summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorRoland Scheidegger <sroland@vmware.com>2010-10-08 21:08:49 +0200
committerRoland Scheidegger <sroland@vmware.com>2010-10-09 00:36:38 +0200
commit2cc6da85d6e60e80d0b5b86fe42f8f82073b5d51 (patch)
tree4148eacacfbb1f33250c202fee70974310677d61
parent318bb080b04c961141a962d440d0e1296f8adda6 (diff)
downloadexternal_mesa3d-2cc6da85d6e60e80d0b5b86fe42f8f82073b5d51.zip
external_mesa3d-2cc6da85d6e60e80d0b5b86fe42f8f82073b5d51.tar.gz
external_mesa3d-2cc6da85d6e60e80d0b5b86fe42f8f82073b5d51.tar.bz2
gallivm: avoid unnecessary URem in linear wrap repeat case
Haven't looked at what code this exactly generates but URem can't be fast. Instead of using two URem only use one and replace the second one with select/add (this is what the corresponding aos code already does).
-rw-r--r--src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c10
1 files changed, 7 insertions, 3 deletions
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
index 242e8d3..b020782 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
@@ -251,19 +251,23 @@ lp_build_sample_wrap_linear(struct lp_build_sample_context *bld,
coord = lp_build_sub(coord_bld, coord, half);
/* convert to int, compute lerp weight */
lp_build_ifloor_fract(coord_bld, coord, &coord0, &weight);
- coord1 = lp_build_add(uint_coord_bld, coord0, uint_coord_bld->one);
/* repeat wrap */
if (is_pot) {
+ coord1 = lp_build_add(uint_coord_bld, coord0, uint_coord_bld->one);
coord0 = LLVMBuildAnd(bld->builder, coord0, length_minus_one, "");
coord1 = LLVMBuildAnd(bld->builder, coord1, length_minus_one, "");
}
else {
/* Add a bias to the texcoord to handle negative coords */
LLVMValueRef bias = lp_build_mul_imm(uint_coord_bld, length, 1024);
+ LLVMValueRef mask;
coord0 = LLVMBuildAdd(bld->builder, coord0, bias, "");
- coord1 = LLVMBuildAdd(bld->builder, coord1, bias, "");
coord0 = LLVMBuildURem(bld->builder, coord0, length, "");
- coord1 = LLVMBuildURem(bld->builder, coord1, length, "");
+ mask = lp_build_compare(bld->builder, int_coord_bld->type,
+ PIPE_FUNC_NOTEQUAL, coord0, length_minus_one);
+ coord1 = LLVMBuildAnd(bld->builder,
+ lp_build_add(uint_coord_bld, coord0, uint_coord_bld->one),
+ mask, "");
}
break;