fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-23 15:30:14 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

freedreno/a5xx: Set num_sp_cores and set PC/VFD_POWER_CNTL accordingly.

Browse source 
Based on libwrap tracing of the blob.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24358>
This commit is contained in:
Emma Anholt 2023-08-08 09:53:16 -07:00 • committed by Marge Bot
parent c9f9d71412
commit c4874b4cee
2 changed files with 35 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

30
src/freedreno/common/freedreno_devices.py
View file

@ -232,8 +232,36 @@ add_gpus([
GPUId(506), GPUId(506),
GPUId(508), GPUId(508),
GPUId(509), GPUId(509),
], GPUInfo(
CHIP.A5XX,
gmem_align_w = 64, gmem_align_h = 32,
tile_align_w = 64, tile_align_h = 32,
tile_max_w = 1024, # max_bitfield_val(7, 0, 5)
tile_max_h = max_bitfield_val(16, 9, 5),
num_vsc_pipes = 16,
cs_shared_mem_size = 32 * 1024,
num_sp_cores = 1,
wave_granularity = 2,
fibers_per_sp = 64 * 16, # Lowest number that didn't fault on spillall fs-varying-array-mat4-col-row-rd.
))
add_gpus([
GPUId(510), GPUId(510),
GPUId(512), GPUId(512),
], GPUInfo(
CHIP.A5XX,
gmem_align_w = 64, gmem_align_h = 32,
tile_align_w = 64, tile_align_h = 32,
tile_max_w = 1024, # max_bitfield_val(7, 0, 5)
tile_max_h = max_bitfield_val(16, 9, 5),
num_vsc_pipes = 16,
cs_shared_mem_size = 32 * 1024,
num_sp_cores = 2,
wave_granularity = 2,
fibers_per_sp = 64 * 16, # Lowest number that didn't fault on spillall fs-varying-array-mat4-col-row-rd.
))
add_gpus([
GPUId(530), GPUId(530),
GPUId(540), GPUId(540),
], GPUInfo( ], GPUInfo(
@ -244,7 +272,7 @@ add_gpus([
tile_max_h = max_bitfield_val(16, 9, 5), tile_max_h = max_bitfield_val(16, 9, 5),
num_vsc_pipes = 16, num_vsc_pipes = 16,
cs_shared_mem_size = 32 * 1024, cs_shared_mem_size = 32 * 1024,
num_sp_cores = 0, # TODO num_sp_cores = 4,
wave_granularity = 2, wave_granularity = 2,
fibers_per_sp = 0, # TODO fibers_per_sp = 0, # TODO
)) ))

10
src/gallium/drivers/freedreno/a5xx/fd5_gmem.c
View file

@ -394,6 +394,7 @@ emit_binning_pass(struct fd_batch *batch) assert_dt
static void static void
fd5_emit_tile_init(struct fd_batch *batch) assert_dt fd5_emit_tile_init(struct fd_batch *batch) assert_dt
{ {
struct fd_context *ctx = batch->ctx;
struct fd_ringbuffer *ring = batch->gmem; struct fd_ringbuffer *ring = batch->gmem;
struct pipe_framebuffer_state *pfb = &batch->framebuffer; struct pipe_framebuffer_state *pfb = &batch->framebuffer;
@ -411,10 +412,10 @@ fd5_emit_tile_init(struct fd_batch *batch) assert_dt
OUT_RING(ring, 0x0); OUT_RING(ring, 0x0);
OUT_PKT4(ring, REG_A5XX_PC_POWER_CNTL, 1); OUT_PKT4(ring, REG_A5XX_PC_POWER_CNTL, 1);
OUT_RING(ring, 0x00000003); /* PC_POWER_CNTL */ OUT_RING(ring, ctx->screen->info->num_sp_cores - 1); /* PC_POWER_CNTL */
OUT_PKT4(ring, REG_A5XX_VFD_POWER_CNTL, 1); OUT_PKT4(ring, REG_A5XX_VFD_POWER_CNTL, 1);
OUT_RING(ring, 0x00000003); /* VFD_POWER_CNTL */ OUT_RING(ring, ctx->screen->info->num_sp_cores - 1); /* VFD_POWER_CNTL */
/* 0x10000000 for BYPASS.. 0x7c13c080 for GMEM: */ /* 0x10000000 for BYPASS.. 0x7c13c080 for GMEM: */
fd_wfi(batch, ring); fd_wfi(batch, ring);
@ -715,6 +716,7 @@ fd5_emit_tile_fini(struct fd_batch *batch) assert_dt
static void static void
fd5_emit_sysmem_prep(struct fd_batch *batch) assert_dt fd5_emit_sysmem_prep(struct fd_batch *batch) assert_dt
{ {
struct fd_context *ctx = batch->ctx;
struct fd_ringbuffer *ring = batch->gmem; struct fd_ringbuffer *ring = batch->gmem;
fd5_emit_restore(batch, ring); fd5_emit_restore(batch, ring);
@ -730,10 +732,10 @@ fd5_emit_sysmem_prep(struct fd_batch *batch) assert_dt
fd5_event_write(batch, ring, PC_CCU_INVALIDATE_COLOR, false); fd5_event_write(batch, ring, PC_CCU_INVALIDATE_COLOR, false);
OUT_PKT4(ring, REG_A5XX_PC_POWER_CNTL, 1); OUT_PKT4(ring, REG_A5XX_PC_POWER_CNTL, 1);
OUT_RING(ring, 0x00000003); /* PC_POWER_CNTL */ OUT_RING(ring, ctx->screen->info->num_sp_cores - 1); /* PC_POWER_CNTL */
OUT_PKT4(ring, REG_A5XX_VFD_POWER_CNTL, 1); OUT_PKT4(ring, REG_A5XX_VFD_POWER_CNTL, 1);
OUT_RING(ring, 0x00000003); /* VFD_POWER_CNTL */ OUT_RING(ring, ctx->screen->info->num_sp_cores - 1); /* VFD_POWER_CNTL */
/* 0x10000000 for BYPASS.. 0x7c13c080 for GMEM: */ /* 0x10000000 for BYPASS.. 0x7c13c080 for GMEM: */
fd_wfi(batch, ring); fd_wfi(batch, ring);