diff options
| author | Eric Anholt <[email protected]> | 2014-11-26 12:44:19 -0800 |
|---|---|---|
| committer | Eric Anholt <[email protected]> | 2014-12-01 11:00:23 -0800 |
| commit | 3fe4d8e1e39b47c9c5c4bfdd87300abd0c336a7e (patch) | |
| tree | 5ac8c5cee06176519262f90dabb3d304c120f655 /src/gallium/drivers/vc4/vc4_qir.h | |
| parent | 6958c404caf3f4b2219ef686e2beeeaf48664905 (diff) | |
vc4: Introduce scheduling of QPU instructions.
This doesn't reschedule much currently, just tries to fit things into the
regfile A/B write-versus-read slots (the cause of the improvements in
shader-db), and hide texture fetch latency by scheduling setup early and
results collection late (haven't performance tested it). This
infrastructure will be important for doing instruction pairing, though.
shader-db2 results:
total instructions in shared programs: 61874 -> 59583 (-3.70%)
instructions in affected programs: 50677 -> 48386 (-4.52%)
Diffstat (limited to 'src/gallium/drivers/vc4/vc4_qir.h')
| -rw-r--r-- | src/gallium/drivers/vc4/vc4_qir.h | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/src/gallium/drivers/vc4/vc4_qir.h b/src/gallium/drivers/vc4/vc4_qir.h index cb02db5272c..0b76a2f246e 100644 --- a/src/gallium/drivers/vc4/vc4_qir.h +++ b/src/gallium/drivers/vc4/vc4_qir.h @@ -148,6 +148,11 @@ struct simple_node { struct simple_node *prev; }; +struct queued_qpu_inst { + struct simple_node link; + uint64_t inst; +}; + struct qinst { struct simple_node link; @@ -368,6 +373,8 @@ bool qir_opt_copy_propagation(struct vc4_compile *c); bool qir_opt_cse(struct vc4_compile *c); bool qir_opt_dead_code(struct vc4_compile *c); +void qpu_schedule_instructions(struct vc4_compile *c); + #define QIR_ALU0(name) \ static inline struct qreg \ qir_##name(struct vc4_compile *c) \ |