On Mon, Dec 22, 2014 at 7:29 PM, Ben Widawsky
<benjamin.widaw...@intel.com> wrote:
> "A source modifier must not be used on src1 for the macro operation. This
> applies to both mul and mach of the macro. If source modifier is required, an
> additional mov instruction may be used before the macro."
>
> Today, we only use MACH in the macro, so this assertion should always hold.
>
> This assertion applies to mul too as mentioned above. It is harder to detect
> mul
> when used in the macro operation, so that is left for now.
A MUL is used in a macro operation if and only if its destination if
the accumulator, right?
Assuming so, we should do the assertions in brw_MUL.
> "Accumulator is an implicit source and thus cannot be an explicit source
> operand."
>
> Signed-off-by: Ben Widawsky <b...@bwidawsk.net>
> ---
> src/mesa/drivers/dri/i965/brw_eu_emit.c | 46
> ++++++++++++++++++++++++++++++++-
> 1 file changed, 45 insertions(+), 1 deletion(-)
>
> diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c
> b/src/mesa/drivers/dri/i965/brw_eu_emit.c
> index 046e293..51bf1b9 100644
> --- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
> +++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
> @@ -1032,7 +1032,6 @@ ALU2(ASR)
> ALU1(FRC)
> ALU1(RNDD)
> ALU2(MAC)
> -ALU2(MACH)
> ALU1(LZD)
> ALU2(DP4)
> ALU2(DPH)
> @@ -1136,6 +1135,51 @@ brw_LINE(struct brw_compile *p, struct brw_reg dest,
> }
>
> brw_inst *
> +brw_MACH(struct brw_compile *p, struct brw_reg dest,
> + struct brw_reg src0, struct brw_reg src1)
> +{
> + const struct brw_context *brw = p->brw;
> + (void) brw; /* Get rid of unused variable warning with NDEBUG */
We do this enough that no comment is needed.
> +
> + /* A source modifier must not be used on src1 for the macro operation. */
> + assert(brw->gen < 8 || !src1.negate);
> + assert(brw->gen < 8 || !src1.abs);
> +
> + /* Source and destination operands must be DWord integers. Source and
> + * destination must be of the same type, signed integer or unsigned
> integer.
> + */
> + assert(brw->gen < 8 || src0.type == src1.type);
> + assert(brw->gen < 8 || type_is_dword(src0.type));
> +
> + /* If dst is UD, src0 and src1 may be UD and/or D. However, if any of src0
> + * and src1 is D, source modifier (abs) must be present to convert it to
> + * match with dst.
> + *
> + * NOTE: There is a src1 exception above
> + */
> + assert(brw->gen < 8 ||
> + dest.type == BRW_REGISTER_TYPE_D ||
> + (src1.type == BRW_REGISTER_TYPE_UD &&
> + src0.type == BRW_REGISTER_TYPE_UD) ||
Weird to test src1 and then src0, but okay.
> + (src0.negate));
Extra parenthesis around src0.negate.
> +
> + /* If dest is D, src0 and src1 must also be D. They cannot be UD as it may
> + * cause unexpected overflow because the computed results are limited to
> 64
> + * bits.
> + */
> + assert(brw->gen < 8 ||
> + dest.type != BRW_REGISTER_TYPE_D ||
> + (src1.type == BRW_REGISTER_TYPE_D && src0.type ==
> BRW_REGISTER_TYPE_D));
src1 and then src0. Weird but okay.
> +
> + assert(src0.file != BRW_ARCHITECTURE_REGISTER_FILE ||
> + src0.nr != BRW_ARF_ACCUMULATOR);
> + assert(src1.file != BRW_ARCHITECTURE_REGISTER_FILE ||
> + src1.nr != BRW_ARF_ACCUMULATOR);
Some tabs in this patch. Spaces only.
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
