ggml-cpu : fix AVX build

slaren · slaren · commit a3b291e81f59 · 2025-08-05T18:29:29.000+02:00
ggml-ci
diff --git a/ggml/src/ggml-cpu/arch/x86/quants.c b/ggml/src/ggml-cpu/arch/x86/quants.c
@@ -269,8 +269,8 @@ static inline __m256 quad_fp16_delta_float(const float x0, const float y0, const
 }
 
 static inline __m256 quad_mx_delta_float(const int8_t x0, const float y0, const int8_t x1, const float y1) {
-    return _mm256_set_m128(_mm_set1_ps(GGML_E8M0(x1) * GGML_CPU_FP16_TO_FP32(y1)),
-                           _mm_set1_ps(GGML_E8M0(x0) * GGML_CPU_FP16_TO_FP32(y0)));
+    return _mm256_set_m128(_mm_set1_ps(GGML_E8M0_TO_FP32_HALF(x1) * GGML_CPU_FP16_TO_FP32(y1)),
+                           _mm_set1_ps(GGML_E8M0_TO_FP32_HALF(x0) * GGML_CPU_FP16_TO_FP32(y0)));
 }
 #endif
 #elif defined(__SSSE3__)

Original file line number	Diff line number	Diff line change
`@@ -269,8 +269,8 @@ static inline __m256 quad_fp16_delta_float(const float x0, const float y0, const`
`269`	`269`	`}`
`270`	`270`
`271`	`271`	`static inline __m256 quad_mx_delta_float(const int8_t x0, const float y0, const int8_t x1, const float y1) {`
`272`		`- return _mm256_set_m128(_mm_set1_ps(GGML_E8M0(x1) * GGML_CPU_FP16_TO_FP32(y1)),`
`273`		`- _mm_set1_ps(GGML_E8M0(x0) * GGML_CPU_FP16_TO_FP32(y0)));`
	`272`	`+ return _mm256_set_m128(_mm_set1_ps(GGML_E8M0_TO_FP32_HALF(x1) * GGML_CPU_FP16_TO_FP32(y1)),`
	`273`	`+ _mm_set1_ps(GGML_E8M0_TO_FP32_HALF(x0) * GGML_CPU_FP16_TO_FP32(y0)));`
`274`	`274`	`}`
`275`	`275`	`#endif`
`276`	`276`	`#elif defined(__SSSE3__)`