From: Kenneth Graunke Date: Wed, 11 Dec 2013 22:44:45 +0000 (-0800) Subject: i965: Create a helper function for emitting PIPE_CONTROL flushes. X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=9420b577ddeefe04c9812e9b62628e2f01be06cf;p=mesa.git i965: Create a helper function for emitting PIPE_CONTROL flushes. These days, we need to emit PIPE_CONTROL flushes all over the place. Being able to do that via a single function call seems convenient. Broadwell will also increase the length of these packets by 1; with the refactoring, we should have to do this in substantially fewer places. v2: Add back forgotten intel_emit_post_sync_nonzero_flush (caught by Eric Anholt). Drop unlikely() from BLT_RING check. Signed-off-by: Kenneth Graunke Reviewed-by: Matt Turner Reviewed-by: Eric Anholt --- diff --git a/src/mesa/drivers/dri/i965/brw_queryobj.c b/src/mesa/drivers/dri/i965/brw_queryobj.c index 64c602fb53e..9f839379bff 100644 --- a/src/mesa/drivers/dri/i965/brw_queryobj.c +++ b/src/mesa/drivers/dri/i965/brw_queryobj.c @@ -52,16 +52,9 @@ brw_write_timestamp(struct brw_context *brw, drm_intel_bo *query_bo, int idx) if (brw->gen >= 6) { /* Emit workaround flushes: */ if (brw->gen == 6) { - /* The timestamp write below is a non-zero post-sync op, which on - * Gen6 necessitates a CS stall. CS stalls need stall at scoreboard - * set. See the comments for intel_emit_post_sync_nonzero_flush(). - */ - BEGIN_BATCH(4); - OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2)); - OUT_BATCH(PIPE_CONTROL_CS_STALL | PIPE_CONTROL_STALL_AT_SCOREBOARD); - OUT_BATCH(0); - OUT_BATCH(0); - ADVANCE_BATCH(); + brw_emit_pipe_control_flush(brw, + PIPE_CONTROL_CS_STALL | + PIPE_CONTROL_STALL_AT_SCOREBOARD); } BEGIN_BATCH(5); diff --git a/src/mesa/drivers/dri/i965/gen6_vs_state.c b/src/mesa/drivers/dri/i965/gen6_vs_state.c index 80129cdf168..ac58e9759c2 100644 --- a/src/mesa/drivers/dri/i965/gen6_vs_state.c +++ b/src/mesa/drivers/dri/i965/gen6_vs_state.c @@ -205,15 +205,10 @@ upload_vs_state(struct brw_context *brw) * what is strictly required to avoid the issue. */ intel_emit_post_sync_nonzero_flush(brw); - - BEGIN_BATCH(4); - OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2)); - OUT_BATCH(PIPE_CONTROL_DEPTH_STALL | - PIPE_CONTROL_INSTRUCTION_FLUSH | - PIPE_CONTROL_STATE_CACHE_INVALIDATE); - OUT_BATCH(0); /* address */ - OUT_BATCH(0); /* write data */ - ADVANCE_BATCH(); + brw_emit_pipe_control_flush(brw, + PIPE_CONTROL_DEPTH_STALL | + PIPE_CONTROL_INSTRUCTION_FLUSH | + PIPE_CONTROL_STATE_CACHE_INVALIDATE); } const struct brw_tracked_state gen6_vs_state = { diff --git a/src/mesa/drivers/dri/i965/intel_batchbuffer.c b/src/mesa/drivers/dri/i965/intel_batchbuffer.c index 59d03f68b29..041b18dd34b 100644 --- a/src/mesa/drivers/dri/i965/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/i965/intel_batchbuffer.c @@ -431,6 +431,32 @@ intel_batchbuffer_data(struct brw_context *brw, brw->batch.used += bytes >> 2; } +/** + * Emit a PIPE_CONTROL with various flushing flags. + * + * The caller is responsible for deciding what flags are appropriate for the + * given generation. + */ +void +brw_emit_pipe_control_flush(struct brw_context *brw, uint32_t flags) +{ + if (brw->gen >= 6) { + BEGIN_BATCH(4); + OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2)); + OUT_BATCH(flags); + OUT_BATCH(0); + OUT_BATCH(0); + ADVANCE_BATCH(); + } else { + BEGIN_BATCH(4); + OUT_BATCH(_3DSTATE_PIPE_CONTROL | flags | (4 - 2)); + OUT_BATCH(0); + OUT_BATCH(0); + OUT_BATCH(0); + ADVANCE_BATCH(); + } +} + /** * Restriction [DevSNB, DevIVB]: * @@ -448,26 +474,9 @@ intel_emit_depth_stall_flushes(struct brw_context *brw) { assert(brw->gen >= 6 && brw->gen <= 7); - BEGIN_BATCH(4); - OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2)); - OUT_BATCH(PIPE_CONTROL_DEPTH_STALL); - OUT_BATCH(0); /* address */ - OUT_BATCH(0); /* write data */ - ADVANCE_BATCH() - - BEGIN_BATCH(4); - OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2)); - OUT_BATCH(PIPE_CONTROL_DEPTH_CACHE_FLUSH); - OUT_BATCH(0); /* address */ - OUT_BATCH(0); /* write data */ - ADVANCE_BATCH(); - - BEGIN_BATCH(4); - OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2)); - OUT_BATCH(PIPE_CONTROL_DEPTH_STALL); - OUT_BATCH(0); /* address */ - OUT_BATCH(0); /* write data */ - ADVANCE_BATCH(); + brw_emit_pipe_control_flush(brw, PIPE_CONTROL_DEPTH_STALL); + brw_emit_pipe_control_flush(brw, PIPE_CONTROL_DEPTH_CACHE_FLUSH); + brw_emit_pipe_control_flush(brw, PIPE_CONTROL_DEPTH_STALL); } /** @@ -565,13 +574,9 @@ intel_emit_post_sync_nonzero_flush(struct brw_context *brw) if (!brw->batch.need_workaround_flush) return; - BEGIN_BATCH(4); - OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2)); - OUT_BATCH(PIPE_CONTROL_CS_STALL | - PIPE_CONTROL_STALL_AT_SCOREBOARD); - OUT_BATCH(0); /* address */ - OUT_BATCH(0); /* write data */ - ADVANCE_BATCH(); + brw_emit_pipe_control_flush(brw, + PIPE_CONTROL_CS_STALL | + PIPE_CONTROL_STALL_AT_SCOREBOARD); BEGIN_BATCH(4); OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2)); @@ -593,46 +598,32 @@ intel_emit_post_sync_nonzero_flush(struct brw_context *brw) void intel_batchbuffer_emit_mi_flush(struct brw_context *brw) { - if (brw->gen >= 6) { - if (brw->batch.ring == BLT_RING) { - BEGIN_BATCH_BLT(4); - OUT_BATCH(MI_FLUSH_DW); - OUT_BATCH(0); - OUT_BATCH(0); - OUT_BATCH(0); - ADVANCE_BATCH(); - } else { - if (brw->gen == 6) { - /* Hardware workaround: SNB B-Spec says: - * - * [Dev-SNB{W/A}]: Before a PIPE_CONTROL with Write Cache - * Flush Enable =1, a PIPE_CONTROL with any non-zero - * post-sync-op is required. - */ - intel_emit_post_sync_nonzero_flush(brw); - } - - BEGIN_BATCH(4); - OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2)); - OUT_BATCH(PIPE_CONTROL_INSTRUCTION_FLUSH | - PIPE_CONTROL_WRITE_FLUSH | - PIPE_CONTROL_DEPTH_CACHE_FLUSH | - PIPE_CONTROL_VF_CACHE_INVALIDATE | - PIPE_CONTROL_TC_FLUSH | - PIPE_CONTROL_NO_WRITE | - PIPE_CONTROL_CS_STALL); - OUT_BATCH(0); /* write address */ - OUT_BATCH(0); /* write data */ - ADVANCE_BATCH(); - } - } else { - BEGIN_BATCH(4); - OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2) | - PIPE_CONTROL_WRITE_FLUSH | - PIPE_CONTROL_NO_WRITE); - OUT_BATCH(0); /* write address */ - OUT_BATCH(0); /* write data */ - OUT_BATCH(0); /* write data */ + if (brw->batch.ring == BLT_RING && brw->gen >= 6) { + BEGIN_BATCH_BLT(4); + OUT_BATCH(MI_FLUSH_DW); + OUT_BATCH(0); + OUT_BATCH(0); + OUT_BATCH(0); ADVANCE_BATCH(); + } else { + int flags = PIPE_CONTROL_NO_WRITE | PIPE_CONTROL_WRITE_FLUSH; + if (brw->gen >= 6) { + flags |= PIPE_CONTROL_INSTRUCTION_FLUSH | + PIPE_CONTROL_DEPTH_CACHE_FLUSH | + PIPE_CONTROL_VF_CACHE_INVALIDATE | + PIPE_CONTROL_TC_FLUSH | + PIPE_CONTROL_CS_STALL; + + if (brw->gen == 6) { + /* Hardware workaround: SNB B-Spec says: + * + * [Dev-SNB{W/A}]: Before a PIPE_CONTROL with Write Cache + * Flush Enable =1, a PIPE_CONTROL with any non-zero + * post-sync-op is required. + */ + intel_emit_post_sync_nonzero_flush(brw); + } + } + brw_emit_pipe_control_flush(brw, flags); } } diff --git a/src/mesa/drivers/dri/i965/intel_batchbuffer.h b/src/mesa/drivers/dri/i965/intel_batchbuffer.h index 2b2693f8ee2..5ed2089049b 100644 --- a/src/mesa/drivers/dri/i965/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/i965/intel_batchbuffer.h @@ -64,6 +64,7 @@ bool intel_batchbuffer_emit_reloc64(struct brw_context *brw, uint32_t read_domains, uint32_t write_domain, uint32_t offset); +void brw_emit_pipe_control_flush(struct brw_context *brw, uint32_t flags); void intel_batchbuffer_emit_mi_flush(struct brw_context *brw); void intel_emit_post_sync_nonzero_flush(struct brw_context *brw); void intel_emit_depth_stall_flushes(struct brw_context *brw);