diff options
| author | Eric Anholt <[email protected]> | 2016-11-30 14:00:28 -0800 |
|---|---|---|
| committer | Eric Anholt <[email protected]> | 2016-11-30 19:58:09 -0800 |
| commit | 51244859e368d9270a1f2f84aec6e97fbc1f5fac (patch) | |
| tree | 973693687a2e72d648f409145a810d64a1c89cfa /src/gallium/drivers/vc4 | |
| parent | 6c34084d8eafff3a764cd1ad49afacd211470f7b (diff) | |
vc4: Avoid false scheduling dependencies for LOAD_IMMs.
Noticed in shaders with branching, where we ended up scheduling delay
slots near the start of a block for the uniforms reset setup.
total instructions in shared programs: 93970 -> 93951 (-0.02%)
instructions in affected programs: 3117 -> 3098 (-0.61%)
3DMMES performance +0.423087% +/- 0.133521% (n=9,10)
Diffstat (limited to 'src/gallium/drivers/vc4')
| -rw-r--r-- | src/gallium/drivers/vc4/vc4_qpu_schedule.c | 5 | ||||
| -rw-r--r-- | src/gallium/drivers/vc4/vc4_qpu_validate.c | 4 |
2 files changed, 9 insertions, 0 deletions
diff --git a/src/gallium/drivers/vc4/vc4_qpu_schedule.c b/src/gallium/drivers/vc4/vc4_qpu_schedule.c index 878cdf95d96..4b2cb9dbd37 100644 --- a/src/gallium/drivers/vc4/vc4_qpu_schedule.c +++ b/src/gallium/drivers/vc4/vc4_qpu_schedule.c @@ -477,6 +477,11 @@ reads_too_soon_after_write(struct choose_scoreboard *scoreboard, uint64_t inst) uint32_t raddr_a = QPU_GET_FIELD(inst, QPU_RADDR_A); uint32_t raddr_b = QPU_GET_FIELD(inst, QPU_RADDR_B); uint32_t sig = QPU_GET_FIELD(inst, QPU_SIG); + + /* Full immediate loads don't read any registers. */ + if (sig == QPU_SIG_LOAD_IMM) + return false; + uint32_t src_muxes[] = { QPU_GET_FIELD(inst, QPU_ADD_A), QPU_GET_FIELD(inst, QPU_ADD_B), diff --git a/src/gallium/drivers/vc4/vc4_qpu_validate.c b/src/gallium/drivers/vc4/vc4_qpu_validate.c index f18ba0364d0..08dd6e5df2e 100644 --- a/src/gallium/drivers/vc4/vc4_qpu_validate.c +++ b/src/gallium/drivers/vc4/vc4_qpu_validate.c @@ -58,6 +58,10 @@ _reads_reg(uint64_t inst, uint32_t r, bool ignore_a, bool ignore_b) if (QPU_GET_FIELD(inst, QPU_SIG) == QPU_SIG_BRANCH) return false; + /* Load immediates don't read any registers. */ + if (QPU_GET_FIELD(inst, QPU_SIG) == QPU_SIG_LOAD_IMM) + return false; + for (int i = 0; i < ARRAY_SIZE(src_regs); i++) { if (!ignore_a && src_regs[i].mux == QPU_MUX_A && |