Good find.
On this sort of issues it's a good idea to search the code for similar cases:
- it looks like this is done differently above:
case TGSI_OPCODE_ABS:
alu.src[0].abs = 1;
if (alu.src[0].neg)
alu.src[0].neg = 0;
break;
so an inline helper to set abs correctly would be cleaner & more consistent.
- looking at src/gallium/docs/source/tgsi.rst there are other opcodes with |..|
in their formula which is not observed, in particular
src/gallium/drivers/r600/r600_shader.c:tgsi_log() does not set abs where the
spec says it should.
Jose
----- Original Message -----
> Need to clear NEG bit because it applies after ABS, e.g. "RSQ ...,
> -1"
> uses -|1| as operand.
>
> Signed-off-by: Vadim Girlin <vadimgir...@gmail.com>
> ---
> src/gallium/drivers/r600/r600_shader.c | 1 +
> 1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/src/gallium/drivers/r600/r600_shader.c
> b/src/gallium/drivers/r600/r600_shader.c
> index f3cbf98..5aad3e3 100644
> --- a/src/gallium/drivers/r600/r600_shader.c
> +++ b/src/gallium/drivers/r600/r600_shader.c
> @@ -1506,6 +1506,7 @@ static int tgsi_rsq(struct r600_shader_ctx
> *ctx)
> for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
> r600_bc_src(&alu.src[i], &ctx->src[i], 0);
> alu.src[i].abs = 1;
> + alu.src[i].neg = 0;
> }
> alu.dst.sel = ctx->temp_reg;
> alu.dst.write = 1;
> --
> 1.7.6
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
