Forcewake binary search & code shrink
Subsystems (2)
Tvrtko Ursulin
maintainercover Forcewake binary search & code shrink
From: Tvrtko Ursulin <[email protected]>
A bunch of cleaned up patches from one of my old branches. Just an idea I had been toying about and probably more a RFC at this point. But it amazingly passed trybot on (almost) the first try so I got tempted to post it.
Basically main motivation was to replace linear search comparison ladders in all mmio accessors with something smarter - a binary search in this case.
Main enabler is moving the comparison ladders from the code into the data tables. Then gradually, patch by patch, I was able to reduce the code and end up with a total saving of approximately 14KiB for the series.
That is kind of nice, but whether the churn is worth it I leave open for dicussion. It is slightly less code so perhaps more maintainable. Although tastes will probably differ.
And I haven't been able to do any benchmarking to see if any performance differece is at all detectable.
Tvrtko Ursulin (13):
drm/i915: Remove redundant hsw_write* mmio functions drm/i915: Keep track of active forcewake domains in a bitmask drm/i915: Do not inline forcewake taking in mmio accessors drm/i915: Data driven register to forcewake domains lookup drm/i915: Sort forcewake mapping tables
drm/i915: Use binary search when looking up forcewake domains drm/i915: Eliminate Gen9 special case
drm/i915: Store the active forcewake range table pointer drm/i915: Remove identical macros
drm/i915: Remove identical mmio read functions drm/i915: Remove identical write mmmio functions drm/i915: Sort the shadow register table
drm/i915: Use binary search when looking for shadowed registers
drivers/gpu/drm/i915/i915_drv.h | 13 + drivers/gpu/drm/i915/intel_uncore.c | 525 ++++++++++++++++-------------------- 2 files changed, 247 insertions(+), 291 deletions(-)
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 1/13 drm/i915: Remove redundant hsw_write* mmio functions
From: Tvrtko Ursulin <[email protected]>
They are completely identical to gen6_write* ones.
--- drivers/gpu/drm/i915/intel_uncore.c | 25 +------------------------ 1 file changed, 1 insertion(+), 24 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +1 -24
@@ -1055,21 +1055,6 @@ gen6_write##x(struct drm_i915_private *dev_priv, i915_reg_t reg, u##x val, boolGEN6_WRITE_FOOTER ; \ } -#define __hsw_write (x) \ -static void \ -hsw_write##x(struct drm_i915_private *dev_priv, i915_reg_t reg, u##x val, bool trace) { \ - u32 __fifo_ret = 0; \ - GEN6_WRITE_HEADER; \ - if (NEEDS_FORCE_WAKE(offset)) { \ - __fifo_ret = __gen6_gt_wait_for_fifo(dev_priv); \ - } \ - __raw_i915_write##x(dev_priv, reg, val); \ - if (unlikely(__fifo_ret)) { \ - gen6_gt_check_fifodbg(dev_priv); \ - } \ - GEN6_WRITE_FOOTER; \ -} -#define __gen8_write (x) \ static void \ gen8_write##x(struct drm_i915_private *dev_priv, i915_reg_t reg, u##x val, bool trace) { \
@@ -1116,9 +1101,6 @@ __chv_write(32) __gen8_write(8) __gen8_write(16) __gen8_write(32) -__hsw_write(8) -__hsw_write(16) -__hsw_write(32) __gen6_write(8) __gen6_write(16)__gen6_write (32 )
@@ -1126,7 +1108,6 @@ __gen6_write(32)#undef __gen9_write#undef __chv_write#undef __gen8_write -#undef __hsw_write#undef __gen6_write#undef GEN6_WRITE_FOOTER#undef GEN6_WRITE_HEADER
@@ -1343,11 +1324,7 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)break ;case 7 :case 6 : -if (IS_HASWELL (dev_priv)) { -ASSIGN_WRITE_MMIO_VFUNCS (hsw); - }else { -ASSIGN_WRITE_MMIO_VFUNCS (gen6); - } +ASSIGN_WRITE_MMIO_VFUNCS (gen6);if (IS_VALLEYVIEW (dev_priv)) {ASSIGN_READ_MMIO_VFUNCS (vlv);
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 2/13 drm/i915: Keep track of active forcewake domains in a bitmask
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
There are current places in the code, and there will be more in the future, which iterate the forcewake domains to find out which ones are currently active.
To save them from doing this iteration, we can cheaply keep a mask of active domains in dev_priv->uncore.fw_domains_active.
This has no cost in terms of object size, even manages to shrink it overall by 368 bytes on my config.
--- drivers/gpu/drm/i915/i915_drv.h | 2 ++ drivers/gpu/drm/i915/intel_uncore.c | 54 ++++++++++++++++--------------------- 2 files changed, 25 insertions(+), 31 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +2 -0
@@ -588,7 +588,9 @@ struct intel_uncore {struct intel_uncore_funcs funcs;unsigned fifo_count; +enum forcewake_domains fw_domains; +enum forcewake_domains fw_domains_active;struct intel_uncore_forcewake_domain {struct drm_i915_private *i915;
drivers/gpu/drm/i915/intel_uncore.c +23 -31
116 lines not on this page · raw message
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 3/13 drm/i915: Do not inline forcewake taking in mmio accessors
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
Once we know we need to take new forcewakes, that being a slow operation, it does not make sense to inline that code into every mmio accessor.
Move it to a separate function and save some code.
v2: Be explicit with noinline and remove stale comment.
(Chris Wilson)
--- drivers/gpu/drm/i915/intel_uncore.c | 24 ++++++++++++++---------- 1 file changed, 14 insertions(+), 10 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +14 -10
@@ -862,11 +862,21 @@ __gen2_read(64)trace_i915_reg_rw (false, reg, val,sizeof (val), trace); \ return val -static inline void __force_wake_auto (struct drm_i915_private *dev_priv, -enum forcewake_domains fw_domains) +static noinline void ___force_wake_auto(struct drm_i915_private *dev_priv, +enum forcewake_domains fw_domains) {struct intel_uncore_forcewake_domain *domain; +for_each_fw_domain_masked (domain, fw_domains, dev_priv) +fw_domain_arm_timer (domain); + + dev_priv->uncore.funcs.force_wake_get (dev_priv, fw_domains); + dev_priv->uncore.fw_domains_active |= fw_domains; +} + +static inline void __force_wake_auto (struct drm_i915_private *dev_priv, +enum forcewake_domains fw_domains) +{if (WARN_ON (!fw_domains))return ;
@@ -874,14 +884,8 @@ static inline void __force_wake_auto(struct drm_i915_private *dev_priv, fw_domains &= dev_priv->uncore.fw_domains; fw_domains &= ~dev_priv->uncore.fw_domains_active; -if (fw_domains) { -/* Ideally GCC would be constant-fold and eliminate this loop */ -for_each_fw_domain_masked (domain, fw_domains, dev_priv) -fw_domain_arm_timer (domain); - - dev_priv->uncore.funcs.force_wake_get (dev_priv, fw_domains); - dev_priv->uncore.fw_domains_active |= fw_domains; - } +if (fw_domains) +___force_wake_auto (dev_priv, fw_domains); }#define __gen6_read (x) \
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 4/13 drm/i915: Data driven register to forcewake domains lookup
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
Move finding the correct forcewake domains to take for register access from code to a mapping table. This will allow more interesting work in the following patches and is easier to review if singled out early.
--- drivers/gpu/drm/i915/intel_uncore.c | 212 ++++++++++++++++++------------------ 1 file changed, 103 insertions(+), 109 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +103 -109
@@ -581,28 +581,52 @@ void assert_forcewakes_inactive(struct drm_i915_private *dev_priv) __fwd; \ }) -#define REG_RANGE (reg, start, end) ((reg) >= (start) && (reg) < (end)) +struct intel_forcewake_range +{ +u32 start; +u32 end; + +enum forcewake_domains domains; +}; -#define FORCEWAKE_VLV_RENDER_RANGE_OFFSET (reg) \ - (REG_RANGE((reg), 0x2000, 0x4000) || \ - REG_RANGE((reg), 0x5000, 0x8000) || \ - REG_RANGE((reg), 0xB000, 0x12000) || \ - REG_RANGE((reg), 0x2E000, 0x30000)) +static enum forcewake_domains +find_fw_domain (u32 offset,const struct intel_forcewake_range *ranges, +unsigned int num_ranges) +{ +unsigned int i; +struct intel_forcewake_range *entry = + (struct intel_forcewake_range *)ranges; + +for (i =0 ; i < num_ranges; i++, entry++) { +if (offset >= entry->start && offset <= entry->end) +return entry->domains; + } + +return -1 ; +} -#define FORCEWAKE_VLV_MEDIA_RANGE_OFFSET (reg) \ - (REG_RANGE((reg), 0x12000, 0x14000) || \ - REG_RANGE((reg), 0x22000, 0x24000) || \ - REG_RANGE((reg), 0x30000, 0x40000)) +#define GEN_FW_RANGE (s, e, d) \ + { .start = (s), .end = (e), .domains = (d) } + +static const struct intel_forcewake_range __vlv_fw_ranges[] = { +GEN_FW_RANGE (0x2000 ,0x3fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x5000 ,0x7fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xb000 ,0x11fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x2e000 ,0x2ffff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x22000 ,0x23fff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x30000 ,0x3ffff ,FORCEWAKE_MEDIA ), +};#define __vlv_reg_read_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (!NEEDS_FORCE_WAKE(offset)) \ - __fwd = 0; \ - else if (FORCEWAKE_VLV_RENDER_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_RENDER; \ - else if (FORCEWAKE_VLV_MEDIA_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_MEDIA; \ + if (NEEDS_FORCE_WAKE((offset))) { \ + __fwd = find_fw_domain(offset, __vlv_fw_ranges, \ + ARRAY_SIZE(__vlv_fw_ranges)); \ + if (__fwd == -1 ) \ + __fwd = 0; \ + } \ __fwd; \ })
@@ -636,104 +660,78 @@ static bool is_gen8_shadowed(u32 offset) __fwd; \ }) -#define FORCEWAKE_CHV_RENDER_RANGE_OFFSET (reg) \ - (REG_RANGE((reg), 0x2000, 0x4000) || \ - REG_RANGE((reg), 0x5200, 0x8000) || \ - REG_RANGE((reg), 0x8300, 0x8500) || \ - REG_RANGE((reg), 0xB000, 0xB480) || \ - REG_RANGE((reg), 0xE000, 0xE800)) - -#define FORCEWAKE_CHV_MEDIA_RANGE_OFFSET (reg) \ - (REG_RANGE((reg), 0x8800, 0x8900) || \ - REG_RANGE((reg), 0xD000, 0xD800) || \ - REG_RANGE((reg), 0x12000, 0x14000) || \ - REG_RANGE((reg), 0x1A000, 0x1C000) || \ - REG_RANGE((reg), 0x1E800, 0x1EA00) || \ - REG_RANGE((reg), 0x30000, 0x38000)) - -#define FORCEWAKE_CHV_COMMON_RANGE_OFFSET (reg) \ - (REG_RANGE((reg), 0x4000, 0x5000) || \ - REG_RANGE((reg), 0x8000, 0x8300) || \ - REG_RANGE((reg), 0x8500, 0x8600) || \ - REG_RANGE((reg), 0x9000, 0xB000) || \ - REG_RANGE((reg), 0xF000, 0x10000)) +static const struct intel_forcewake_range __chv_fw_ranges[] = { +GEN_FW_RANGE (0x2000 ,0x3fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x5200 ,0x7fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8300 ,0x84ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xe000 ,0xe7ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8800 ,0x88ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xd000 ,0xd7ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x1a000 ,0x1bfff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x1e800 ,0x1e9ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x30000 ,0x37fff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x4000 ,0x4fff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x8000 ,0x82ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x8500 ,0x85ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x9000 ,0xafff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xf000 ,0xffff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +};#define __chv_reg_read_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (!NEEDS_FORCE_WAKE(offset)) \ - __fwd = 0; \ - else if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_RENDER; \ - else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_MEDIA; \ - else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_RENDER | FORCEWAKE_MEDIA; \ + if (NEEDS_FORCE_WAKE((offset))) { \ + __fwd = find_fw_domain(offset, __chv_fw_ranges, \ + ARRAY_SIZE(__chv_fw_ranges)); \ + if (__fwd == -1 ) \ + __fwd = 0; \ + } \ __fwd; \ })#define __chv_reg_write_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (!NEEDS_FORCE_WAKE(offset) || is_gen8_shadowed(offset)) \ - __fwd = 0; \ - else if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_RENDER; \ - else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_MEDIA; \ - else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_RENDER | FORCEWAKE_MEDIA; \ + if (NEEDS_FORCE_WAKE((offset)) && !is_gen8_shadowed(offset)) { \ + __fwd = find_fw_domain(offset, __chv_fw_ranges, \ + ARRAY_SIZE(__chv_fw_ranges)); \ + if (__fwd == -1 ) \ + __fwd = 0; \ + } \ __fwd; \ }) -#define FORCEWAKE_GEN9_UNCORE_RANGE_OFFSET (reg) \ - REG_RANGE((reg), 0xB00, 0x2000) - -#define FORCEWAKE_GEN9_RENDER_RANGE_OFFSET (reg) \ - (REG_RANGE((reg), 0x2000, 0x2700) || \ - REG_RANGE((reg), 0x3000, 0x4000) || \ - REG_RANGE((reg), 0x5200, 0x8000) || \ - REG_RANGE((reg), 0x8140, 0x8160) || \ - REG_RANGE((reg), 0x8300, 0x8500) || \ - REG_RANGE((reg), 0x8C00, 0x8D00) || \ - REG_RANGE((reg), 0xB000, 0xB480) || \ - REG_RANGE((reg), 0xE000, 0xE900) || \ - REG_RANGE((reg), 0x24400, 0x24800)) - -#define FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET (reg) \ - (REG_RANGE((reg), 0x8130, 0x8140) || \ - REG_RANGE((reg), 0x8800, 0x8A00) || \ - REG_RANGE((reg), 0xD000, 0xD800) || \ - REG_RANGE((reg), 0x12000, 0x14000) || \ - REG_RANGE((reg), 0x1A000, 0x1EA00) || \ - REG_RANGE((reg), 0x30000, 0x40000)) - -#define FORCEWAKE_GEN9_COMMON_RANGE_OFFSET (reg) \ - REG_RANGE((reg), 0x9400, 0x9800) - -#define FORCEWAKE_GEN9_BLITTER_RANGE_OFFSET (reg) \ - ((reg) < 0x40000 && \ - !FORCEWAKE_GEN9_UNCORE_RANGE_OFFSET(reg) && \ - !FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg) && \ - !FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg) && \ - !FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) - -#define SKL_NEEDS_FORCE_WAKE (reg) \ - ((reg) < 0x40000 && !FORCEWAKE_GEN9_UNCORE_RANGE_OFFSET(reg)) +static const struct intel_forcewake_range __gen9_fw_ranges[] = { +GEN_FW_RANGE (0xb00 ,0x1fff ,0 ),/* uncore range */ +GEN_FW_RANGE (0x2000 ,0x26ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x3000 ,0x3fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x5200 ,0x7fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8140 ,0x815f ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8300 ,0x84ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8c00 ,0x8cff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xe000 ,0xe8ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x24400 ,0x247ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x9400 ,0x97ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x8130 ,0x813f ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x8800 ,0x89ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xd000 ,0xd7ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x1a000 ,0x1e9ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x30000 ,0x3ffff ,FORCEWAKE_MEDIA ), +};#define __gen9_reg_read_fw_domains (offset) \ ({ \ - enum forcewake_domains __fwd; \ - if (!SKL_NEEDS_FORCE_WAKE(offset)) \ - __fwd = 0; \ - else if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_RENDER; \ - else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_MEDIA; \ - else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_RENDER | FORCEWAKE_MEDIA; \ - else \ - __fwd = FORCEWAKE_BLITTER; \ + enum forcewake_domains __fwd = 0; \ + if (NEEDS_FORCE_WAKE((offset))) { \ + __fwd = find_fw_domain(offset, __gen9_fw_ranges, \ + ARRAY_SIZE(__gen9_fw_ranges)); \ + if (__fwd == -1 ) \ + __fwd = FORCEWAKE_BLITTER; \ + } \ __fwd; \ })
@@ -759,17 +757,13 @@ static bool is_gen9_shadowed(u32 offset)#define __gen9_reg_write_fw_domains (offset) \ ({ \ - enum forcewake_domains __fwd; \ - if (!SKL_NEEDS_FORCE_WAKE(offset) || is_gen9_shadowed(offset)) \ - __fwd = 0; \ - else if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_RENDER; \ - else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_MEDIA; \ - else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(offset)) \ - __fwd = FORCEWAKE_RENDER | FORCEWAKE_MEDIA; \ - else \ - __fwd = FORCEWAKE_BLITTER; \ + enum forcewake_domains __fwd = 0; \ + if (NEEDS_FORCE_WAKE((offset)) && !is_gen9_shadowed(offset)) { \ + __fwd = find_fw_domain(offset, __gen9_fw_ranges, \ + ARRAY_SIZE(__gen9_fw_ranges)); \ + if (__fwd == -1 ) \ + __fwd = FORCEWAKE_BLITTER; \ + } \ __fwd; \ })
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 5/13 drm/i915: Sort forcewake mapping tables
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
Sorting the tables (verified at runtime to help during development) is another prerequisite for interesting work which will follow.
--- drivers/gpu/drm/i915/intel_uncore.c | 55 ++++++++++++++++++++++++++++--------- 1 file changed, 42 insertions(+), 13 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +42 -13
@@ -605,16 +605,34 @@ find_fw_domain(u32 offset, const struct intel_forcewake_range *ranges,return -1 ; } +static void +intel_fw_table_check (const struct intel_forcewake_range *ranges, +unsigned int num_ranges) +{ +unsigned int i; +struct intel_forcewake_range *entry = + (struct intel_forcewake_range *)ranges; +s32 prev =-1 ; + +for (i =0 ; i < num_ranges; i++, entry++) { +WARN_ON_ONCE (prev >= (s32 )entry->start); + prev = entry->start; +WARN_ON_ONCE (prev >= (s32 )entry->end); + prev = entry->end; + } +} +#define GEN_FW_RANGE (s, e, d) \ { .start = (s), .end = (e), .domains = (d) } +/* *Must* be sorted by offset ranges! */ static const struct intel_forcewake_range __vlv_fw_ranges[] = {GEN_FW_RANGE (0x2000 ,0x3fff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x5000 ,0x7fff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0xb000 ,0x11fff ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0x2e000 ,0x2ffff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x22000 ,0x23fff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x2e000 ,0x2ffff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x30000 ,0x3ffff ,FORCEWAKE_MEDIA ), };
@@ -660,23 +678,24 @@ static bool is_gen8_shadowed(u32 offset) __fwd; \ }) +/* *Must* be sorted by offset ranges! */ static const struct intel_forcewake_range __chv_fw_ranges[] = {GEN_FW_RANGE (0x2000 ,0x3fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x4000 ,0x4fff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x5200 ,0x7fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8000 ,0x82ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8300 ,0x84ff ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0xe000 ,0xe7ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8500 ,0x85ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8800 ,0x88ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x9000 ,0xafff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0xd000 ,0xd7ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xe000 ,0xe7ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xf000 ,0xffff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x1a000 ,0x1bfff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x1e800 ,0x1e9ff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x30000 ,0x37fff ,FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x4000 ,0x4fff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x8000 ,0x82ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x8500 ,0x85ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x9000 ,0xafff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0xf000 ,0xffff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), };#define __chv_reg_read_fw_domains (offset) \
@@ -703,23 +722,24 @@ static const struct intel_forcewake_range __chv_fw_ranges[] = { __fwd; \ }) +/* *Must* be sorted by offset ranges! */ static const struct intel_forcewake_range __gen9_fw_ranges[] = {GEN_FW_RANGE (0xb00 ,0x1fff ,0 ),/* uncore range */ GEN_FW_RANGE (0x2000 ,0x26ff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x3000 ,0x3fff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x5200 ,0x7fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8130 ,0x813f ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8140 ,0x815f ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x8300 ,0x84ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8800 ,0x89ff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8c00 ,0x8cff ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0xe000 ,0xe8ff ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0x24400 ,0x247ff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x9400 ,0x97ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x8130 ,0x813f ,FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x8800 ,0x89ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0xd000 ,0xd7ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xe000 ,0xe8ff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x1a000 ,0x1e9ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x24400 ,0x247ff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x30000 ,0x3ffff ,FORCEWAKE_MEDIA ), };
@@ -1299,11 +1319,17 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)switch (INTEL_INFO (dev_priv)->gen) {default :case 9 : +intel_fw_table_check (__gen9_fw_ranges, +ARRAY_SIZE (__gen9_fw_ranges)); +ASSIGN_WRITE_MMIO_VFUNCS (gen9);ASSIGN_READ_MMIO_VFUNCS (gen9);break ;case 8 :if (IS_CHERRYVIEW (dev_priv)) { +intel_fw_table_check (__chv_fw_ranges, +ARRAY_SIZE (__chv_fw_ranges)); +ASSIGN_WRITE_MMIO_VFUNCS (chv);ASSIGN_READ_MMIO_VFUNCS (chv);
@@ -1317,6 +1343,9 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)ASSIGN_WRITE_MMIO_VFUNCS (gen6);if (IS_VALLEYVIEW (dev_priv)) { +intel_fw_table_check (__vlv_fw_ranges, +ARRAY_SIZE (__vlv_fw_ranges)); +ASSIGN_READ_MMIO_VFUNCS (vlv); }else {ASSIGN_READ_MMIO_VFUNCS (gen6);
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 05/13] drm/i915: Sort forcewake mapping tables
Quoted 16 lines · Tvrtko Ursulin
From: Tvrtko Ursulin <[email protected]>
Sorting the tables (verified at runtime to help during development) is another prerequisite for interesting work which will follow.
Signed-off-by Tvrtko Ursulin <[email protected]>--- drivers/gpu/drm/i915/intel_uncore.c | 55 ++++++++++++++++++++++++++++--------- 1 file changed, 42 insertions(+), 13 deletions(-)drivers/gpu/drm/i915/intel_uncore.c +0
-0
drivers/gpu/drm/i915/intel_uncore.c find_fw_domain(u32 offset, const struct intel_forcewake_range *ranges, · line 614
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +7
-0@@ -605,16 +605,34 @@ find_fw_domain(u32 offset, const struct intel_forcewake_range *ranges,return -1 ; } +static void +intel_fw_table_check (const struct intel_forcewake_range *ranges, +unsigned int num_ranges) +{ +unsigned int i; +struct intel_forcewake_range *entry = + (struct intel_forcewake_range *)ranges;
Why the const cast away?
drivers/gpu/drm/i915/intel_uncore.c find_fw_domain(u32 offset, const struct intel_forcewake_range *ranges, · line 628
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +9
-0+s32 prev =-1 ; + +for (i =0 ; i < num_ranges; i++, entry++) { +WARN_ON_ONCE (prev >= (s32 )entry->start); + prev = entry->start; +WARN_ON_ONCE (prev >= (s32 )entry->end); + prev = entry->end; + } +/* *Must* be sorted by offset ranges! */
/* *Must* be sorted by offset ranges! See intel_fw_table_check() */
drivers/gpu/drm/i915/intel_uncore.c find_fw_domain(u32 offset, const struct intel_forcewake_range *ranges, · line 637
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +1
-1static const struct intel_forcewake_range __vlv_fw_ranges[] = {GEN_FW_RANGE (0x2000 ,0x3fff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x5000 ,0x7fff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0xb000 ,0x11fff ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0x2e000 ,0x2ffff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x22000 ,0x23fff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x2e000 ,0x2ffff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x30000 ,0x3ffff ,FORCEWAKE_MEDIA ), };
Ok.
Quoted 20 lines
drivers/gpu/drm/i915/intel_uncore.c +8
-2@@ -660,23 +678,24 @@ static bool is_gen8_shadowed(u32 offset) __fwd; \ }) +/* *Must* be sorted by offset ranges! */ static const struct intel_forcewake_range __chv_fw_ranges[] = {GEN_FW_RANGE (0x2000 ,0x3fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x4000 ,0x4fff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x5200 ,0x7fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8000 ,0x82ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8300 ,0x84ff ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0xe000 ,0xe7ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8500 ,0x85ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8800 ,0x88ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x9000 ,0xafff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0xd000 ,0xd7ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xe000 ,0xe7ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xf000 ,0xffff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ),
drivers/gpu/drm/i915/intel_uncore.c static bool is_gen8_shadowed(u32 offset) · line 699
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +0
-5GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x1a000 ,0x1bfff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x1e800 ,0x1e9ff ,FORCEWAKE_MEDIA ), GEN_FW_RANGE(0x30000, 0x37fff, FORCEWAKE_MEDIA), -GEN_FW_RANGE (0x4000 ,0x4fff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x8000 ,0x82ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x8500 ,0x85ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x9000 ,0xafff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), - GEN_FW_RANGE(0xf000, 0xffff, FORCEWAKE_RENDER | FORCEWAKE_MEDIA), };
Ok.
Quoted 20 lines
drivers/gpu/drm/i915/intel_uncore.c +3
-3#define __chv_reg_read_fw_domains (offset) \@@ -703,23 +722,24 @@ static const struct intel_forcewake_range __chv_fw_ranges[] = { __fwd; \ }) +/* *Must* be sorted by offset ranges! */ static const struct intel_forcewake_range __gen9_fw_ranges[] = {GEN_FW_RANGE (0xb00 ,0x1fff ,0 ),/* uncore range */ GEN_FW_RANGE (0x2000 ,0x26ff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x3000 ,0x3fff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x5200 ,0x7fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8130 ,0x813f ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8140 ,0x815f ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x8300 ,0x84ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8800 ,0x89ff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8c00 ,0x8cff ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0xe000 ,0xe8ff ,FORCEWAKE_RENDER ), -GEN_FW_RANGE (0x24400 ,0x247ff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x9400 ,0x97ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ),
drivers/gpu/drm/i915/intel_uncore.c static const struct intel_forcewake_range __chv_fw_ranges[] = { · line 744
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +3
-2-GEN_FW_RANGE (0x8130 ,0x813f ,FORCEWAKE_MEDIA ), -GEN_FW_RANGE (0x8800 ,0x89ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0xd000 ,0xd7ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xe000 ,0xe8ff ,FORCEWAKE_RENDER ),GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x1a000 ,0x1e9ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x24400 ,0x247ff ,FORCEWAKE_RENDER ), GEN_FW_RANGE(0x30000, 0x3ffff, FORCEWAKE_MEDIA), };
Ok.
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 6/13 drm/i915: Use binary search when looking up forcewake domains
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
Instead of the existing linear seach, now that we have sorted range tables, we can do a binary search on them for some potential miniscule performance gain, but more importantly for elegance and code size. Hopefully the perfomance gain is sufficient to offset the function calls which were not there before.
--- drivers/gpu/drm/i915/intel_uncore.c | 28 ++++++++++++++++++++-------- 1 file changed, 20 insertions(+), 8 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +20 -8
@@ -26,6 +26,7 @@#include "i915_vgpu.h" #include <linux/pm_runtime.h> +#include <linux/bsearch.h> #define FORCEWAKE_ACK_TIMEOUT_MS 50
@@ -589,20 +590,31 @@ struct intel_forcewake_rangeenum forcewake_domains domains; }; +static int fw_range_cmp (const void *key,const void *elt) +{ +struct intel_forcewake_range *entry = + (struct intel_forcewake_range *)elt; +u32 offset = (u32)((unsigned long )key); + +if (offset < entry->start) +return -1 ; +else if (offset > entry->end) +return 1 ; +else +return 0 ; +} +static enum forcewake_domains find_fw_domain (u32 offset,const struct intel_forcewake_range *ranges,unsigned int num_ranges) { -unsigned int i; -struct intel_forcewake_range *entry = - (struct intel_forcewake_range *)ranges; +struct intel_forcewake_range *entry; -for (i =0 ; i < num_ranges; i++, entry++) { -if (offset >= entry->start && offset <= entry->end) -return entry->domains; - } + entry =bsearch ((void *)(unsigned long )offset, (const void *)ranges, + num_ranges,sizeof (struct intel_forcewake_range ), + fw_range_cmp); -return -1 ; +return entry ? entry->domains :-1 ; } static void
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/13] drm/i915: Use binary search when looking up forcewake domains
Quoted 50 lines · Tvrtko Ursulin
From: Tvrtko Ursulin <[email protected]>
Instead of the existing linear seach, now that we have sorted range tables, we can do a binary search on them for some potential miniscule performance gain, but more importantly for elegance and code size. Hopefully the perfomance gain is sufficient to offset the function calls which were not there before.
Signed-off-by Tvrtko Ursulin <[email protected]>--- drivers/gpu/drm/i915/intel_uncore.c | 28 ++++++++++++++++++++-------- 1 file changed, 20 insertions(+), 8 deletions(-)drivers/gpu/drm/i915/intel_uncore.c +15
-2@@ -26,6 +26,7 @@#include "i915_vgpu.h" #include <linux/pm_runtime.h> +#include <linux/bsearch.h> #define FORCEWAKE_ACK_TIMEOUT_MS 50@@ -589,20 +590,31 @@ struct intel_forcewake_rangeenum forcewake_domains domains; }; +static int fw_range_cmp (const void *key,const void *elt) +{ +struct intel_forcewake_range *entry = + (struct intel_forcewake_range *)elt; +u32 offset = (u32)((unsigned long )key); + +if (offset < entry->start) +return -1 ; +else if (offset > entry->end) +return 1 ; +else +return 0 ; +} +static enum forcewake_domains find_fw_domain (u32 offset,const struct intel_forcewake_range *ranges,unsigned int num_ranges) { -unsigned int i; - struct intel_forcewake_range *entry =
drivers/gpu/drm/i915/intel_uncore.c struct intel_forcewake_range · line 615
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +4
-5- (struct intel_forcewake_range *)ranges; +struct intel_forcewake_range *entry; -for (i =0 ; i < num_ranges; i++, entry++) { -if (offset >= entry->start && offset <= entry->end) -return entry->domains; - } + entry =bsearch ((void *)(unsigned long )offset, (const void *)ranges, + num_ranges,sizeof (struct intel_forcewake_range ), + fw_range_cmp);
How much for bsearch() to be turned into a generator macro?
drivers/gpu/drm/i915/intel_uncore.c struct intel_forcewake_range · line 618
Quoted 3 lines
drivers/gpu/drm/i915/intel_uncore.c +1
-1-return -1 ; +return entry ? entry->domains :-1 ; }
Looks ok, maybe pass in the default value to return if !entry, saves the double check.
-Chris
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 1/1 lib: Typesafe generator macro for bsearch
Sometimes a callout to a generic bsearch() library function is substantial overhead for a small search utility. For these situations, macro generate a type-specific bsearch routine.
--- include/linux/bsearch.h | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
include/linux/bsearch.h +18 -0
@@ -6,4 +6,22 @@void *bsearch (const void *key,const void *base,size_t num,size_t size,int (*cmp)(const void *key,const void *elt)); +#define BSEARCH (key, base, num, cmp) ({ \ + unsigned long start__ = 0, end__ = (num); \ + typeof(base) result__ = NULL; \ + while (start__ < end__) { \ + unsigned long mid__ = (start__ + end__) / 2; \ + int ret__ = (cmp)((key), (base) + mid__); \ + if (ret__ < 0) { \ + end__ = mid__; \ + } else if (ret__ > 0) { \ + start__ = mid__ + 1; \ + } else { \ + result__ = (base) + mid__; \ + break; \ + } \ + } \ + result__; \ +}) +#endif /* _LINUX_BSEARCH_H */
Signature · 6 lines
2.9.3 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/13] drm/i915: Use binary search when looking up forcewake domains
In reply to Chris WilsonQuoted 54 lines · Chris Wilson
Quoted 53 lines · Tvrtko Ursulin
From: Tvrtko Ursulin <[email protected]>
Instead of the existing linear seach, now that we have sorted range tables, we can do a binary search on them for some potential miniscule performance gain, but more importantly for elegance and code size. Hopefully the perfomance gain is sufficient to offset the function calls which were not there before.
Signed-off-by Tvrtko Ursulin <[email protected]>--- drivers/gpu/drm/i915/intel_uncore.c | 28 ++++++++++++++++++++-------- 1 file changed, 20 insertions(+), 8 deletions(-)drivers/gpu/drm/i915/intel_uncore.c +16
-3@@ -26,6 +26,7 @@#include "i915_vgpu.h" #include <linux/pm_runtime.h> +#include <linux/bsearch.h> #define FORCEWAKE_ACK_TIMEOUT_MS 50@@ -589,20 +590,31 @@ struct intel_forcewake_rangeenum forcewake_domains domains; }; +static int fw_range_cmp (const void *key,const void *elt) +{ +struct intel_forcewake_range *entry = + (struct intel_forcewake_range *)elt; +u32 offset = (u32)((unsigned long )key); + +if (offset < entry->start) +return -1 ; +else if (offset > entry->end) +return 1 ; +else +return 0 ; +} +static enum forcewake_domains find_fw_domain (u32 offset,const struct intel_forcewake_range *ranges,unsigned int num_ranges) { -unsigned int i; -struct intel_forcewake_range *entry = - (struct intel_forcewake_range *)ranges; +struct intel_forcewake_range *entry;
Quoted 8 lines
drivers/gpu/drm/i915/intel_uncore.c struct intel_forcewake_range · line 615
Quoted 7 lines
drivers/gpu/drm/i915/intel_uncore.c +3
-4-for (i =0 ; i < num_ranges; i++, entry++) { -if (offset >= entry->start && offset <= entry->end) -return entry->domains; - } + entry =bsearch ((void *)(unsigned long )offset, (const void *)ranges, + num_ranges,sizeof (struct intel_forcewake_range ), + fw_range_cmp);How much for
bsearch()to be turned into a generator macro?
By default it is a small code size win (128 bytes). It makes find_fw_domain a function with an inlined comparator (so one function call less per search iteration than using library bsearch) and inlines is_gen8_shadowed completely.
Forcing find_fw_domain to be fully inline adds approximately 1k.
I am not sure - you think it is worth doing some of the above? Function calls are supposed to be cheap so perhaps just with the default inlining, but then it is either pushing the core patch or having a local copy of a macro.
Quoted 5 lines
drivers/gpu/drm/i915/intel_uncore.c struct intel_forcewake_range · line 618
Quoted 3 lines
drivers/gpu/drm/i915/intel_uncore.c +1
-1-return -1 ; +return entry ? entry->domains :-1 ; }Looks ok, maybe pass in the default value to return if !entry, saves the double check.
It goes away later in the series so it is fine. I just wanted to be gradual so it is easy to review.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/13] drm/i915: Use binary search when looking up forcewake domains
Quoted 13 lines · Tvrtko Ursulin
Quoted 5 lines · Chris Wilson
drivers/gpu/drm/i915/intel_uncore.c struct intel_forcewake_range · line 615
Quoted 3 lines · Tvrtko Ursulin
drivers/gpu/drm/i915/intel_uncore.c +3
-0+ entry =bsearch ((void *)(unsigned long )offset, (const void *)ranges, + num_ranges,sizeof (struct intel_forcewake_range ), + fw_range_cmp);How much for
bsearch()to be turned into a generator macro?By default it is a small code size win (128 bytes). It makes
find_fw_domaina function with an inlined comparator (so one function call less per search iteration than using library bsearch) and inlinesis_gen8_shadowedcompletely.
Quoted 6 lines
Forcing
find_fw_domainto be fully inline adds approximately 1k.I am not sure - you think it is worth doing some of the above? Function calls are supposed to be cheap so perhaps just with the default inlining, but then it is either pushing the core patch or having a local copy of a macro.
But I wonder if we get better branch predictor from inlining.
Happy enough with the default inlining, whilst we are in the noise compared to the actual fw and mmio, we might as well try to keep the irq paths trim at least.
-Chris
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 7/13 drm/i915: Eliminate Gen9 special case
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
If we insert blitter forcewake domain entries in the range table we can eliminate that special case and simplify the code in a few macros. This will enable more unification later.
--- drivers/gpu/drm/i915/intel_uncore.c | 42 ++++++++++++++++++------------------- 1 file changed, 21 insertions(+), 21 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +21 -21
@@ -614,7 +614,7 @@ find_fw_domain(u32 offset, const struct intel_forcewake_range *ranges, num_ranges,sizeof (struct intel_forcewake_range ), fw_range_cmp); -return entry ? entry->domains :-1 ; +return entry ? entry->domains :0 ; } static void
@@ -651,12 +651,9 @@ static const struct intel_forcewake_range __vlv_fw_ranges[] = {#define __vlv_reg_read_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset))) { \ + if (NEEDS_FORCE_WAKE((offset))) \ __fwd = find_fw_domain(offset, __vlv_fw_ranges, \ ARRAY_SIZE(__vlv_fw_ranges)); \ - if (__fwd == -1 ) \ - __fwd = 0; \ - } \ __fwd; \ })
@@ -713,57 +710,63 @@ static const struct intel_forcewake_range __chv_fw_ranges[] = {#define __chv_reg_read_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset))) { \ + if (NEEDS_FORCE_WAKE((offset))) \ __fwd = find_fw_domain(offset, __chv_fw_ranges, \ ARRAY_SIZE(__chv_fw_ranges)); \ - if (__fwd == -1 ) \ - __fwd = 0; \ - } \ __fwd; \ })#define __chv_reg_write_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset)) && !is_gen8_shadowed(offset)) { \ + if (NEEDS_FORCE_WAKE((offset)) && !is_gen8_shadowed(offset)) \ __fwd = find_fw_domain(offset, __chv_fw_ranges, \ ARRAY_SIZE(__chv_fw_ranges)); \ - if (__fwd == -1 ) \ - __fwd = 0; \ - } \ __fwd; \ })/* *Must* be sorted by offset ranges! */ static const struct intel_forcewake_range __gen9_fw_ranges[] = { +GEN_FW_RANGE (0x0 ,0xaff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0xb00 ,0x1fff ,0 ),/* uncore range */ GEN_FW_RANGE (0x2000 ,0x26ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x2700 ,0x2fff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x3000 ,0x3fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x4000 ,0x51ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x5200 ,0x7fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8000 ,0x812f ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x8130 ,0x813f ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8140 ,0x815f ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8160 ,0x82ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x8300 ,0x84ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8500 ,0x87ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x8800 ,0x89ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x8a00 ,0x8bff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x8c00 ,0x8cff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8d00 ,0x93ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x9400 ,0x97ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x9800 ,0xafff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xb480 ,0xbfff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0xd000 ,0xd7ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xd800 ,0xdfff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0xe000 ,0xe8ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xe900 ,0x11fff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x14000 ,0x19fff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x1a000 ,0x1e9ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x1ea00 ,0x243ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x24400 ,0x247ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x24800 ,0x2ffff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x30000 ,0x3ffff ,FORCEWAKE_MEDIA ), };#define __gen9_reg_read_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset))) { \ + if (NEEDS_FORCE_WAKE((offset))) \ __fwd = find_fw_domain(offset, __gen9_fw_ranges, \ ARRAY_SIZE(__gen9_fw_ranges)); \ - if (__fwd == -1 ) \ - __fwd = FORCEWAKE_BLITTER; \ - } \ __fwd; \ })
@@ -790,12 +793,9 @@ static bool is_gen9_shadowed(u32 offset)#define __gen9_reg_write_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset)) && !is_gen9_shadowed(offset)) { \ + if (NEEDS_FORCE_WAKE((offset)) && !is_gen9_shadowed(offset)) \ __fwd = find_fw_domain(offset, __gen9_fw_ranges, \ ARRAY_SIZE(__gen9_fw_ranges)); \ - if (__fwd == -1 ) \ - __fwd = FORCEWAKE_BLITTER; \ - } \ __fwd; \ })
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 07/13] drm/i915: Eliminate Gen9 special case
Quoted 100 lines · Tvrtko Ursulin
From: Tvrtko Ursulin <[email protected]>
If we insert blitter forcewake domain entries in the range table we can eliminate that special case and simplify the code in a few macros. This will enable more unification later.
Signed-off-by Tvrtko Ursulin <[email protected]>--- drivers/gpu/drm/i915/intel_uncore.c | 42 ++++++++++++++++++------------------- 1 file changed, 21 insertions(+), 21 deletions(-)drivers/gpu/drm/i915/intel_uncore.c +19
-13@@ -614,7 +614,7 @@ find_fw_domain(u32 offset, const struct intel_forcewake_range *ranges, num_ranges,sizeof (struct intel_forcewake_range ), fw_range_cmp); -return entry ? entry->domains :-1 ; +return entry ? entry->domains :0 ; } static void@@ -651,12 +651,9 @@ static const struct intel_forcewake_range __vlv_fw_ranges[] = {#define __vlv_reg_read_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset))) { \ + if (NEEDS_FORCE_WAKE((offset))) \ __fwd = find_fw_domain(offset, __vlv_fw_ranges, \ ARRAY_SIZE(__vlv_fw_ranges)); \ - if (__fwd == -1 ) \ - __fwd = 0; \ - } \ __fwd; \ })@@ -713,57 +710,63 @@ static const struct intel_forcewake_range __chv_fw_ranges[] = {#define __chv_reg_read_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset))) { \ + if (NEEDS_FORCE_WAKE((offset))) \ __fwd = find_fw_domain(offset, __chv_fw_ranges, \ ARRAY_SIZE(__chv_fw_ranges)); \ - if (__fwd == -1 ) \ - __fwd = 0; \ - } \ __fwd; \ })#define __chv_reg_write_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset)) && !is_gen8_shadowed(offset)) { \ + if (NEEDS_FORCE_WAKE((offset)) && !is_gen8_shadowed(offset)) \ __fwd = find_fw_domain(offset, __chv_fw_ranges, \ ARRAY_SIZE(__chv_fw_ranges)); \ - if (__fwd == -1 ) \ - __fwd = 0; \ - } \ __fwd; \ })/* *Must* be sorted by offset ranges! */ static const struct intel_forcewake_range __gen9_fw_ranges[] = { +GEN_FW_RANGE (0x0 ,0xaff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0xb00 ,0x1fff ,0 ),/* uncore range */ GEN_FW_RANGE (0x2000 ,0x26ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x2700 ,0x2fff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x3000 ,0x3fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x4000 ,0x51ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x5200 ,0x7fff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8000 ,0x812f ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x8130 ,0x813f ,FORCEWAKE_MEDIA ),GEN_FW_RANGE (0x8140 ,0x815f ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8160 ,0x82ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x8300 ,0x84ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8500 ,0x87ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x8800 ,0x89ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x8a00 ,0x8bff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x8c00 ,0x8cff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x8d00 ,0x93ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x9400 ,0x97ff ,FORCEWAKE_RENDER |FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x9800 ,0xafff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0xb000 ,0xb47f ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xb480 ,0xbfff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0xd000 ,0xd7ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0xd800 ,0xdfff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0xe000 ,0xe8ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0xe900 ,0x11fff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x12000 ,0x13fff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x14000 ,0x19fff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x1a000 ,0x1e9ff ,FORCEWAKE_MEDIA ), +GEN_FW_RANGE (0x1ea00 ,0x243ff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x24400 ,0x247ff ,FORCEWAKE_RENDER ), +GEN_FW_RANGE (0x24800 ,0x2ffff ,FORCEWAKE_BLITTER ),GEN_FW_RANGE (0x30000 ,0x3ffff ,FORCEWAKE_MEDIA ), };
drivers/gpu/drm/i915/intel_uncore.c static const struct intel_forcewake_range __chv_fw_ranges[] = { · line 765
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +1
-3#define __gen9_reg_read_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset))) { \ + if (NEEDS_FORCE_WAKE((offset))) \ __fwd = find_fw_domain(offset, __gen9_fw_ranges, \ ARRAY_SIZE(__gen9_fw_ranges)); \ - if (__fwd == -1 ) \ - __fwd = FORCEWAKE_BLITTER; \
Hmm. Leaving the holes halves the table size, and saves approximately one pass... :|
Being explicit in the declaration is better, so lgtm. -Chris
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 8/13 drm/i915: Store the active forcewake range table pointer
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
If we store this in the uncore structure we are on a good way to show more commonality between the per-platform implementations.
--- drivers/gpu/drm/i915/i915_drv.h | 11 ++++++ drivers/gpu/drm/i915/intel_uncore.c | 74 ++++++++++++++++++------------------- 2 files changed, 47 insertions(+), 38 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +11 -0
@@ -582,9 +582,20 @@ struct intel_uncore_funcs { uint32_t val, bool trace); }; +struct intel_forcewake_range +{ +u32 start; +u32 end; + +enum forcewake_domains domains; +}; +struct intel_uncore {spinlock_t lock;/** lock is also taken in irq contexts. */ +struct intel_forcewake_range *fw_domains_table; +unsigned int fw_domains_table_entries; +struct intel_uncore_funcs funcs;unsigned fifo_count;
drivers/gpu/drm/i915/intel_uncore.c +36 -38
156 lines not on this page · raw message
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 08/13] drm/i915: Store the active forcewake range table pointer
Quoted 16 lines · Tvrtko Ursulin
From: Tvrtko Ursulin <[email protected]>
If we store this in the uncore structure we are on a good way to show more commonality between the per-platform implementations.
Signed-off-by Tvrtko Ursulin <[email protected]>--- drivers/gpu/drm/i915/i915_drv.h | 11 ++++++ drivers/gpu/drm/i915/intel_uncore.c | 74 ++++++++++++++++++------------------- 2 files changed, 47 insertions(+), 38 deletions(-)drivers/gpu/drm/i915/i915_drv.h +0
-0
drivers/gpu/drm/i915/i915_drv.h struct intel_uncore_funcs { · line 586
Quoted 5 lines
drivers/gpu/drm/i915/i915_drv.h +2
-0@@ -582,9 +582,20 @@ struct intel_uncore_funcs { uint32_t val, bool trace); }; +struct intel_forcewake_range +{
struct intel_forcewake_range {
drivers/gpu/drm/i915/i915_drv.h struct intel_uncore_funcs { · line 597
Quoted 11 lines
drivers/gpu/drm/i915/i915_drv.h +8
-0+u32 start; +u32 end; + +enum forcewake_domains domains; +}; +struct intel_uncore {spinlock_t lock;/** lock is also taken in irq contexts. */ +struct intel_forcewake_range *fw_domains_table; +unsigned int fw_domains_table_entries;
constify?
Quoted 167 lines
drivers/gpu/drm/i915/i915_drv.h +1
-0+struct intel_uncore_funcs funcs;unsigned fifo_count;drivers/gpu/drm/i915/intel_uncore.c +34
-38@@ -582,14 +582,6 @@ void assert_forcewakes_inactive(struct drm_i915_private *dev_priv) __fwd; \ }) -struct intel_forcewake_range -{ -u32 start; -u32 end; - -enum forcewake_domains domains; -}; -static int fw_range_cmp (const void *key,const void *elt) { struct intel_forcewake_range *entry =@@ -605,28 +597,36 @@ static int fw_range_cmp(const void *key, const void *elt) }static enum forcewake_domains -find_fw_domain (u32 offset,const struct intel_forcewake_range *ranges, -unsigned int num_ranges) +find_fw_domain (struct drm_i915_private *dev_priv,u32 offset) { -struct intel_forcewake_range *entry; +struct intel_forcewake_range *table, *entry; +unsigned int num_entries; - entry =bsearch ((void *)(unsigned long )offset, (const void *)ranges, - num_ranges,sizeof (struct intel_forcewake_range ), + table = dev_priv->uncore.fw_domains_table; + num_entries = dev_priv->uncore.fw_domains_table_entries; + + entry =bsearch ((void *)(unsigned long )offset, (const void *)table, + num_entries,sizeof (struct intel_forcewake_range ), fw_range_cmp);return entry ? entry->domains :0 ; }static void -intel_fw_table_check (const struct intel_forcewake_range *ranges, -unsigned int num_ranges) +intel_fw_table_check (struct drm_i915_private *dev_priv) {unsigned int i; -struct intel_forcewake_range *entry = - (struct intel_forcewake_range *)ranges; -s32 prev =-1 ; +struct intel_forcewake_range *entry; +unsigned int num_ranges; +s32 prev; -for (i =0 ; i < num_ranges; i++, entry++) { + entry = dev_priv->uncore.fw_domains_table; +if (!entry) +return ; + + num_ranges = dev_priv->uncore.fw_domains_table_entries; + +for (i =0 , prev =-1 ; i < num_ranges; i++, entry++) {WARN_ON_ONCE (prev >= (s32 )entry->start); prev = entry->start;WARN_ON_ONCE (prev >= (s32 )entry->end);@@ -652,8 +652,7 @@ static const struct intel_forcewake_range __vlv_fw_ranges[] = { ({ \enum forcewake_domains __fwd =0 ; \if (NEEDS_FORCE_WAKE ((offset))) \ - __fwd =find_fw_domain (offset, __vlv_fw_ranges, \ -ARRAY_SIZE (__vlv_fw_ranges)); \ + __fwd =find_fw_domain (dev_priv, offset); \ __fwd; \ })@@ -711,8 +710,7 @@ static const struct intel_forcewake_range __chv_fw_ranges[] = { ({ \enum forcewake_domains __fwd =0 ; \if (NEEDS_FORCE_WAKE ((offset))) \ - __fwd =find_fw_domain (offset, __chv_fw_ranges, \ -ARRAY_SIZE (__chv_fw_ranges)); \ + __fwd =find_fw_domain (dev_priv, offset); \ __fwd; \ })@@ -720,8 +718,7 @@ static const struct intel_forcewake_range __chv_fw_ranges[] = { ({ \enum forcewake_domains __fwd =0 ; \if (NEEDS_FORCE_WAKE ((offset)) && !is_gen8_shadowed (offset)) \ - __fwd =find_fw_domain (offset, __chv_fw_ranges, \ -ARRAY_SIZE (__chv_fw_ranges)); \ + __fwd =find_fw_domain (dev_priv, offset); \ __fwd; \ })@@ -765,8 +762,7 @@ static const struct intel_forcewake_range __gen9_fw_ranges[] = { ({ \enum forcewake_domains __fwd =0 ; \if (NEEDS_FORCE_WAKE ((offset))) \ - __fwd =find_fw_domain (offset, __gen9_fw_ranges, \ -ARRAY_SIZE (__gen9_fw_ranges)); \ + __fwd =find_fw_domain (dev_priv, offset); \ __fwd; \ })@@ -794,8 +790,7 @@ static bool is_gen9_shadowed(u32 offset) ({ \enum forcewake_domains __fwd =0 ; \if (NEEDS_FORCE_WAKE ((offset)) && !is_gen9_shadowed (offset)) \ - __fwd =find_fw_domain (offset, __gen9_fw_ranges, \ -ARRAY_SIZE (__gen9_fw_ranges)); \ + __fwd =find_fw_domain (dev_priv, offset); \ __fwd; \ })@@ -1318,6 +1313,13 @@ static void intel_uncore_fw_domains_init(struct drm_i915_private *dev_priv)WARN_ON (dev_priv->uncore.fw_domains ==0 ); } +#define ASSIGN_FW_DOMAINS_TABLE (d) \ +{ \ + dev_priv->uncore.fw_domains_table = \ + (struct intel_forcewake_range *)(d); \ + dev_priv->uncore.fw_domains_table_entries = ARRAY_SIZE((d)); \ +} +void intel_uncore_init (struct drm_i915_private *dev_priv) {i915_check_vgpu (dev_priv);@@ -1331,17 +1333,13 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)switch (INTEL_INFO (dev_priv)->gen) {default :case 9 : -intel_fw_table_check (__gen9_fw_ranges, -ARRAY_SIZE (__gen9_fw_ranges)); - +ASSIGN_FW_DOMAINS_TABLE (__gen9_fw_ranges);ASSIGN_WRITE_MMIO_VFUNCS (gen9);ASSIGN_READ_MMIO_VFUNCS (gen9);break ;case 8 :if (IS_CHERRYVIEW (dev_priv)) { -intel_fw_table_check (__chv_fw_ranges, -ARRAY_SIZE (__chv_fw_ranges)); - +ASSIGN_FW_DOMAINS_TABLE (__chv_fw_ranges);ASSIGN_WRITE_MMIO_VFUNCS (chv);ASSIGN_READ_MMIO_VFUNCS (chv);@@ -1355,9 +1353,7 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)ASSIGN_WRITE_MMIO_VFUNCS (gen6);if (IS_VALLEYVIEW (dev_priv)) { -intel_fw_table_check (__vlv_fw_ranges, -ARRAY_SIZE (__vlv_fw_ranges)); - +ASSIGN_FW_DOMAINS_TABLE (__vlv_fw_ranges);ASSIGN_READ_MMIO_VFUNCS (vlv); }else {ASSIGN_READ_MMIO_VFUNCS (gen6);
drivers/gpu/drm/i915/intel_uncore.c void intel_uncore_init(struct drm_i915_private *dev_priv) · line 1374
Quoted 4 lines
drivers/gpu/drm/i915/intel_uncore.c +1
-0@@ -1375,6 +1371,8 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)break ; } +intel_fw_table_check (dev_priv);
Looking better...
-Chris
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 9/13 drm/i915: Remove identical macros
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
Remove some macros which are now obviously identical.
--- drivers/gpu/drm/i915/intel_uncore.c | 30 +++++++----------------------- 1 file changed, 7 insertions(+), 23 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +7 -23
@@ -648,7 +648,7 @@ static const struct intel_forcewake_range __vlv_fw_ranges[] = { GEN_FW_RANGE(0x30000, 0x3ffff, FORCEWAKE_MEDIA), }; -#define __vlv_reg_read_fw_domains (offset) \ +#define __fwtable_reg_read_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ if (NEEDS_FORCE_WAKE((offset))) \
@@ -706,14 +706,6 @@ static const struct intel_forcewake_range __chv_fw_ranges[] = { GEN_FW_RANGE(0x30000, 0x37fff, FORCEWAKE_MEDIA), }; -#define __chv_reg_read_fw_domains (offset) \ -({ \ - enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset))) \ - __fwd = find_fw_domain(dev_priv, offset); \ - __fwd; \ -}) -#define __chv_reg_write_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \
@@ -758,14 +750,6 @@ static const struct intel_forcewake_range __gen9_fw_ranges[] = { GEN_FW_RANGE(0x30000, 0x3ffff, FORCEWAKE_MEDIA), }; -#define __gen9_reg_read_fw_domains (offset) \ -({ \ - enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset))) \ - __fwd = find_fw_domain(dev_priv, offset); \ - __fwd; \ -}) -static const i915_reg_t gen9_shadowed_regs[] = {RING_TAIL (RENDER_RING_BASE ),RING_TAIL (GEN6_BSD_RING_BASE ),
@@ -926,7 +910,7 @@ static u##x \ vlv_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \ enum forcewake_domains fw_engine; \ GEN6_READ_HEADER(x); \ - fw_engine = __vlv_reg_read_fw_domains(offset); \ + fw_engine = __fwtable_reg_read_fw_domains(offset); \ if (fw_engine) \ __force_wake_auto(dev_priv, fw_engine); \ val = __raw_i915_read##x(dev_priv, reg); \
@@ -938,7 +922,7 @@ static u##x \ chv_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \ enum forcewake_domains fw_engine; \ GEN6_READ_HEADER(x); \ - fw_engine = __chv_reg_read_fw_domains(offset); \ + fw_engine = __fwtable_reg_read_fw_domains(offset); \ if (fw_engine) \ __force_wake_auto(dev_priv, fw_engine); \ val = __raw_i915_read##x(dev_priv, reg); \
@@ -950,7 +934,7 @@ static u##x \ gen9_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \ enum forcewake_domains fw_engine; \ GEN6_READ_HEADER(x); \ - fw_engine = __gen9_reg_read_fw_domains(offset); \ + fw_engine = __fwtable_reg_read_fw_domains(offset); \ if (fw_engine) \ __force_wake_auto(dev_priv, fw_engine); \ val = __raw_i915_read##x(dev_priv, reg); \
@@ -1829,18 +1813,18 @@ intel_uncore_forcewake_for_read(struct drm_i915_private *dev_priv,switch (INTEL_GEN (dev_priv)) {case 9 : - fw_domains =__gen9_reg_read_fw_domains (i915_mmio_reg_offset (reg)); + fw_domains =__fwtable_reg_read_fw_domains (i915_mmio_reg_offset (reg));break ;case 8 :if (IS_CHERRYVIEW (dev_priv)) - fw_domains =__chv_reg_read_fw_domains (i915_mmio_reg_offset (reg)); + fw_domains =__fwtable_reg_read_fw_domains (i915_mmio_reg_offset (reg));else fw_domains =__gen6_reg_read_fw_domains (i915_mmio_reg_offset (reg));break ;case 7 :case 6 :if (IS_VALLEYVIEW (dev_priv)) - fw_domains =__vlv_reg_read_fw_domains (i915_mmio_reg_offset (reg)); + fw_domains =__fwtable_reg_read_fw_domains (i915_mmio_reg_offset (reg));else fw_domains =__gen6_reg_read_fw_domains (i915_mmio_reg_offset (reg));break ;
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 09/13] drm/i915: Remove identical macros
Quoted 9 lines · Tvrtko Ursulin
vlv_read##x(struct
drm_i915_private*dev_priv,i915_reg_treg, bool trace) { \enum
forcewake_domainsfw_engine; \
GEN6_READ_HEADER(x); \drivers/gpu/drm/i915/intel_uncore.c +1
-1- fw_engine =__vlv_reg_read_fw_domains (offset); \ + fw_engine =__fwtable_reg_read_fw_domains (offset); \if (
fw_engine) \
__force_wake_auto(dev_priv, fw_engine); \
val = __raw_i915_read##x(dev_priv, reg); \
I notice you in later patch remove the now equal funcs.
Quoted 19 lines
(quoted diff) +0
-0@@ -1829,18 +1813,18 @@ intel_uncore_forcewake_for_read(struct drm_i915_private *dev_priv,
switch (INTEL_GEN(dev_priv)) {
case 9:(quoted diff) +1
-1- fw_domains = __gen9_reg_read_fw_domains(i915_mmio_reg_offset(reg)); + fw_domains = __fwtable_reg_read_fw_domains(i915_mmio_reg_offset(reg));break;
case 8:
if (IS_CHERRYVIEW(dev_priv))(quoted diff) +1
-1- fw_domains = __chv_reg_read_fw_domains(i915_mmio_reg_offset(reg)); + fw_domains = __fwtable_reg_read_fw_domains(i915_mmio_reg_offset(reg));else
fw_domains= __gen6_reg_read_fw_domains(i915_mmio_reg_offset(reg)); break;
case 7:
case 6:
if (IS_VALLEYVIEW(dev_priv))drivers/gpu/drm/i915/intel_uncore.c +1
-1- fw_domains =__vlv_reg_read_fw_domains (i915_mmio_reg_offset (reg)); + fw_domains =__fwtable_reg_read_fw_domains (i915_mmio_reg_offset (reg));
Quoted 3 lines
else
fw_domains= __gen6_reg_read_fw_domains(i915_mmio_reg_offset(reg)); break;
Drop the whole switch, just do if (IS_CHERRYVIEW || IS_VALLEYVIEW) fwtable_ else gen6_. Maybe some HAS_FWTABLE() prop would work too, but that's not must.
With switch removed,
Regards, Joonas
Signature · 7 lines
Joonas Lahtinen Open Source Technology Center Intel Corporation _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 10/13 drm/i915: Remove identical mmio read functions
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
It is now obvious VLV, CHV and Gen9 mmio read fcuntions are completely identical so we can remove the three copies and just keep the newly named generic implementation.
--- drivers/gpu/drm/i915/intel_uncore.c | 54 +++++++------------------------------ 1 file changed, 10 insertions(+), 44 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +10 -44
@@ -905,9 +905,9 @@ gen6_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \GEN6_READ_FOOTER ; \ } -#define __vlv_read (x) \ +#define __fwtbl_read (x) \ static u##x \ -vlv_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \ +fwtbl_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \ enum forcewake_domains fw_engine; \ GEN6_READ_HEADER(x); \ fw_engine = __fwtable_reg_read_fw_domains(offset); \
@@ -917,50 +917,16 @@ vlv_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \GEN6_READ_FOOTER ; \ } -#define __chv_read (x) \ -static u##x \ -chv_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \ - enum forcewake_domains fw_engine; \ - GEN6_READ_HEADER(x); \ - fw_engine = __fwtable_reg_read_fw_domains(offset); \ - if (fw_engine) \ - __force_wake_auto(dev_priv, fw_engine); \ - val = __raw_i915_read##x(dev_priv, reg); \ - GEN6_READ_FOOTER; \ -} - -#define __gen9_read (x) \ -static u##x \ -gen9_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \ - enum forcewake_domains fw_engine; \ - GEN6_READ_HEADER(x); \ - fw_engine = __fwtable_reg_read_fw_domains(offset); \ - if (fw_engine) \ - __force_wake_auto(dev_priv, fw_engine); \ - val = __raw_i915_read##x(dev_priv, reg); \ - GEN6_READ_FOOTER; \ -} - -__gen9_read(8) -__gen9_read(16) -__gen9_read(32) -__gen9_read(64) -__chv_read(8) -__chv_read(16) -__chv_read(32) -__chv_read(64) -__vlv_read(8) -__vlv_read(16) -__vlv_read(32) -__vlv_read(64) +__fwtbl_read(8) +__fwtbl_read(16) +__fwtbl_read(32) +__fwtbl_read(64) __gen6_read(8) __gen6_read(16) __gen6_read(32) __gen6_read(64) -#undef __gen9_read -#undef __chv_read -#undef __vlv_read +#undef __fwtbl_read#undef __gen6_read#undef GEN6_READ_FOOTER#undef GEN6_READ_HEADER
@@ -1319,13 +1285,13 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)case 9 :ASSIGN_FW_DOMAINS_TABLE (__gen9_fw_ranges);ASSIGN_WRITE_MMIO_VFUNCS (gen9); -ASSIGN_READ_MMIO_VFUNCS (gen9); +ASSIGN_READ_MMIO_VFUNCS (fwtbl);break ;case 8 :if (IS_CHERRYVIEW (dev_priv)) {ASSIGN_FW_DOMAINS_TABLE (__chv_fw_ranges);ASSIGN_WRITE_MMIO_VFUNCS (chv); -ASSIGN_READ_MMIO_VFUNCS (chv); +ASSIGN_READ_MMIO_VFUNCS (fwtbl); }else {ASSIGN_WRITE_MMIO_VFUNCS (gen8);
@@ -1338,7 +1304,7 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)if (IS_VALLEYVIEW (dev_priv)) {ASSIGN_FW_DOMAINS_TABLE (__vlv_fw_ranges); -ASSIGN_READ_MMIO_VFUNCS (vlv); +ASSIGN_READ_MMIO_VFUNCS (fwtbl); }else {ASSIGN_READ_MMIO_VFUNCS (gen6); }
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 10/13] drm/i915: Remove identical mmio read functions
drivers/gpu/drm/i915/intel_uncore.c gen6_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \ · line 908
Quoted 7 lines · Tvrtko Ursulin
(quoted diff) +0
-0@@ -905,9 +905,9 @@ gen6_read##x(struct drm_i915_private *dev_priv, i915_reg_t reg, bool trace) { \ > GEN6_READ_FOOTER; \ }drivers/gpu/drm/i915/intel_uncore.c +1
-1-#define __vlv_read (x) \ +#define __fwtbl_read (x) \
Consistency is the key, use __fwtable or __fwtbl in the series, not both.
With the unification,
Regards, Joonas
Signature · 7 lines
Joonas Lahtinen Open Source Technology Center Intel Corporation _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 11/13 drm/i915: Remove identical write mmmio functions
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
We notice two identical copies of the shadow register table and following from that removal can also unify CHV and Gen9 write mmio functions and macros into a single implementation.
--- drivers/gpu/drm/i915/intel_uncore.c | 69 +++++++------------------------------ 1 file changed, 12 insertions(+), 57 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +12 -57
@@ -706,7 +706,7 @@ static const struct intel_forcewake_range __chv_fw_ranges[] = { GEN_FW_RANGE(0x30000, 0x37fff, FORCEWAKE_MEDIA), }; -#define __chv_reg_write_fw_domains (offset) \ +#define __fwtbl_reg_write_fw_domains (offset) \ ({ \ enum forcewake_domains __fwd = 0; \ if (NEEDS_FORCE_WAKE((offset)) && !is_gen8_shadowed(offset)) \
@@ -750,34 +750,6 @@ static const struct intel_forcewake_range __gen9_fw_ranges[] = { GEN_FW_RANGE(0x30000, 0x3ffff, FORCEWAKE_MEDIA), }; -static const i915_reg_t gen9_shadowed_regs[] = { -RING_TAIL (RENDER_RING_BASE ), -RING_TAIL (GEN6_BSD_RING_BASE ), -RING_TAIL (VEBOX_RING_BASE ), -RING_TAIL (BLT_RING_BASE ), -GEN6_RPNSWREQ , -GEN6_RC_VIDEO_FREQ , -/* TODO: Other registers are not yet used */ -}; - -static bool is_gen9_shadowed (u32 offset) -{ -int i; -for (i =0 ; i <ARRAY_SIZE (gen9_shadowed_regs); i++) -if (offset == gen9_shadowed_regs[i].reg) -return true; - -return false; -} - -#define __gen9_reg_write_fw_domains (offset) \ -({ \ - enum forcewake_domains __fwd = 0; \ - if (NEEDS_FORCE_WAKE((offset)) && !is_gen9_shadowed(offset)) \ - __fwd = find_fw_domain(dev_priv, offset); \ - __fwd; \ -}) -static void ilk_dummy_write (struct drm_i915_private *dev_priv) {
@@ -1034,37 +1006,21 @@ gen8_write##x(struct drm_i915_private *dev_priv, i915_reg_t reg, u##x val, boolGEN6_WRITE_FOOTER ; \ } -#define __chv_write (x) \ -static void \ -chv_write##x(struct drm_i915_private *dev_priv, i915_reg_t reg, u##x val, bool trace) { \ - enum forcewake_domains fw_engine; \ - GEN6_WRITE_HEADER; \ - fw_engine = __chv_reg_write_fw_domains(offset); \ - if (fw_engine) \ - __force_wake_auto(dev_priv, fw_engine); \ - __raw_i915_write##x(dev_priv, reg, val); \ - GEN6_WRITE_FOOTER; \ -} - -#define __gen9_write (x) \ +#define __fwtbl_write (x) \ static void \ -gen9_write##x(struct drm_i915_private *dev_priv, i915_reg_t reg, u##x val, \ - bool trace) { \ +fwtbl_write##x(struct drm_i915_private *dev_priv, i915_reg_t reg, u##x val, bool trace) { \ enum forcewake_domains fw_engine; \ GEN6_WRITE_HEADER; \ - fw_engine = __gen9_reg_write_fw_domains(offset); \ + fw_engine = __fwtbl_reg_write_fw_domains(offset); \ if (fw_engine) \ __force_wake_auto(dev_priv, fw_engine); \ __raw_i915_write##x(dev_priv, reg, val); \ GEN6_WRITE_FOOTER; \ } -__gen9_write(8) -__gen9_write(16) -__gen9_write(32) -__chv_write(8) -__chv_write(16) -__chv_write(32) +__fwtbl_write(8) +__fwtbl_write(16) +__fwtbl_write(32) __gen8_write(8) __gen8_write(16)__gen8_write (32 )
@@ -1072,8 +1028,7 @@ __gen6_write(8) __gen6_write(16) __gen6_write(32) -#undef __gen9_write -#undef __chv_write +#undef __fwtbl_write#undef __gen8_write#undef __gen6_write#undef GEN6_WRITE_FOOTER
@@ -1284,13 +1239,13 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)default :case 9 :ASSIGN_FW_DOMAINS_TABLE (__gen9_fw_ranges); -ASSIGN_WRITE_MMIO_VFUNCS (gen9); +ASSIGN_WRITE_MMIO_VFUNCS (fwtbl);ASSIGN_READ_MMIO_VFUNCS (fwtbl);break ;case 8 :if (IS_CHERRYVIEW (dev_priv)) {ASSIGN_FW_DOMAINS_TABLE (__chv_fw_ranges); -ASSIGN_WRITE_MMIO_VFUNCS (chv); +ASSIGN_WRITE_MMIO_VFUNCS (fwtbl);ASSIGN_READ_MMIO_VFUNCS (fwtbl); }else {
@@ -1819,11 +1774,11 @@ intel_uncore_forcewake_for_write(struct drm_i915_private *dev_priv,switch (INTEL_GEN (dev_priv)) {case 9 : - fw_domains =__gen9_reg_write_fw_domains (i915_mmio_reg_offset (reg)); + fw_domains =__fwtbl_reg_write_fw_domains (i915_mmio_reg_offset (reg));break ;case 8 :if (IS_CHERRYVIEW (dev_priv)) - fw_domains =__chv_reg_write_fw_domains (i915_mmio_reg_offset (reg)); + fw_domains =__fwtbl_reg_write_fw_domains (i915_mmio_reg_offset (reg));else fw_domains =__gen8_reg_write_fw_domains (i915_mmio_reg_offset (reg));break ;
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 11/13] drm/i915: Remove identical write mmmio functions
Quoted 7 lines · Tvrtko Ursulin
From: Tvrtko Ursulin <[email protected]>
We notice two identical copies of the shadow register table and following from that removal can also unify CHV and Gen9 write mmio functions and macros into a single implementation.
Signed-off-by Tvrtko Ursulin <[email protected]>
Regards, Joonas
Signature · 7 lines
Joonas Lahtinen Open Source Technology Center Intel Corporation _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 12/13 drm/i915: Sort the shadow register table
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
Also verify the order at runtime. This was we can start using binary search on it in a following patch.
--- drivers/gpu/drm/i915/intel_uncore.c | 28 ++++++++++++++++++++++------ 1 file changed, 22 insertions(+), 6 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +22 -6
@@ -657,15 +657,29 @@ static const struct intel_forcewake_range __vlv_fw_ranges[] = { })static const i915_reg_t gen8_shadowed_regs[] = { -GEN6_RPNSWREQ , -GEN6_RC_VIDEO_FREQ , -RING_TAIL (RENDER_RING_BASE ), -RING_TAIL (GEN6_BSD_RING_BASE ), -RING_TAIL (VEBOX_RING_BASE ), -RING_TAIL (BLT_RING_BASE ), +RING_TAIL (RENDER_RING_BASE ),/* 0x2000 (base) */ +GEN6_RPNSWREQ ,/* 0xA008 */ +GEN6_RC_VIDEO_FREQ ,/* 0xA00C */ +RING_TAIL (GEN6_BSD_RING_BASE ),/* 0x12000 (base) */ +RING_TAIL (VEBOX_RING_BASE ),/* 0x1a000 (base) */ +RING_TAIL (BLT_RING_BASE ),/* 0x22000 (base) */ /* TODO: Other registers are not yet used */ }; +static void intel_shadow_table_check (void ) +{ +i915_reg_t *reg = (i915_reg_t *)gen8_shadowed_regs; +s32 prev =-1 ; +u32 offset; +unsigned int i; + +for (i =0 ; i <ARRAY_SIZE (gen8_shadowed_regs); i++, reg++) { + offset =i915_mmio_reg_offset (*reg); +WARN_ON_ONCE (prev >= (s32 )offset); + prev = offset; + } +} +static bool is_gen8_shadowed (u32 offset) {int i;
@@ -1277,6 +1291,8 @@ void intel_uncore_init(struct drm_i915_private *dev_priv) }intel_fw_table_check (dev_priv); +if (INTEL_GEN (dev_priv) >=8 ) +intel_shadow_table_check ();if (intel_vgpu_active (dev_priv)) {ASSIGN_WRITE_MMIO_VFUNCS (vgpu);
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 12/13] drm/i915: Sort the shadow register table
Quoted 22 lines · Tvrtko Ursulin
From: Tvrtko Ursulin <[email protected]>
Also verify the order at runtime. This was we can start using binary search on it in a following patch.
Signed-off-by Tvrtko Ursulin <[email protected]>--- drivers/gpu/drm/i915/intel_uncore.c | 28 ++++++++++++++++++++++------ 1 file changed, 22 insertions(+), 6 deletions(-)drivers/gpu/drm/i915/intel_uncore.c +0
-4@@ -657,15 +657,29 @@ static const struct intel_forcewake_range __vlv_fw_ranges[] = { })static const i915_reg_t gen8_shadowed_regs[] = { -GEN6_RPNSWREQ , -GEN6_RC_VIDEO_FREQ , -RING_TAIL (RENDER_RING_BASE ), -RING_TAIL (GEN6_BSD_RING_BASE ),
drivers/gpu/drm/i915/intel_uncore.c static const struct intel_forcewake_range __vlv_fw_ranges[] = { · line 667
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +6
-2- RING_TAIL(VEBOX_RING_BASE), - RING_TAIL(BLT_RING_BASE), +RING_TAIL (RENDER_RING_BASE ),/* 0x2000 (base) */ +GEN6_RPNSWREQ ,/* 0xA008 */ +GEN6_RC_VIDEO_FREQ ,/* 0xA00C */ +RING_TAIL (GEN6_BSD_RING_BASE ),/* 0x12000 (base) */ +RING_TAIL (VEBOX_RING_BASE ),/* 0x1a000 (base) */ + RING_TAIL(BLT_RING_BASE), /* 0x22000 (base) */ /* TODO: Other registers are not yet used */ };
Ok.
-Chris
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 12/13] drm/i915: Sort the shadow register table
In reply to Tvrtko UrsulinQuoted 4 lines · Tvrtko Ursulin
drivers/gpu/drm/i915/intel_uncore.c +4
-0+static void intel_shadow_table_check (void ) +{ +i915_reg_t *reg = (i915_reg_t *)gen8_shadowed_regs; +s32 prev =-1 ;
drivers/gpu/drm/i915/intel_uncore.c static const struct intel_forcewake_range __vlv_fw_ranges[] = { · line 682
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +10
-0+u32 offset; +unsigned int i; + +for (i =0 ; i <ARRAY_SIZE (gen8_shadowed_regs); i++, reg++) { + offset =i915_mmio_reg_offset (*reg); +WARN_ON_ONCE (prev >= (s32 )offset); + prev = offset; + } +} +
BUILD_BUG_ON would gain extra points, but this is fine too :)
Regards, Joonas
Signature · 7 lines
Joonas Lahtinen Open Source Technology Center Intel Corporation _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 12/13] drm/i915: Sort the shadow register table
Quoted 16 lines · Joonas Lahtinen
Quoted 4 lines · Tvrtko Ursulin
drivers/gpu/drm/i915/intel_uncore.c +4
-0+static void intel_shadow_table_check (void ) +{ +i915_reg_t *reg = (i915_reg_t *)gen8_shadowed_regs; +s32 prev =-1 ;drivers/gpu/drm/i915/intel_uncore.c static const struct intel_forcewake_range __vlv_fw_ranges[] = { · line 682
Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +10
-0+u32 offset; +unsigned int i; + +for (i =0 ; i <ARRAY_SIZE (gen8_shadowed_regs); i++, reg++) { + offset =i915_mmio_reg_offset (*reg); +WARN_ON_ONCE (prev >= (s32 )offset); + prev = offset; + } +} +
BUILD_BUG_ONwould gain extra points, but this is fine too :)
It is possible to do that, how?
Opinions of whether these checks should always be there or maybe under #ifdef CONFIG_DRM_I915_DEBUG perhaps?
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Joonas Lahtinen
maintainerRe: [PATCH 12/13] drm/i915: Sort the shadow register table
Quoted 3 lines · Tvrtko Ursulin
It is possible to do that, how?
I think it'd enable uncomfortable amount of macros. So maybe not worth the bother :)
Quoted 4 lines
Opinions of whether these checks should always be there or maybe under
#ifdefCONFIG_DRM_I915_DEBUGperhaps?
I think I915_DEBUG around would be good idea, just omit the function body if not enabled.
Regards, Joonas
Signature · 7 lines
Joonas Lahtinen Open Source Technology Center Intel Corporation _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tvrtko Ursulin
maintainerPATCH 13/13 drm/i915: Use binary search when looking for shadowed registers
In reply to Tvrtko UrsulinFrom: Tvrtko Ursulin <[email protected]>
Simply replace the linear search with the kernel's binary search implementation. There is only six registers currently in that table so this may not be that interesting. It adds a function call so hopefully remains performance neutral for now.
--- drivers/gpu/drm/i915/intel_uncore.c | 26 +++++++++++++++++++++----- 1 file changed, 21 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +21 -5
@@ -680,14 +680,30 @@ static void intel_shadow_table_check(void) } } +static int mmio_reg_cmp (const void *key,const void *elt) +{ +u32 offset = (u32 )(unsigned long )key; +i915_reg_t *reg = (i915_reg_t *)elt; + +if (offset <i915_mmio_reg_offset (*reg)) +return -1 ; +else if (offset >i915_mmio_reg_offset (*reg)) +return 1 ; +else +return 0 ; +} +static bool is_gen8_shadowed (u32 offset) { -int i; -for (i =0 ; i <ARRAY_SIZE (gen8_shadowed_regs); i++) -if (offset == gen8_shadowed_regs[i].reg) -return true; +i915_reg_t *reg; -return false; + reg =bsearch ((void *)(unsigned long )offset, + (const void *)gen8_shadowed_regs, +ARRAY_SIZE (gen8_shadowed_regs), +sizeof (i915_reg_t), + mmio_reg_cmp); + +return reg ? true : false; }#define __gen8_reg_write_fw_domains (offset) \
Signature · 6 lines
2.7.4 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 13/13] drm/i915: Use binary search when looking for shadowed registers
Quoted 1 line · Tvrtko Ursulin
+static int
mmio_reg_cmp(const void *key, const void *elt)
+ {
<SNIP>
drivers/gpu/drm/i915/intel_uncore.c static void intel_shadow_table_check(void) · line 693
Quoted 7 lines
drivers/gpu/drm/i915/intel_uncore.c +6
-0+if (offset <i915_mmio_reg_offset (*reg)) +return -1 ; +else if (offset >i915_mmio_reg_offset (*reg)) +return 1 ; +else +return 0 ;
I was about to suggest return a - b; but then realized it might overflow. After a bit of searching, that is done in regcache_default_cmp though. So not sure if it's a bug in there or did I miss something. This is safe way at least.
drivers/gpu/drm/i915/intel_uncore.c static void intel_shadow_table_check(void) · line 700
Quoted 10 lines
static bool
is_gen8_shadowed(u32 offset)
{drivers/gpu/drm/i915/intel_uncore.c static void intel_shadow_table_check(void) · line 698
Quoted 5 lines
drivers/gpu/drm/i915/intel_uncore.c +1
-4-int i; -for (i =0 ; i <ARRAY_SIZE (gen8_shadowed_regs); i++) -if (offset == gen8_shadowed_regs[i].reg) -return true; +i915_reg_t *reg;Quoted 1 line
drivers/gpu/drm/i915/intel_uncore.c +0
-1-return false;drivers/gpu/drm/i915/intel_uncore.c +1
-0+ reg = bsearch((void *)(unsigned long)offset,
(unsigned long) is not required, as we don't have sign bits to extend.
drivers/gpu/drm/i915/intel_uncore.c static void intel_shadow_table_check(void) · line 706
Quoted 6 lines
drivers/gpu/drm/i915/intel_uncore.c +6
-0+ (const void *)gen8_shadowed_regs, + ARRAY_SIZE(gen8_shadowed_regs), + sizeof(i915_reg_t), + mmio_reg_cmp); + +return reg ? true : false;
As I previously learnt, no need to explicitly convert bool types anymore, so this can just be return bsearch(...);
With the last two fixed,
Regards, Joonas
Signature · 7 lines
Joonas Lahtinen Open Source Technology Center Intel Corporation _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 13/13] drm/i915: Use binary search when looking for shadowed registers
Quoted 23 lines · Joonas Lahtinen
Quoted 1 line · Tvrtko Ursulin
+static int
mmio_reg_cmp(const void *key, const void *elt)+ {
<SNIP>drivers/gpu/drm/i915/intel_uncore.c static void intel_shadow_table_check(void) · line 693
Quoted 6 lines
drivers/gpu/drm/i915/intel_uncore.c +6
-0+if (offset <i915_mmio_reg_offset (*reg)) +return -1 ; +else if (offset >i915_mmio_reg_offset (*reg)) +return 1 ; +else +return 0 ;I was about to suggest return a - b; but then realized it might overflow. After a bit of searching, that is done in
regcache_default_cmpthough. So not sure if it's a bug in there or did I miss something. This is safe way at least.Quoted 8 lines
static bool
is_gen8_shadowed(u32 offset)
{Quoted 5 lines
drivers/gpu/drm/i915/intel_uncore.c +1
-4-int i; -for (i =0 ; i <ARRAY_SIZE (gen8_shadowed_regs); i++) -if (offset == gen8_shadowed_regs[i].reg) -return true; +i915_reg_t *reg;
Quoted 3 lines
drivers/gpu/drm/i915/intel_uncore.c static void intel_shadow_table_check(void) · line 700
Quoted 2 lines
Quoted 1 line
drivers/gpu/drm/i915/intel_uncore.c +0
-1-return false;drivers/gpu/drm/i915/intel_uncore.c +1
-0+ reg = bsearch((void *)(unsigned long)offset,(unsigned long) is not required, as we don't have sign bits to extend.
Gcc warns if I omit that.
Quoted 11 lines
drivers/gpu/drm/i915/intel_uncore.c static void intel_shadow_table_check(void) · line 706
Quoted 7 lines
drivers/gpu/drm/i915/intel_uncore.c +6
-0+ (const void *)gen8_shadowed_regs, +ARRAY_SIZE (gen8_shadowed_regs), +sizeof (i915_reg_t), + mmio_reg_cmp); + +return reg ? true : false;As I previously learnt, no need to explicitly convert bool types anymore, so this can just be return bsearch(...);
With the last two fixed,
I've done this one.
Quoted 1 line
Reviewed-by Joonas Lahtinen <[email protected]>
Thanks!
Tvrtko
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Jani Nikula
maintainerQuoted 9 lines · Tvrtko Ursulin
From: Tvrtko Ursulin <[email protected]>
A bunch of cleaned up patches from one of my old branches. Just an idea I had been toying about and probably more a RFC at this point. But it amazingly passed trybot on (almost) the first try so I got tempted to post it.
Basically main motivation was to replace linear search comparison ladders in all mmio accessors with something smarter - a binary search in this case.
Quoted 7 lines
Main enabler is moving the comparison ladders from the code into the data tables. Then gradually, patch by patch, I was able to reduce the code and end up with a total saving of approximately 14KiB for the series.
That is kind of nice, but whether the churn is worth it I leave open for dicussion. It is slightly less code so perhaps more maintainable. Although tastes will probably differ.
Glanced through it quickly, I think the new stuff looks prettier and more manageable. I'm not opposed.
BR,
Jani.
Quoted 22 lines
And I haven't been able to do any benchmarking to see if any performance differece is at all detectable.
Tvrtko Ursulin (13):
drm/i915: Remove redundant hsw_write* mmio functions drm/i915: Keep track of active forcewake domains in a bitmask drm/i915: Do not inline forcewake taking in mmio accessors drm/i915: Data driven register to forcewake domains lookup drm/i915: Sort forcewake mapping tables
drm/i915: Use binary search when looking up forcewake domains drm/i915: Eliminate Gen9 special case
drm/i915: Store the active forcewake range table pointer drm/i915: Remove identical macros
drm/i915: Remove identical mmio read functions drm/i915: Remove identical write mmmio functions drm/i915: Sort the shadow register table
drm/i915: Use binary search when looking for shadowed registersdrivers/gpu/drm/i915/i915_drv.h | 13 + drivers/gpu/drm/i915/intel_uncore.c | 525 ++++++++++++++++-------------------- 2 files changed, 247 insertions(+), 291 deletions(-)
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Patchwork
CI✗ Fi.CI.BAT: warning for Forcewake binary search & code shrink
In reply to Tvrtko Ursulin== Series Details ==
== Series Details ==
Series: Forcewake binary search & code shrink
URL : https://patchwork.freedesktop.org/series/13080/
State : warning
== Summary ==
Series 13080v1 Forcewake binary search & code shrink
https://patchwork.freedesktop.org/api/1.0/series/13080/revisions/1/mbox/
Test kms_pipe_crc_basic:
Subgroup read-crc-pipe-b:
pass -> DMESG-WARN (fi-ilk-650)fi-bdw-5557u total:244 pass:229 dwarn:0 dfail:0 fail:0 skip:15 fi-bsw-n3050 total:244 pass:202 dwarn:0 dfail:0 fail:0 skip:42 fi-bxt-t5700 total:244 pass:214 dwarn:0 dfail:0 fail:0 skip:30 fi-byt-n2820 total:244 pass:208 dwarn:0 dfail:0 fail:1 skip:35 fi-hsw-4770 total:244 pass:222 dwarn:0 dfail:0 fail:0 skip:22 fi-hsw-4770r total:244 pass:222 dwarn:0 dfail:0 fail:0 skip:22 fi-ilk-650 total:244 pass:181 dwarn:1 dfail:0 fail:2 skip:60 fi-ivb-3520m total:244 pass:219 dwarn:0 dfail:0 fail:0 skip:25 fi-ivb-3770 total:244 pass:207 dwarn:0 dfail:0 fail:0 skip:37 fi-skl-6260u total:244 pass:230 dwarn:0 dfail:0 fail:0 skip:14 fi-skl-6700hq total:244 pass:221 dwarn:1 dfail:0 fail:0 skip:22 fi-skl-6700k total:244 pass:219 dwarn:1 dfail:0 fail:0 skip:24 fi-skl-6770hq total:244 pass:228 dwarn:1 dfail:0 fail:1 skip:14 fi-snb-2520m total:244 pass:208 dwarn:0 dfail:0 fail:0 skip:36 fi-snb-2600 total:244 pass:207 dwarn:0 dfail:0 fail:0 skip:37
fi-byt-j1900 failed to connect after reboot
Results at /archive/results/CI_IGT_test/Patchwork_2594/
2fbc239494fa3883e164cc89d35f54d22f0c9e1f drm-intel-nightly: 2016y-09m-29d-13h-31m-39s UTC integration manifest d042c9c drm/i915: Use binary search when looking for shadowed registers
dbafe06 drm/i915: Sort the shadow register table
b9121e6 drm/i915: Remove identical write mmmio functions
0780db3 drm/i915: Remove identical mmio read functions
b1f3e47 drm/i915: Remove identical macros
777472d drm/i915: Store the active forcewake range table pointer
20718aa drm/i915: Eliminate Gen9 special case
09c6e3b drm/i915: Use binary search when looking up forcewake domains
98dd975 drm/i915: Sort forcewake mapping tables
f89fd58 drm/i915: Data driven register to forcewake domains lookup
bac5870 drm/i915: Do not inline forcewake taking in mmio accessors
6b67148 drm/i915: Keep track of active forcewake domains in a bitmask
cf28017 drm/i915: Remove redundant hsw_write* mmio functions
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 9 lines · Tvrtko Ursulin
From: Tvrtko Ursulin <[email protected]>
A bunch of cleaned up patches from one of my old branches. Just an idea I had been toying about and probably more a RFC at this point. But it amazingly passed trybot on (almost) the first try so I got tempted to post it.
Basically main motivation was to replace linear search comparison ladders in all mmio accessors with something smarter - a binary search in this case.
Quoted 7 lines
Main enabler is moving the comparison ladders from the code into the data tables. Then gradually, patch by patch, I was able to reduce the code and end up with a total saving of approximately 14KiB for the series.
That is kind of nice, but whether the churn is worth it I leave open for dicussion. It is slightly less code so perhaps more maintainable. Although tastes will probably differ.
More declarative approach, yes please. Looks really nice. -Chris
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Patchwork
CI✗ Fi.CI.BAT: failure for Forcewake binary search & code shrink (rev2)
In reply to Tvrtko Ursulin== Series Details ==
== Series Details ==
Series: Forcewake binary search & code shrink (rev2)
URL : https://patchwork.freedesktop.org/series/13080/
State : failure
== Summary ==
LD fs/btrfs/built-in.o CC arch/x86/kernel/cpu/capflags.o LD arch/x86/kernel/cpu/built-in.o LD arch/x86/kernel/built-in.o LD drivers/md/md-mod.o LD drivers/md/built-in.o LD [M] drivers/net/ethernet/intel/igb/igb.o LD drivers/usb/core/usbcore.o LD drivers/usb/core/built-in.o LD arch/x86/built-in.o LD drivers/usb/host/xhci-hcd.o LD fs/ext4/ext4.o LD fs/ext4/built-in.o LD net/core/built-in.o LD fs/built-in.o LD drivers/usb/host/built-in.o LD drivers/usb/built-in.o LD [M] drivers/net/ethernet/intel/e1000e/e1000e.o LD net/ipv4/built-in.o LD net/built-in.o LD drivers/net/ethernet/built-in.o LD drivers/net/built-in.o
scripts/Makefile.build:440: recipe for target 'drivers/gpu/drm/i915' failed make[3]: *** [drivers/gpu/drm/i915] Error 2 scripts/Makefile.build:440: recipe for target 'drivers/gpu/drm' failed make[2]: *** [drivers/gpu/drm] Error 2 scripts/Makefile.build:440: recipe for target 'drivers/gpu' failed make[1]: *** [drivers/gpu] Error 2
Makefile:968: recipe for target 'drivers' failed
make: *** [drivers] Error 2
Full logs at /archive/deploy/logs/Patchwork_2598
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx