));
     } else {
       #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-        r_.u16 = a_.u16 >> HEDLEY_STATIC_CAST(int16_t, imm8);
+        r_.u16 = a_.u16 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(16, imm8);
       #else
         SIMDE_VECTORIZE
         for (size_t i = 0 ; i < (sizeof(r_.u16) / sizeof(r_.u16[0])) ; i++) {
@@ -4867,7 +4867,7 @@ simde_mm256_srli_epi32 (simde__m256i a, const int imm8)
       r_.altivec_u32[i] = vec_sr(a_.altivec_u32[i], sv);
     }
   #elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-    r_.u32 = a_.u32 >> HEDLEY_STATIC_CAST(int16_t, imm8);
+    r_.u32 = a_.u32 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(16, imm8);
   #else
     SIMDE_VECTORIZE
     for (size_t i = 0 ; i < (sizeof(r_.u32) / sizeof(r_.u32[0])) ; i++) {
@@ -4899,7 +4899,7 @@ simde_mm256_srli_epi64 (simde__m256i a, const int imm8)
     a_ = simde__m256i_to_private(a);
 
 #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-  r_.u64 = a_.u64 >> HEDLEY_STATIC_CAST(int32_t, imm8);
+  r_.u64 = a_.u64 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(32, imm8);
 #else
   SIMDE_VECTORIZE
   for (size_t i = 0 ; i < (sizeof(r_.u64) / sizeof(r_.u64[0])) ; i++) {

x86/avx512/srli.h

@@ -58,7 +58,7 @@ simde_mm512_srli_epi16 (simde__m512i a, const unsigned int imm8)
       return simde_mm512_setzero_si512();
 
     #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-      r_.u16 = a_.u16 >> HEDLEY_STATIC_CAST(int16_t, imm8);
+      r_.u16 = a_.u16 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(16, imm8);
     #else
       SIMDE_VECTORIZE
       for (size_t i = 0 ; i < (sizeof(r_.u16) / sizeof(r_.u16[0])) ; i++) {

x86/sse2.h

@@ -5722,7 +5722,7 @@ simde_mm_slli_epi16 (simde__m128i a, const int imm8)
     a_ = simde__m128i_to_private(a);
 
   #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-    r_.i16 = a_.i16 << (imm8 & 0xff);
+    r_.i16 = a_.i16 << SIMDE_CAST_VECTOR_SHIFT_COUNT(8, imm8 & 0xff);
   #else
     const int s = (imm8 > HEDLEY_STATIC_CAST(int, sizeof(r_.i16[0]) * CHAR_BIT) - 1) ? 0 : imm8;
     SIMDE_VECTORIZE
@@ -5879,7 +5879,7 @@ simde_mm_srli_epi16 (simde__m128i a, const int imm8)
     a_ = simde__m128i_to_private(a);
 
   #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-    r_.u16 = a_.u16 >> imm8;
+    r_.u16 = a_.u16 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(8, imm8);
   #else
     SIMDE_VECTORIZE
     for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
@@ -5928,7 +5928,7 @@ simde_mm_srli_epi32 (simde__m128i a, const int imm8)
     a_ = simde__m128i_to_private(a);
 
   #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
-    r_.u32 = a_.u32 >> (imm8 & 0xff);
+    r_.u32 = a_.u32 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(8, imm8 & 0xff);
   #else
     SIMDE_VECTORIZE
     for (size_t i = 0 ; i < (sizeof(r_.i32) / sizeof(r_.i32[0])) ; i++) {
@@ -5992,7 +5992,7 @@ simde_mm_srli_epi64 (simde__m128i a, const int imm8)
     r_.neon_u64 = vshlq_u64(a_.neon_u64, vdupq_n_s64(-imm8));
   #else
     #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && !defined(SIMDE_BUG_GCC_94488)
-      r_.u64 = a_.u64 >> imm8;
+      r_.u64 = a_.u64 >> SIMDE_CAST_VECTOR_SHIFT_COUNT(8, imm8);
     #else
       SIMDE_VECTORIZE
       for (size_t i = 0 ; i < (sizeof(r_.i64) / sizeof(r_.i64[0])) ; i++) {