[Mesa-dev] [PATCH 15/16] i965/fs: Add support for removing MOV.NZ instructions.
Jason Ekstrand
jason at jlekstrand.net
Tue Jan 20 14:58:47 PST 2015
On Mon, Jan 19, 2015 at 3:31 PM, Matt Turner <mattst88 at gmail.com> wrote:
> For some reason, we occasionally write the flag register with a MOV.NZ
> instruction:
>
> add(8) g25<1>F -g6<0,1,0>F g15<8,8,1>F
> cmp.l.f0(8) g26<1>D g25<8,8,1>F 0F
> mov.nz.f0(8) null g26<8,8,1>D
>
> A MOV.NZ instruction on the result of a CMP is like comparing for
> equality with true in C. It's useless. Removing it allows us to
> generate:
>
> add.l.f0(8) null -g6<0,1,0>F g15<8,8,1>F
>
> total instructions in shared programs: 5955701 -> 5951657 (-0.07%)
> instructions in affected programs: 302910 -> 298866 (-1.34%)
> GAINED: 1
> LOST: 0
> ---
> .../drivers/dri/i965/brw_fs_cmod_propagation.cpp | 23 ++++++++++++++--
> .../drivers/dri/i965/test_fs_cmod_propagation.cpp | 32
> ++++++++++++++++++++++
> 2 files changed, 52 insertions(+), 3 deletions(-)
>
> diff --git a/src/mesa/drivers/dri/i965/brw_fs_cmod_propagation.cpp
> b/src/mesa/drivers/dri/i965/brw_fs_cmod_propagation.cpp
> index b521350..dd89512 100644
> --- a/src/mesa/drivers/dri/i965/brw_fs_cmod_propagation.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_fs_cmod_propagation.cpp
> @@ -57,12 +57,20 @@ opt_cmod_propagation_local(fs_visitor *v, bblock_t
> *block)
> foreach_inst_in_block_reverse_safe(fs_inst, inst, block) {
> ip--;
>
> - if (inst->opcode != BRW_OPCODE_CMP ||
> + if ((inst->opcode != BRW_OPCODE_CMP &&
> + inst->opcode != BRW_OPCODE_MOV) ||
> inst->predicate != BRW_PREDICATE_NONE ||
> !inst->dst.is_null() ||
> inst->src[0].file != GRF ||
> - inst->src[0].abs ||
> - !inst->src[1].is_zero())
> + inst->src[0].abs)
> + continue;
> +
> + if (inst->opcode == BRW_OPCODE_CMP && !inst->src[1].is_zero())
> + continue;
> +
> + if (inst->opcode == BRW_OPCODE_MOV &&
> + (inst->conditional_mod != BRW_CONDITIONAL_NZ ||
> + inst->src[0].negate))
>
I think negate is ok here. I'm not 100% sure on the symantics of move.nz,
but if it's a "!= 0" then negation shouldn't matter. If it only considers
the bottom bit then negation shouldn't matter there either.
> continue;
>
> bool read_flag = false;
> @@ -72,6 +80,15 @@ opt_cmod_propagation_local(fs_visitor *v, bblock_t
> *block)
> scan_inst->dst.reg == inst->src[0].reg &&
> scan_inst->dst.reg_offset == inst->src[0].reg_offset &&
> !scan_inst->is_partial_write()) {
> + if (inst->opcode == BRW_OPCODE_MOV) {
> + if (!scan_inst->writes_flag())
> + break;
> +
> + inst->remove(block);
> + progress = true;
> + break;
> + }
> +
> enum brw_conditional_mod cond =
> inst->src[0].negate ?
> brw_invert_cmod(inst->conditional_mod)
> : inst->conditional_mod;
> diff --git a/src/mesa/drivers/dri/i965/test_fs_cmod_propagation.cpp
> b/src/mesa/drivers/dri/i965/test_fs_cmod_propagation.cpp
> index 15f685e..9541597 100644
> --- a/src/mesa/drivers/dri/i965/test_fs_cmod_propagation.cpp
> +++ b/src/mesa/drivers/dri/i965/test_fs_cmod_propagation.cpp
> @@ -343,3 +343,35 @@ TEST_F(cmod_propagation_test, negate)
> EXPECT_EQ(BRW_OPCODE_ADD, instruction(block0, 0)->opcode);
> EXPECT_EQ(BRW_CONDITIONAL_L, instruction(block0, 0)->conditional_mod);
> }
> +
> +TEST_F(cmod_propagation_test, movnz)
> +{
> + fs_reg dest(v, glsl_type::float_type);
> + fs_reg src0(v, glsl_type::float_type);
> + fs_reg src1(v, glsl_type::float_type);
> + v->emit(BRW_OPCODE_CMP, dest, src0, src1)
> + ->conditional_mod = BRW_CONDITIONAL_GE;
> + v->emit(BRW_OPCODE_MOV, v->reg_null_f, dest)
> + ->conditional_mod = BRW_CONDITIONAL_NZ;
> +
> + /* = Before =
> + *
> + * 0: cmp.ge.f0(8) dest src0 src1
> + * 1: mov.nz.f0(8) null dest
> + *
> + * = After =
> + * 0: cmp.ge.f0(8) dest src0 src1
> + */
> +
> + v->calculate_cfg();
> + bblock_t *block0 = v->cfg->blocks[0];
> +
> + EXPECT_EQ(0, block0->start_ip);
> + EXPECT_EQ(1, block0->end_ip);
> +
> + EXPECT_TRUE(cmod_propagation(v));
> + EXPECT_EQ(0, block0->start_ip);
> + EXPECT_EQ(0, block0->end_ip);
> + EXPECT_EQ(BRW_OPCODE_CMP, instruction(block0, 0)->opcode);
> + EXPECT_EQ(BRW_CONDITIONAL_GE, instruction(block0, 0)->conditional_mod);
> +}
> --
> 2.0.4
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20150120/fe8b8270/attachment.html>
More information about the mesa-dev
mailing list