diff options
| author | Eric Anholt <[email protected]> | 2009-11-11 11:58:12 -0800 |
|---|---|---|
| committer | Eric Anholt <[email protected]> | 2009-11-13 13:18:56 -0800 |
| commit | 91bd593109c71310fb7e101c5f73a14f1bbd5f93 (patch) | |
| tree | cad0b0e49e78fbe49fa836cc961e01c0180e224f /src | |
| parent | d6690ce15fb8c7c6abf1bc0d847c1d2da2c33904 (diff) | |
i965: Avoid moving the current value back into the accumulator for MAD.
This is a 2.9% (+/-.3%) performance win for my GL demo, which hits MAD
sequences for matrix transforms.
Diffstat (limited to 'src')
| -rw-r--r-- | src/mesa/drivers/dri/i965/brw_vs_emit.c | 35 |
1 files changed, 34 insertions, 1 deletions
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c index 15154c3b8e1..f7b07266362 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c @@ -1271,6 +1271,38 @@ post_vs_emit( struct brw_vs_compile *c, } } +static GLboolean +accumulator_contains(struct brw_vs_compile *c, struct brw_reg val) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *prev_insn = &p->store[p->nr_insn - 1]; + + if (p->nr_insn == 0) + return GL_FALSE; + + if (val.address_mode != BRW_ADDRESS_DIRECT) + return GL_FALSE; + + switch (prev_insn->header.opcode) { + case BRW_OPCODE_MOV: + case BRW_OPCODE_MAC: + case BRW_OPCODE_MUL: + if (prev_insn->header.access_mode == BRW_ALIGN_16 && + prev_insn->header.execution_size == val.width && + prev_insn->bits1.da1.dest_reg_file == val.file && + prev_insn->bits1.da1.dest_reg_type == val.type && + prev_insn->bits1.da1.dest_address_mode == val.address_mode && + prev_insn->bits1.da1.dest_reg_nr == val.nr && + prev_insn->bits1.da16.dest_subreg_nr == val.subnr / 16 && + prev_insn->bits1.da16.dest_writemask == 0xf) + return GL_TRUE; + else + return GL_FALSE; + default: + return GL_FALSE; + } +} + static uint32_t get_predicate(const struct prog_instruction *inst) { @@ -1449,7 +1481,8 @@ void brw_vs_emit(struct brw_vs_compile *c ) unalias3(c, dst, args[0], args[1], args[2], emit_lrp_noalias); break; case OPCODE_MAD: - brw_MOV(p, brw_acc_reg(), args[2]); + if (!accumulator_contains(c, args[2])) + brw_MOV(p, brw_acc_reg(), args[2]); brw_MAC(p, dst, args[0], args[1]); break; case OPCODE_MAX: |