[Mesa-dev] [PATCH] intel: Don't propagate conditional modifiers if a UD source is negated
Ian Romanick
idr at freedesktop.org
Tue Oct 9 16:24:10 UTC 2018
On 10/09/2018 09:00 AM, Jason Ekstrand wrote:
> This fixes a bug uncovered by my NIR integer division by constant
> optimization series.
>
> Fixes: 19f9cb72c8b "i965/fs: Add pass to propagate conditional..."
> Fixes: 627f94b72e0 "i965/vec4: adding vec4_cmod_propagation..."
> Cc: Matt Turner <mattst88 at gmail.com>
> ---
> .../compiler/brw_fs_cmod_propagation.cpp | 25 ++++++++++++++++---
> src/intel/compiler/brw_reg.h | 9 +++++++
> .../compiler/brw_vec4_cmod_propagation.cpp | 24 ++++++++++++++++--
> 3 files changed, 53 insertions(+), 5 deletions(-)
>
> diff --git a/src/intel/compiler/brw_fs_cmod_propagation.cpp b/src/intel/compiler/brw_fs_cmod_propagation.cpp
> index 5fb522f810f..4fdd04a9983 100644
> --- a/src/intel/compiler/brw_fs_cmod_propagation.cpp
> +++ b/src/intel/compiler/brw_fs_cmod_propagation.cpp
> @@ -25,6 +25,25 @@
> #include "brw_cfg.h"
> #include "brw_eu.h"
>
> +static bool
> +can_add_cmod_to_inst(const fs_inst *inst)
> +{
> + if (!inst->can_do_cmod())
> + return false;
> +
> + /* The accumulator result appears to get used for the conditional modifier
> + * generation. When negating a UD value, there is a 33rd bit generated for
> + * the sign in the accumulator value, so now you can't check, for example,
> + * equality with a 32-bit value. See piglit fs-op-neg-uvec4.
> + */
> + for (unsigned i = 0; i < inst->sources; i++) {
> + if (type_is_unsigned_int(inst->src[i].type) && inst->src[i].negate)
> + return false;
> + }
> +
> + return true;
> +}
> +
This probably should go after the @file header comment. :)
Also... it looks like this patch replaces every caller of
::can_do_cmod() with can_add_cmod_to_inst. Maybe just change
can_do_cmod? If you do that, I'd support changing the name to
can_add_cmod_to_inst.
> /** @file brw_fs_cmod_propagation.cpp
> *
> * Implements a pass that propagates the conditional modifier from a CMP x 0.0
> @@ -91,7 +110,7 @@ cmod_propagate_cmp_to_add(const gen_device_info *devinfo, bblock_t *block,
> negate ? brw_swap_cmod(inst->conditional_mod)
> : inst->conditional_mod;
>
> - if (scan_inst->can_do_cmod() &&
> + if (can_add_cmod_to_inst(scan_inst) &&
> ((!read_flag && scan_inst->conditional_mod == BRW_CONDITIONAL_NONE) ||
> scan_inst->conditional_mod == cond)) {
> scan_inst->conditional_mod = cond;
> @@ -146,7 +165,7 @@ cmod_propagate_not(const gen_device_info *devinfo, bblock_t *block,
> scan_inst->exec_size != inst->exec_size)
> break;
>
> - if (scan_inst->can_do_cmod() &&
> + if (can_add_cmod_to_inst(scan_inst) &&
> ((!read_flag && scan_inst->conditional_mod == BRW_CONDITIONAL_NONE) ||
> scan_inst->conditional_mod == cond)) {
> scan_inst->conditional_mod = cond;
> @@ -316,7 +335,7 @@ opt_cmod_propagation_local(const gen_device_info *devinfo, bblock_t *block)
> inst->src[0].negate ? brw_swap_cmod(inst->conditional_mod)
> : inst->conditional_mod;
>
> - if (scan_inst->can_do_cmod() &&
> + if (can_add_cmod_to_inst(scan_inst) &&
> ((!read_flag && scan_inst->conditional_mod == BRW_CONDITIONAL_NONE) ||
> scan_inst->conditional_mod == cond)) {
> scan_inst->conditional_mod = cond;
> diff --git a/src/intel/compiler/brw_reg.h b/src/intel/compiler/brw_reg.h
> index ac12ab3d2dd..46d66198a1d 100644
> --- a/src/intel/compiler/brw_reg.h
> +++ b/src/intel/compiler/brw_reg.h
> @@ -376,6 +376,15 @@ brw_int_type(unsigned sz, bool is_signed)
> }
> }
>
> +static inline bool
> +type_is_unsigned_int(enum brw_reg_type tp)
> +{
> + return tp == BRW_REGISTER_TYPE_UB ||
> + tp == BRW_REGISTER_TYPE_UW ||
> + tp == BRW_REGISTER_TYPE_UD ||
> + tp == BRW_REGISTER_TYPE_UQ;
> +}
> +
> /**
> * Construct a brw_reg.
> * \param file one of the BRW_x_REGISTER_FILE values
> diff --git a/src/intel/compiler/brw_vec4_cmod_propagation.cpp b/src/intel/compiler/brw_vec4_cmod_propagation.cpp
> index a1d46dc8dca..392e283d1f8 100644
> --- a/src/intel/compiler/brw_vec4_cmod_propagation.cpp
> +++ b/src/intel/compiler/brw_vec4_cmod_propagation.cpp
> @@ -46,6 +46,26 @@ writemasks_incompatible(const vec4_instruction *earlier,
> (later->dst.writemask & ~earlier->dst.writemask) != 0;
> }
>
> +static bool
> +can_add_cmod_to_inst(const vec4_instruction *inst)
> +{
> + if (!inst->can_do_cmod())
> + return false;
> +
> + /* The accumulator result appears to get used for the conditional modifier
> + * generation. When negating a UD value, there is a 33rd bit generated for
> + * the sign in the accumulator value, so now you can't check, for example,
> + * equality with a 32-bit value. See piglit fs-op-neg-uvec4.
> + */
> + for (unsigned i = 0; i < 3; i++) {
> + if (inst->src[i].file != BAD_FILE &&
> + type_is_unsigned_int(inst->src[i].type) && inst->src[i].negate)
> + return false;
> + }
> +
> + return true;
> +}
> +
> static bool
> opt_cmod_propagation_local(bblock_t *block)
> {
> @@ -132,7 +152,7 @@ opt_cmod_propagation_local(bblock_t *block)
> negate ? brw_swap_cmod(inst->conditional_mod)
> : inst->conditional_mod;
>
> - if (scan_inst->can_do_cmod() &&
> + if (can_add_cmod_to_inst(scan_inst) &&
> ((!read_flag && scan_inst->conditional_mod == BRW_CONDITIONAL_NONE) ||
> scan_inst->conditional_mod == cond)) {
> scan_inst->conditional_mod = cond;
> @@ -229,7 +249,7 @@ opt_cmod_propagation_local(bblock_t *block)
> inst->src[0].negate ? brw_swap_cmod(inst->conditional_mod)
> : inst->conditional_mod;
>
> - if (scan_inst->can_do_cmod() &&
> + if (can_add_cmod_to_inst(scan_inst) &&
> ((!read_flag && scan_inst->conditional_mod == BRW_CONDITIONAL_NONE) ||
> scan_inst->conditional_mod == cond)) {
> scan_inst->conditional_mod = cond;
>
More information about the mesa-dev
mailing list