From a626565178fb810f439fbdfdaf6cc28a973a6a81 Mon Sep 17 00:00:00 2001 From: =?utf8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 21 Jan 2010 15:51:37 -0800 Subject: [PATCH] vmware/core: Flush preemptively the FIFO commands to keep the GMR working set within reasonable bounds. --- .../winsys/drm/vmware/core/vmw_context.c | 38 ++++++++++++++++++- 1 file changed, 37 insertions(+), 1 deletion(-) diff --git a/src/gallium/winsys/drm/vmware/core/vmw_context.c b/src/gallium/winsys/drm/vmware/core/vmw_context.c index b6997588de4..87a3df923fc 100644 --- a/src/gallium/winsys/drm/vmware/core/vmw_context.c +++ b/src/gallium/winsys/drm/vmware/core/vmw_context.c @@ -73,6 +73,19 @@ struct vmw_svga_winsys_context struct pb_validate *validate; uint32_t last_fence; + + /** + * The amount of GMR that is referred by the commands currently batched + * in the context. + */ + uint32_t seen_regions; + + /** + * Whether this context should fail to reserve more commands, not because it + * ran out of command space, but because a substantial ammount of GMR was + * referred. + */ + boolean preemptive_flush; }; @@ -124,6 +137,8 @@ vmw_swc_flush(struct svga_winsys_context *swc, #ifdef DEBUG vswc->must_flush = FALSE; #endif + vswc->preemptive_flush = FALSE; + vswc->seen_regions = 0; if(pfence) *pfence = fence; @@ -151,7 +166,8 @@ vmw_swc_reserve(struct svga_winsys_context *swc, if(nr_bytes > vswc->command.size) return NULL; - if(vswc->command.used + nr_bytes > vswc->command.size || + if(vswc->preemptive_flush || + vswc->command.used + nr_bytes > vswc->command.size || vswc->surface.used + nr_relocs > vswc->surface.size) { #ifdef DEBUG vswc->must_flush = TRUE; @@ -220,6 +236,26 @@ vmw_swc_region_relocation(struct svga_winsys_context *swc, ret = pb_validate_add_buffer(vswc->validate, buf, flags); /* TODO: Update pipebuffer to reserve buffers and not fail here */ assert(ret == PIPE_OK); + + /* + * Flush preemptively the FIFO commands to keep the GMR working set within + * the GMR pool size. + * + * This is necessary for applications like SPECviewperf that generate huge + * amounts of immediate vertex data, so that we don't pile up too much of + * that vertex data neither in the guest nor in the host. + * + * Note that in the current implementation if a region is referred twice in + * a command stream, it will be accounted twice. We could detect repeated + * regions and count only once, but there is no incentive to do that, since + * regions are typically short-lived; always referred in a single command; + * and at the worst we just flush the commands a bit sooner, which for the + * SVGA virtual device it's not a performance issue since flushing commands + * to the FIFO won't cause flushing in the host. + */ + vswc->seen_regions += buf->base.size; + if(vswc->seen_regions >= VMW_GMR_POOL_SIZE/2) + vswc->preemptive_flush = TRUE; } -- 2.30.2