[Mesa-dev] [PATCH 4/5] radeonsi: fix register count calculation

[Christian König]

On 31.07.2012 15:32, Michel Dänzer wrote:
> On Mit, 2012-07-25 at 19:58 +0200, Christian König wrote:
>> Signed-off-by: Christian König <deathsimple at vodafone.de>
>> ---
>>   src/gallium/drivers/radeonsi/si_state_draw.c |    8 ++++----
>>   1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
>> index 0d9f009..74ed01f 100644
>> --- a/src/gallium/drivers/radeonsi/si_state_draw.c
>> +++ b/src/gallium/drivers/radeonsi/si_state_draw.c
>> @@ -86,8 +86,8 @@ static void si_pipe_shader_vs(struct pipe_context *ctx, struct si_pipe_shader *s
>>   	assert(num_sgprs <= 104);
>>   
>>   	si_pm4_set_reg(pm4, R_00B128_SPI_SHADER_PGM_RSRC1_VS,
>> -		       S_00B128_VGPRS((shader->num_vgprs - 1) / 4) |
>> -		       S_00B128_SGPRS((num_sgprs - 1) / 8));
>> +		       S_00B128_VGPRS((align(shader->num_vgprs,4) / 4) - 1) |
>> +		       S_00B128_SGPRS((align(num_sgprs, 8) / 8) - 1));
>>   	si_pm4_set_reg(pm4, R_00B12C_SPI_SHADER_PGM_RSRC2_VS,
>>   		       S_00B12C_USER_SGPR(num_user_sgprs));
>>   
>> @@ -192,8 +192,8 @@ static void si_pipe_shader_ps(struct pipe_context *ctx, struct si_pipe_shader *s
>>   	assert(num_sgprs <= 104);
>>   
>>   	si_pm4_set_reg(pm4, R_00B028_SPI_SHADER_PGM_RSRC1_PS,
>> -		       S_00B028_VGPRS((shader->num_vgprs - 1) / 4) |
>> -		       S_00B028_SGPRS((num_sgprs - 1) / 8));
>> +		       S_00B028_VGPRS((align(shader->num_vgprs, 4) / 4) - 1) |
>> +		       S_00B028_SGPRS((align(num_sgprs, 8) / 8) - 1));
>>   	si_pm4_set_reg(pm4, R_00B02C_SPI_SHADER_PGM_RSRC2_PS,
>>   		       S_00B02C_USER_SGPR(num_user_sgprs));
>>   
> The previous code was correct according to my reading of the register
> spec: 0 means 4/8 V/SGPRs, 63/15 means 256/128.

Came to the same conclusion after sleeping a night over it. So I removed 
the patch from my patchset.

Sorry for the noise,
Christian.