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

freedreno/a3xx/compiler: fix neg mov's 

create_mov() was fixed up to handle neg/abs properly for interal mov's,
using absneg.f, but forgot to fix it for TGSI MOV's.  The problem with
using add.f to handle negated mov's is that we can only take a single
const reg src.  So:

  MOV TEMP[n], -CONST[m]

would turn into:

  add.f Rdst, (neg)CONST[m], 0.0

which would not work.  Anyways, just remove the extra code and always
use create_mov() which DTRT.

This fixes piglit vs-op-neg-int test.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
This commit is contained in:
Rob Clark 2014-04-09 09:01:32 -04:00
parent 4d641803e8
commit 9604e31dc9
1 changed files with 1 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
src/gallium/drivers/freedreno/a3xx/fd3_compiler.c
View File

@ -1599,26 +1599,7 @@ instr_cat1(const struct instr_translater *t,
{
struct tgsi_dst_register *dst = get_dst(ctx, inst);
struct tgsi_src_register *src = &inst->Src[0].Register;
/* mov instructions can't handle a negate on src: */
if (src->Negate) {
struct tgsi_src_register constval;
struct ir3_instruction *instr;
/* since right now, we are using uniformly either TYPE_F16 or
* TYPE_F32, and we don't utilize the conversion possibilities
* of mov instructions, we can get away with substituting an
* add.f which can handle negate. Might need to revisit this
* in the future if we start supporting widening/narrowing or
* conversion to/from integer..
*/
instr = instr_create(ctx, 2, OPC_ADD_F);
get_immediate(ctx, &constval, fui(0.0));
vectorize(ctx, instr, dst, 2, src, 0, &constval, 0);
} else {
create_mov(ctx, dst, src);
/* create_mov() generates vector sequence, so no vectorize() */
}
create_mov(ctx, dst, src);
put_dst(ctx, inst, dst);
}