diff options
author | Eric Anholt <[email protected]> | 2014-10-09 15:02:00 +0200 |
---|---|---|
committer | Eric Anholt <[email protected]> | 2014-10-09 21:47:06 +0200 |
commit | 0401f55fffc6e77807e6987e23d2709a1599d61e (patch) | |
tree | 5ce87cfe7978a07094587776668f141f78e217e9 | |
parent | 1cd8c1aab04c4da9aa6cbbd52460607b8416ce1b (diff) |
vc4: Optimize fmul(x, 0) and fmul(x, 1).
This was being generated frequently by matrix multiplies of 2 and
3-channel vertex attributes (which have the 0 or 1 loaded in the shader).
-rw-r--r-- | src/gallium/drivers/vc4/vc4_opt_algebraic.c | 45 |
1 files changed, 45 insertions, 0 deletions
diff --git a/src/gallium/drivers/vc4/vc4_opt_algebraic.c b/src/gallium/drivers/vc4/vc4_opt_algebraic.c index 8c6fcef5936..435814e6912 100644 --- a/src/gallium/drivers/vc4/vc4_opt_algebraic.c +++ b/src/gallium/drivers/vc4/vc4_opt_algebraic.c @@ -33,6 +33,7 @@ */ #include "vc4_qir.h" +#include "util/u_math.h" static bool debug; @@ -77,6 +78,16 @@ is_zero(struct vc4_compile *c, struct qinst **defs, struct qreg reg) c->uniform_data[reg.index] == 0); } +static bool +is_1f(struct vc4_compile *c, struct qinst **defs, struct qreg reg) +{ + reg = follow_movs(defs, reg); + + return (reg.file == QFILE_UNIF && + c->uniform_contents[reg.index] == QUNIFORM_CONSTANT && + c->uniform_data[reg.index] == fui(1.0)); +} + static void replace_with_mov(struct vc4_compile *c, struct qinst *inst, struct qreg arg) { @@ -87,6 +98,30 @@ replace_with_mov(struct vc4_compile *c, struct qinst *inst, struct qreg arg) dump_to(c, inst); } +static bool +fmul_replace_zero(struct vc4_compile *c, + struct qinst **defs, + struct qinst *inst, + int arg) +{ + if (!is_zero(c, defs, inst->src[arg])) + return false; + replace_with_mov(c, inst, inst->src[arg]); + return true; +} + +static bool +fmul_replace_one(struct vc4_compile *c, + struct qinst **defs, + struct qinst *inst, + int arg) +{ + if (!is_1f(c, defs, inst->src[arg])) + return false; + replace_with_mov(c, inst, inst->src[1 - arg]); + return true; +} + bool qir_opt_algebraic(struct vc4_compile *c) { @@ -177,6 +212,16 @@ qir_opt_algebraic(struct vc4_compile *c) } break; + case QOP_FMUL: + if (fmul_replace_zero(c, defs, inst, 0) || + fmul_replace_zero(c, defs, inst, 1) || + fmul_replace_one(c, defs, inst, 0) || + fmul_replace_one(c, defs, inst, 1)) { + progress = true; + break; + } + break; + default: break; } |