gallivm: allow arch rounding with avx512
authorTim Rowley <timothy.o.rowley@intel.com>
Wed, 1 Nov 2017 18:22:47 +0000 (13:22 -0500)
committerTim Rowley <timothy.o.rowley@intel.com>
Thu, 2 Nov 2017 15:24:54 +0000 (10:24 -0500)
Fixes piglit vs-roundeven-{float,vec[234]} with simd16 VS.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
src/gallium/auxiliary/gallivm/lp_bld_arit.c

index cf1958b3b6335779568cde7874f7099235a9c563..a1edd349f1f1897d0323c1215b4a3d4d300914ef 100644 (file)
@@ -1953,7 +1953,8 @@ arch_rounding_available(const struct lp_type type)
 {
    if ((util_cpu_caps.has_sse4_1 &&
        (type.length == 1 || type.width*type.length == 128)) ||
-       (util_cpu_caps.has_avx && type.width*type.length == 256))
+       (util_cpu_caps.has_avx && type.width*type.length == 256) ||
+       (util_cpu_caps.has_avx512f && type.width*type.length == 512))
       return TRUE;
    else if ((util_cpu_caps.has_altivec &&
             (type.width == 32 && type.length == 4)))