[Mesa-dev] [PATCH v2 1/4] i965/gen10: Implement WaSampleOffsetIZ workaround
Anuj Phogat
anuj.phogat at gmail.com
Fri Nov 3 01:07:11 UTC 2017
On Thu, Nov 2, 2017 at 10:59 AM, Nanley Chery <nanleychery at gmail.com> wrote:
> On Wed, Nov 01, 2017 at 03:48:25PM -0700, Anuj Phogat wrote:
>> There are few other (duplicate) workarounds which have similar recommendations:
>> WaFlushHangWhenNonPipelineStateAndMarkerStalled
>> WaCSStallBefore3DSamplePattern
>> WaPipeControlBefore3DStateSamplePattern
>>
>> WaPipeControlBefore3DStateSamplePattern has some extra recommendations if
>> driver is using mid batch context restore. Ignoring it for now because We're
>> not doing mid-batch context restore in Mesa.
>>
>> This workaround doesn't fix any of the piglit hangs we've seen
>> on CNL. But it might be fixing something we haven't tested yet.
>>
>> V2: Use brw_load_register_imm32() to program CACHE_MODE_0.
>> Get rid of brw_flush_gpu_caches().
>>
>> Cc: Nanley Chery <nanley.g.chery at intel.com>
>> Signed-off-by: Anuj Phogat <anuj.phogat at gmail.com>
>> Reviewed-by: Rafael Antognolli <rafael.antognolli at intel.com>
>> ---
>> src/mesa/drivers/dri/i965/brw_context.h | 2 ++
>> src/mesa/drivers/dri/i965/brw_defines.h | 1 +
>> src/mesa/drivers/dri/i965/brw_pipe_control.c | 41 ++++++++++++++++++++++
>> src/mesa/drivers/dri/i965/gen8_multisample_state.c | 8 +++++
>> 4 files changed, 52 insertions(+)
>>
>> diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
>> index 0102f15424..1030b2b313 100644
>> --- a/src/mesa/drivers/dri/i965/brw_context.h
>> +++ b/src/mesa/drivers/dri/i965/brw_context.h
>> @@ -1656,6 +1656,8 @@ void brw_emit_post_sync_nonzero_flush(struct brw_context *brw);
>> void brw_emit_depth_stall_flushes(struct brw_context *brw);
>> void gen7_emit_vs_workaround_flush(struct brw_context *brw);
>> void gen7_emit_cs_stall_flush(struct brw_context *brw);
>> +void gen10_emit_wa_cs_stall_flush(struct brw_context *brw);
>> +void gen10_emit_wa_lri_to_cache_mode_zero(struct brw_context *brw);
>
> These functions are only used in one file. What do you think about
> making them static?
>
Agreed. Fixed in v3.
>>
>> /* brw_queryformat.c */
>> void brw_query_internal_format(struct gl_context *ctx, GLenum target,
>> diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
>> index 4abb790612..270cdf29db 100644
>> --- a/src/mesa/drivers/dri/i965/brw_defines.h
>> +++ b/src/mesa/drivers/dri/i965/brw_defines.h
>> @@ -1609,6 +1609,7 @@ enum brw_pixel_shader_coverage_mask_mode {
>> #define GEN7_GPGPU_DISPATCHDIMY 0x2504
>> #define GEN7_GPGPU_DISPATCHDIMZ 0x2508
>>
>> +#define GEN7_CACHE_MODE_0 0x7000
>> #define GEN7_CACHE_MODE_1 0x7004
>> # define GEN9_FLOAT_BLEND_OPTIMIZATION_ENABLE (1 << 4)
>> # define GEN8_HIZ_NP_PMA_FIX_ENABLE (1 << 11)
>> diff --git a/src/mesa/drivers/dri/i965/brw_pipe_control.c b/src/mesa/drivers/dri/i965/brw_pipe_control.c
>> index 460b8f73b6..6ebe1443d5 100644
>> --- a/src/mesa/drivers/dri/i965/brw_pipe_control.c
>> +++ b/src/mesa/drivers/dri/i965/brw_pipe_control.c
>> @@ -279,6 +279,47 @@ gen7_emit_cs_stall_flush(struct brw_context *brw)
>> }
>>
>> /**
>> + * From Gen10 Workarounds page in h/w specs:
>> + * WaSampleOffsetIZ:
>> + * Prior to the 3DSTATE_SAMPLE_PATTERN driver must ensure there are no
>> + * markers in the pipeline by programming a PIPE_CONTROL with stall.
>> + */
>> +void
>> +gen10_emit_wa_cs_stall_flush(struct brw_context *brw)
>> +{
>> + const struct gen_device_info *devinfo = &brw->screen->devinfo;
>
> While build-testing this with a release build, I expected the compiler
> to emit an unused variable warning for the devinfo variables, but for
> some reason it did not.
>
> With or without those minor points addressed, this patch is
> Reviewed-by: Nanley Chery <nanley.g.chery at intel.com>
>
>> + assert(devinfo->gen == 10);
>> + brw_emit_pipe_control_flush(brw,
>> + PIPE_CONTROL_CS_STALL |
>> + PIPE_CONTROL_STALL_AT_SCOREBOARD);
>> +}
>> +
>> +/**
>> + * From Gen10 Workarounds page in h/w specs:
>> + * WaSampleOffsetIZ:
>> + * When 3DSTATE_SAMPLE_PATTERN is programmed, driver must then issue an
>> + * MI_LOAD_REGISTER_IMM command to an offset between 0x7000 and 0x7FFF(SVL)
>> + * after the command to ensure the state has been delivered prior to any
>> + * command causing a marker in the pipeline.
>> + */
>> +void
>> +gen10_emit_wa_lri_to_cache_mode_zero(struct brw_context *brw)
>> +{
>> + const struct gen_device_info *devinfo = &brw->screen->devinfo;
>> + assert(devinfo->gen == 10);
>> +
>> + /* Before changing the value of CACHE_MODE_0 register, GFX pipeline must
>> + * be idle; i.e., full flush is required.
>> + */
>> + brw_emit_pipe_control_flush(brw,
>> + PIPE_CONTROL_CACHE_FLUSH_BITS |
>> + PIPE_CONTROL_CACHE_INVALIDATE_BITS);
>> +
>> + /* Write to CACHE_MODE_0 (0x7000) */
>> + brw_load_register_imm32(brw, GEN7_CACHE_MODE_0, 0);
>> +}
>> +
>> +/**
>> * Emits a PIPE_CONTROL with a non-zero post-sync operation, for
>> * implementing two workarounds on gen6. From section 1.4.7.1
>> * "PIPE_CONTROL" of the Sandy Bridge PRM volume 2 part 1:
>> diff --git a/src/mesa/drivers/dri/i965/gen8_multisample_state.c b/src/mesa/drivers/dri/i965/gen8_multisample_state.c
>> index 3afa586275..5235fc2cf9 100644
>> --- a/src/mesa/drivers/dri/i965/gen8_multisample_state.c
>> +++ b/src/mesa/drivers/dri/i965/gen8_multisample_state.c
>> @@ -33,6 +33,11 @@
>> void
>> gen8_emit_3dstate_sample_pattern(struct brw_context *brw)
>> {
>> + const struct gen_device_info *devinfo = &brw->screen->devinfo;
>> +
>> + if (devinfo->gen == 10)
>> + gen10_emit_wa_cs_stall_flush(brw);
>> +
>> BEGIN_BATCH(9);
>> OUT_BATCH(_3DSTATE_SAMPLE_PATTERN << 16 | (9 - 2));
>>
>> @@ -52,4 +57,7 @@ gen8_emit_3dstate_sample_pattern(struct brw_context *brw)
>> /* 1x and 2x MSAA */
>> OUT_BATCH(brw_multisample_positions_1x_2x);
>> ADVANCE_BATCH();
>> +
>> + if (devinfo->gen == 10)
>> + gen10_emit_wa_lri_to_cache_mode_zero(brw);
>> }
>> --
>> 2.13.5
>>
>> _______________________________________________
>> mesa-dev mailing list
>> mesa-dev at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
More information about the mesa-dev
mailing list