gallivm: Use native packs and unpacks for the lerps

For the texturing packs, things looked pretty terrible. For every
lerp, we were repacking the values, and while those look sort of cheap
with 128bit, with 256bit we end up with 2 of them instead of just 1 but
worse, plus 2 extracts too (the unpack, however, works fine with a
single instruction, albeit only with llvm 3.8 - the vpmovzxbw).

Ideally we'd use more clever pack for llvmpipe backend conversion too
since we actually use the "wrong" shuffle (which is more work) when doing
the fs twiddle just so we end up with the wrong order for being able to
do native pack when converting from 2x8f -> 1x16b. But this requires some
refactoring, since the untwiddle is separate from conversion.

This is only used for avx2 256bit pack/unpack for now.

Improves openarena scores by 8% or so, though overall it's still pretty
disappointing how much faster 256bit vectors are even with avx2 (or
rather, aren't...). And, of course, eliminating the needless
packs/unpacks in the first place would eliminate most of that advantage
(not quite all) from this patch.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

1 files changed, 7 insertions, 7 deletions

diff --git a/src/gallium/auxiliary/gallivm/lp_bld_arit.c b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
index f5cacc4..3ea0734 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_arit.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
@@ -1046,14 +1046,14 @@ lp_build_mul(struct lp_build_context *bld,
       struct lp_type wide_type = lp_wider_type(type);
       LLVMValueRef al, ah, bl, bh, abl, abh, ab;
 
-      lp_build_unpack2(bld->gallivm, type, wide_type, a, &al, &ah);
-      lp_build_unpack2(bld->gallivm, type, wide_type, b, &bl, &bh);
+      lp_build_unpack2_native(bld->gallivm, type, wide_type, a, &al, &ah);
+      lp_build_unpack2_native(bld->gallivm, type, wide_type, b, &bl, &bh);
 
       /* PMULLW, PSRLW, PADDW */
       abl = lp_build_mul_norm(bld->gallivm, wide_type, al, bl);
       abh = lp_build_mul_norm(bld->gallivm, wide_type, ah, bh);
 
-      ab = lp_build_pack2(bld->gallivm, wide_type, type, abl, abh);
+      ab = lp_build_pack2_native(bld->gallivm, wide_type, type, abl, abh);
 
       return ab;
    }
@@ -1350,9 +1350,9 @@ lp_build_lerp(struct lp_build_context *bld,
 
       lp_build_context_init(&wide_bld, bld->gallivm, wide_type);
 
-      lp_build_unpack2(bld->gallivm, type, wide_type, x,  &xl,  &xh);
-      lp_build_unpack2(bld->gallivm, type, wide_type, v0, &v0l, &v0h);
-      lp_build_unpack2(bld->gallivm, type, wide_type, v1, &v1l, &v1h);
+      lp_build_unpack2_native(bld->gallivm, type, wide_type, x,  &xl,  &xh);
+      lp_build_unpack2_native(bld->gallivm, type, wide_type, v0, &v0l, &v0h);
+      lp_build_unpack2_native(bld->gallivm, type, wide_type, v1, &v1l, &v1h);
 
       /*
        * Lerp both halves.
@@ -1363,7 +1363,7 @@ lp_build_lerp(struct lp_build_context *bld,
       resl = lp_build_lerp_simple(&wide_bld, xl, v0l, v1l, flags);
       resh = lp_build_lerp_simple(&wide_bld, xh, v0h, v1h, flags);
 
-      res = lp_build_pack2(bld->gallivm, wide_type, type, resl, resh);
+      res = lp_build_pack2_native(bld->gallivm, wide_type, type, resl, resh);
    } else {
       res = lp_build_lerp_simple(bld, x, v0, v1, flags);
    }