AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

gallivm: implement the correct version of LRP

Browse Source 
The previous version has precision issues. This can be a problem
with tessellation. Sadly, I can't find the article where I read it
anymore. I'm not sure if the unsafe-fp-math flag would be enough to revert
this.

v2: added the comment
This commit is contained in:
Marek Olšák
2015-10-10 22:05:58 +02:00
parent a2197cac7f
commit 529c5e7740
1 changed files with 13 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/gallium/auxiliary/gallivm/lp_bld_tgsi_action.c
+13 -6
View File
@@ -538,12 +538,19 @@ lrp_emit(
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
LLVMValueRef tmp;
tmp = lp_build_emit_llvm_binary(bld_base, TGSI_OPCODE_SUB,
emit_data->args[1],
emit_data->args[2]);
emit_data->output[emit_data->chan] = lp_build_emit_llvm_ternary(bld_base,
TGSI_OPCODE_MAD, emit_data->args[0], tmp, emit_data->args[2]);
struct lp_build_context *bld = &bld_base->base;
LLVMValueRef inv, a, b;
/* This uses the correct version: (1 - t)*a + t*b
*
* An alternative version is "a + t*(b-a)". The problem is this version
* doesn't return "b" for t = 1, because "a + (b-a)" isn't equal to "b"
* because of the floating-point rounding.
*/
inv = lp_build_sub(bld, bld_base->base.one, emit_data->args[0]);
a = lp_build_mul(bld, emit_data->args[1], emit_data->args[0]);
b = lp_build_mul(bld, emit_data->args[2], inv);
emit_data->output[emit_data->chan] = lp_build_add(bld, a, b);
}
/* TGSI_OPCODE_MAD */