fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-25 04:20:08 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

r300: fix negate mask computation when merging movs

Browse source 
The main problem here is we can have a negate bit set for an unused
channel, so we can't just OR together the negates when channel merging.
Right now the bug is hidden because how we run the pass order, but
that will change in a later commit. Add some helpers for merging of the
negates, they will be also used more in a later commits. As a bonus
construct the new source separatelly and only rewrite the original
instructions after checking that the final swizzle is valid.

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Filip Gawin <filip@gawin.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17560>
This commit is contained in:
Pavel Ondračka 2022-07-12 10:17:13 +02:00
parent aeae7b812c
commit 6286e48e5d
1 changed files with 28 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

38
src/gallium/drivers/r300/compiler/radeon_optimize.c
View file

@ -900,12 +900,28 @@ static unsigned int merge_swizzles(unsigned int swz1, unsigned int swz2) {
return new_swz;
}
/* Sets negate to 0 for unused channels. */
static unsigned int clean_negate(struct rc_src_register src)
{
unsigned int new_negate = 0;
for (unsigned int chan = 0; chan < 4; chan++) {
unsigned int swz = GET_SWZ(src.Swizzle, chan);
if (swz != RC_SWIZZLE_UNUSED)
new_negate |= src.Negate & (1 << chan);
}
return new_negate;
}
static unsigned int merge_negates(struct rc_src_register src1, struct rc_src_register src2)
{
return clean_negate(src1) | clean_negate(src2);
}
static int merge_movs(struct radeon_compiler * c, struct rc_instruction * inst)
{
unsigned int orig_dst_reg = inst->U.I.DstReg.Index;
unsigned int orig_dst_file = inst->U.I.DstReg.File;
unsigned int orig_dst_wmask = inst->U.I.DstReg.WriteMask;
unsigned int orig_src_reg = inst->U.I.SrcReg[0].Index;
unsigned int orig_src_file = inst->U.I.SrcReg[0].File;
struct rc_instruction * cur = inst;
@ -944,17 +960,19 @@ static int merge_movs(struct radeon_compiler * c, struct rc_instruction * inst)
orig_src_file == RC_FILE_NONE) {
cur->U.I.DstReg.WriteMask |= orig_dst_wmask;
struct rc_src_register src;
if (cur->U.I.SrcReg[0].File == RC_FILE_NONE) {
cur->U.I.SrcReg[0].File = orig_src_file;
cur->U.I.SrcReg[0].Index = orig_src_reg;
cur->U.I.SrcReg[0].Abs = inst->U.I.SrcReg[0].Abs;
cur->U.I.SrcReg[0].RelAddr = inst->U.I.SrcReg[0].RelAddr;
src = inst->U.I.SrcReg[0];
} else {
src = cur->U.I.SrcReg[0];
}
cur->U.I.SrcReg[0].Swizzle =
merge_swizzles(cur->U.I.SrcReg[0].Swizzle,
inst->U.I.SrcReg[0].Swizzle);
cur->U.I.SrcReg[0].Negate |= inst->U.I.SrcReg[0].Negate;
src.Swizzle = merge_swizzles(cur->U.I.SrcReg[0].Swizzle,
inst->U.I.SrcReg[0].Swizzle);
src.Negate = merge_negates(inst->U.I.SrcReg[0], cur->U.I.SrcReg[0]);
if (!c->SwizzleCaps->IsNative(RC_OPCODE_MOV, src))
return 0;
cur->U.I.DstReg.WriteMask |= orig_dst_wmask;
cur->U.I.SrcReg[0] = src;
/* finally delete the original mov */
rc_remove_instruction(inst);