fix(avx): workaround for missing _mm256_setr_m128i in GCC < 9

Author: xingchensong

ggml.c

@@ -188,6 +188,11 @@ typedef double ggml_float;
 #else
 #if !defined(__riscv)
 #include <immintrin.h>
+#if (defined(__GNUC__) && __GNUC__ >= 9) || defined(__INTEL_COMPILER)
+#define MM256_SET_M128I(a, b) _mm256_set_m128i((a), (b))
+#else
+#define MM256_SET_M128I(a, b) _mm256_insertf128_si256(_mm256_castsi128_si256(b), (a), 1)
+#endif
 #endif
 #endif
 #endif
@@ -533,7 +538,7 @@ static inline __m256i bytes_from_bits_32(const uint8_t * x) {
 static inline __m256i bytes_from_nibbles_32(const uint8_t * rsi)
 {
     const __m128i tmp = _mm_loadu_si128((const __m128i *)rsi);
-    const __m256i bytes = _mm256_set_m128i(_mm_srli_epi16(tmp, 4), tmp);
+    const __m256i bytes = MM256_SET_M128I(_mm_srli_epi16(tmp, 4), tmp);
     const __m256i lowMask = _mm256_set1_epi8( 0xF );
     return _mm256_and_si256(lowMask, bytes);
 }
@@ -606,7 +611,7 @@ static inline __m256i bytes_from_bits_32(const uint8_t * x) {
     bytesh = _mm_or_si128(bytesh, bit_mask);
     bytesl = _mm_cmpeq_epi8(bytesl, _mm_set1_epi64x(-1));
     bytesh = _mm_cmpeq_epi8(bytesh, _mm_set1_epi64x(-1));
-    return _mm256_set_m128i(bytesh, bytesl);
+    return MM256_SET_M128I(bytesh, bytesl);
 }
 
 // Unpack 32 4-bit fields into 32 bytes
@@ -619,15 +624,15 @@ static inline __m256i bytes_from_nibbles_32(const uint8_t * rsi)
     const __m128i lowMask = _mm_set1_epi8(0xF);
     tmpl = _mm_and_si128(lowMask, tmpl);
     tmph = _mm_and_si128(lowMask, tmph);
-    return _mm256_set_m128i(tmph, tmpl);
+    return MM256_SET_M128I(tmph, tmpl);
 }
 
 // add int16_t pairwise and return as float vector
 static inline __m256 sum_i16_pairs_float(const __m128i xh, const __m128i xl) {
     const __m128i ones = _mm_set1_epi16(1);
     const __m128i summed_pairsl = _mm_madd_epi16(ones, xl);
     const __m128i summed_pairsh = _mm_madd_epi16(ones, xh);
-    const __m256i summed_pairs = _mm256_set_m128i(summed_pairsh, summed_pairsl);
+    const __m256i summed_pairs = MM256_SET_M128I(summed_pairsh, summed_pairsl);
     return _mm256_cvtepi32_ps(summed_pairs);
 }
 
@@ -2290,7 +2295,7 @@ static void ggml_vec_dot_q4_0_q8_0(const int n, float * restrict s, const void *
         const __m128i i32_1 = mul_sum_i8_pairs(bx, by);
 
         // Convert int32_t to float
-        __m256 p = _mm256_cvtepi32_ps(_mm256_set_m128i(i32_0, i32_1));
+        __m256 p = _mm256_cvtepi32_ps(MM256_SET_M128I(i32_0, i32_1));
 
         // Apply the scale, and accumulate
         acc = _mm256_add_ps(_mm256_mul_ps( d, p ), acc);
@@ -2766,7 +2771,7 @@ static void ggml_vec_dot_q5_0_q8_0(const int n, float * restrict s, const void *
         __m128i bxh = _mm256_extractf128_si256(bx, 1);
         bxl = _mm_or_si128(bxl, bxhil);
         bxh = _mm_or_si128(bxh, bxhih);
-        bx = _mm256_set_m128i(bxh, bxl);
+        bx = MM256_SET_M128I(bxh, bxl);
 
         const __m256i by = _mm256_loadu_si256((const __m256i *)y[i].qs);
 
@@ -3022,7 +3027,7 @@ static void ggml_vec_dot_q5_1_q8_1(const int n, float * restrict s, const void *
         __m128i bxh = _mm256_extractf128_si256(bx, 1);
         bxl = _mm_or_si128(bxl, bxhil);
         bxh = _mm_or_si128(bxh, bxhih);
-        bx = _mm256_set_m128i(bxh, bxl);
+        bx = MM256_SET_M128I(bxh, bxl);
 
         const __m256 dy = _mm256_set1_ps(y[i].d);
         const __m256i by = _mm256_loadu_si256((const __m256i *)y[i].qs);