[Mesa-dev] [PATCH] r600/sb: Do not distribute neg in expr_handler::fold_assoc() when folding multiplications.

Glenn Kennard glenn.kennard at gmail.com
Wed Mar 9 22:19:57 UTC 2016


On Wed, 09 Mar 2016 09:58:48 +0100, Xavier B <xavierb at gmail.com> wrote:

> From: xavier <xavierb at gmail.com>
>
> Previously it was doing this transformation for a Trine 3 shader:
>      MUL     R6.x.12,    R13.x.23, 0.5|3f000000
> -    MULADD     R4.x.12,    -R6.x.12, 2|40000000, 1|3f800000
> +    MULADD     R4.x.12,    -R13.x.23, -1|bf800000, 1|3f800000
>
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94412
> Signed-off-by: Xavier Bouchoux <xavierb at gmail.com>
> ---
>  src/gallium/drivers/r600/sb/sb_expr.cpp | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/src/gallium/drivers/r600/sb/sb_expr.cpp b/src/gallium/drivers/r600/sb/sb_expr.cpp
> index 556a05d..3dd3a48 100644
> --- a/src/gallium/drivers/r600/sb/sb_expr.cpp
> +++ b/src/gallium/drivers/r600/sb/sb_expr.cpp
> @@ -598,9 +598,13 @@ bool expr_handler::fold_assoc(alu_node *n) {
> 	unsigned op = n->bc.op;
>  	bool allow_neg = false, cur_neg = false;
> +	bool distribute_neg = false;
> 	switch(op) {
>  	case ALU_OP2_ADD:
> +		distribute_neg = true;

> +		allow_neg = true;

I'm not sure this change belongs in this patch, or even if its correct.

> +		break;
>  	case ALU_OP2_MUL:
>  	case ALU_OP2_MUL_IEEE:
>  		allow_neg = true;
> @@ -632,7 +636,7 @@ bool expr_handler::fold_assoc(alu_node *n) {
>  		if (v1->is_const()) {
>  			literal arg = v1->get_const_value();
>  			apply_alu_src_mod(a->bc, 1, arg);
> -			if (cur_neg)
> +			if (cur_neg && distribute_neg)
>  				arg.f = -arg.f;
> 			if (a == n)
> @@ -660,7 +664,7 @@ bool expr_handler::fold_assoc(alu_node *n) {
>  		if (v0->is_const()) {
>  			literal arg = v0->get_const_value();
>  			apply_alu_src_mod(a->bc, 0, arg);
> -			if (cur_neg)
> +			if (cur_neg && distribute_neg)
>  				arg.f = -arg.f;
> 			if (last_arg == 0) {


With the allow_neg change removed, patch is
Reviewed-by: Glenn Kennard <glenn.kennard at gmail.com>


More information about the mesa-dev mailing list