From 622afc8dbdca3ac58137d888184f22371cb58c68 Mon Sep 17 00:00:00 2001 From: "Kristian H. Kristensen" Date: Wed, 16 Oct 2019 12:08:19 -0700 Subject: [PATCH] freedreno/a6xx: Implement PIPE_QUERY_PRIMITIVES_GENERATED for GS When we don't have streamout enabled, we have to read this register to get the number of primitives emitted. Signed-off-by: Kristian H. Kristensen --- src/freedreno/registers/a6xx.xml | 32 +++++++++++++++++ src/freedreno/registers/adreno_pm4.xml | 2 ++ .../drivers/freedreno/a6xx/fd6_query.c | 34 +++++++++++++++++++ 3 files changed, 68 insertions(+) diff --git a/src/freedreno/registers/a6xx.xml b/src/freedreno/registers/a6xx.xml index 4d83a5dd2cf..b7cfecdc121 100644 --- a/src/freedreno/registers/a6xx.xml +++ b/src/freedreno/registers/a6xx.xml @@ -1372,6 +1372,38 @@ to upconvert to 32b float internally? + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/src/freedreno/registers/adreno_pm4.xml b/src/freedreno/registers/adreno_pm4.xml index 23705a7cad9..f94e8e4e10b 100644 --- a/src/freedreno/registers/adreno_pm4.xml +++ b/src/freedreno/registers/adreno_pm4.xml @@ -16,6 +16,8 @@ xsi:schemaLocation="http://nouveau.freedesktop.org/ rules-ng.xsd"> + + diff --git a/src/gallium/drivers/freedreno/a6xx/fd6_query.c b/src/gallium/drivers/freedreno/a6xx/fd6_query.c index d950144bc7e..db249fc6ef7 100644 --- a/src/gallium/drivers/freedreno/a6xx/fd6_query.c +++ b/src/gallium/drivers/freedreno/a6xx/fd6_query.c @@ -256,6 +256,8 @@ struct PACKED fd6_primitives_sample { struct { uint64_t generated, emitted; } start[4], stop[4], result; + + uint64_t prim_start[16], prim_stop[16], prim_emitted; }; @@ -268,6 +270,17 @@ static void primitive_counts_resume(struct fd_acc_query *aq, struct fd_batch *batch) { struct fd_ringbuffer *ring = batch->draw; + const unsigned count = 1; + + fd_wfi(batch, ring); + + OUT_PKT7(ring, CP_REG_TO_MEM, 3); + OUT_RING(ring, CP_REG_TO_MEM_0_64B | + CP_REG_TO_MEM_0_CNT(count - 1) | + CP_REG_TO_MEM_0_REG(REG_A6XX_RBBM_PRIMCTR_8_LO)); + primitives_relocw(ring, aq, prim_start); + + fd6_event_write(batch, ring, START_PRIMITIVE_CTRS, false); OUT_PKT4(ring, REG_A6XX_VPC_SO_STREAM_COUNTS_LO, 2); primitives_relocw(ring, aq, start[0]); @@ -279,6 +292,25 @@ static void primitive_counts_pause(struct fd_acc_query *aq, struct fd_batch *batch) { struct fd_ringbuffer *ring = batch->draw; + const unsigned count = 1; + + fd_wfi(batch, ring); + + /* snapshot the end values: */ + OUT_PKT7(ring, CP_REG_TO_MEM, 3); + OUT_RING(ring, CP_REG_TO_MEM_0_64B | + CP_REG_TO_MEM_0_CNT(count - 1) | + CP_REG_TO_MEM_0_REG(REG_A6XX_RBBM_PRIMCTR_8_LO)); + primitives_relocw(ring, aq, prim_stop); + + /* result += stop - start: */ + OUT_PKT7(ring, CP_MEM_TO_MEM, 9); + OUT_RING(ring, CP_MEM_TO_MEM_0_DOUBLE | + CP_MEM_TO_MEM_0_NEG_C | 0x40000000); + primitives_relocw(ring, aq, result.generated); + primitives_reloc(ring, aq, prim_emitted); + primitives_reloc(ring, aq, prim_stop); + primitives_reloc(ring, aq, prim_start); OUT_PKT4(ring, REG_A6XX_VPC_SO_STREAM_COUNTS_LO, 2); primitives_relocw(ring, aq, stop[0]); @@ -304,6 +336,8 @@ primitive_counts_pause(struct fd_acc_query *aq, struct fd_batch *batch) primitives_reloc(ring, aq, result.generated); primitives_reloc(ring, aq, stop[aq->base.index].generated); primitives_reloc(ring, aq, start[aq->base.index].generated); + + fd6_event_write(batch, ring, STOP_PRIMITIVE_CTRS, false); } static void -- 2.30.2