Mesa (master): nvc0: properly align NVE4_COMPUTE_MP_TEMP_SIZE

[Christoph Bumiller]

Module: Mesa
Branch: master
Commit: ef6d5ee9f34bce89c8bb8ff001be4c70a2a5421d
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=ef6d5ee9f34bce89c8bb8ff001be4c70a2a5421d

Author: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Date:   Thu Jul 25 10:35:33 2013 +0200

nvc0: properly align NVE4_COMPUTE_MP_TEMP_SIZE

MP_TEMP_SIZE must be aligned to 0x8000, while TEMP_SIZE on NVE4_3D
must be aligned to 0x20000, so perform both alignments to be sure
we allocate enough space (actually the bo will most likely use 128
KiB pages and not aligning to that would be a waste anyway).

Cc: "9.2" mesa-stable at lists.freedesktop.org

---

 src/gallium/drivers/nvc0/nvc0_screen.c  |    1 +
 src/gallium/drivers/nvc0/nve4_compute.c |    4 ++--
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/src/gallium/drivers/nvc0/nvc0_screen.c b/src/gallium/drivers/nvc0/nvc0_screen.c
index 201d12f..171a302 100644
--- a/src/gallium/drivers/nvc0/nvc0_screen.c
+++ b/src/gallium/drivers/nvc0/nvc0_screen.c
@@ -512,6 +512,7 @@ nvc0_screen_resize_tls_area(struct nvc0_screen *screen,
    }
 
    size *= (screen->base.device->chipset >= 0xe0) ? 64 : 48; /* max warps */
+   size  = align(size, 0x8000);
    size *= screen->mp_count;
 
    size = align(size, 1 << 17);
diff --git a/src/gallium/drivers/nvc0/nve4_compute.c b/src/gallium/drivers/nvc0/nve4_compute.c
index 258564b..abadd60 100644
--- a/src/gallium/drivers/nvc0/nve4_compute.c
+++ b/src/gallium/drivers/nvc0/nve4_compute.c
@@ -78,11 +78,11 @@ nve4_screen_compute_setup(struct nvc0_screen *screen,
     */
    BEGIN_NVC0(push, NVE4_COMPUTE(MP_TEMP_SIZE_HIGH(0)), 3);
    PUSH_DATAh(push, screen->tls->size / screen->mp_count);
-   PUSH_DATA (push, screen->tls->size / screen->mp_count);
+   PUSH_DATA (push, (screen->tls->size / screen->mp_count) & ~0x7fff);
    PUSH_DATA (push, 0xff);
    BEGIN_NVC0(push, NVE4_COMPUTE(MP_TEMP_SIZE_HIGH(1)), 3);
    PUSH_DATAh(push, screen->tls->size / screen->mp_count);
-   PUSH_DATA (push, screen->tls->size / screen->mp_count);
+   PUSH_DATA (push, (screen->tls->size / screen->mp_count) & ~0x7fff);
    PUSH_DATA (push, 0xff);
 
    /* Unified address space ? Who needs that ? Certainly not OpenCL.