Skip to contentSkip to conversation

drm/radeon: Inline r100_mm_rreg, -wreg, v3

PATCH Applied Lauri Kasanen 3 participants 5 messages Apr 20, 2014 – Jul 10, 2014

Subsystems (2)
DRM drivers Radeon and amdgpu DRM drivers
lore
Share & tools
lore

Lauri Kasanen

author
#0 lore raw

PATCH 1/1 drm/radeon: Inline r100_mm_rreg, -wreg, v3

This was originally un-inlined by Andi Kleen in 2011 citing size concerns. Indeed, a first attempt at inlining it grew radeon.ko by 7%.

However, 2% of cpu is spent in this function. Simply inlining it gave 1% more fps in Urban Terror.

v2: We know the minimum MMIO size. Adding it to the if allows the compiler to optimize the branch out, improving both performance and size.

The v2 patch decreases radeon.ko size by 2%. I didn't re-benchmark, but common sense says perf is now more than 1% better.

v3: Also change _wreg, make the threshold a define.

Inlining _wreg increased the size a bit compared to v2, so now radeon.ko is only 1% smaller.

Signed-off-by Lauri Kasanen <[email protected]>
---
 drivers/gpu/drm/radeon/r100.c   | 33 ---------------------------------
 drivers/gpu/drm/radeon/radeon.h | 40 ++++++++++++++++++++++++++++++++++++----
 2 files changed, 36 insertions(+), 37 deletions(-)
drivers/gpu/drm/radeon/r100.c +0 -33
index b6c3264..a4e7871 100644
@@ -4086,39 +4086,6 @@ int r100_init(struct radeon_device *rdev)
 	return 0;
 }
 
-uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg,
-		      bool always_indirect)
-{
-	if (reg < rdev->rmmio_size && !always_indirect)
-		return readl(((void __iomem *)rdev->rmmio) + reg);
-	else {
-		unsigned long flags;
-		uint32_t ret;
-
-		spin_lock_irqsave(&rdev->mmio_idx_lock, flags);
-		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
-		ret = readl(((void __iomem *)rdev->rmmio) + RADEON_MM_DATA);
-		spin_unlock_irqrestore(&rdev->mmio_idx_lock, flags);
-
-		return ret;
-	}
-}
-
-void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v,
-		  bool always_indirect)
-{
-	if (reg < rdev->rmmio_size && !always_indirect)
-		writel(v, ((void __iomem *)rdev->rmmio) + reg);
-	else {
-		unsigned long flags;
-
-		spin_lock_irqsave(&rdev->mmio_idx_lock, flags);
-		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
-		writel(v, ((void __iomem *)rdev->rmmio) + RADEON_MM_DATA);
-		spin_unlock_irqrestore(&rdev->mmio_idx_lock, flags);
-	}
-}
-
 u32 r100_io_rreg(struct radeon_device *rdev, u32 reg)
 {
 	if (reg < rdev->rio_mem_size)
drivers/gpu/drm/radeon/radeon.h +36 -4
index f21db7a..a749b6c 100644
@@ -2328,10 +2328,42 @@ int radeon_device_init(struct radeon_device *rdev,
 void radeon_device_fini(struct radeon_device *rdev);
 int radeon_gpu_wait_for_idle(struct radeon_device *rdev);
 
-uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg,
-		      bool always_indirect);
-void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v,
-		  bool always_indirect);
+#define RADEON_MIN_MMIO_SIZE 0x10000
+
+static inline uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg,
+				    bool always_indirect)
+{
+	/* The mmio size is 64kb at minimum. Allows the if to be optimized out. */
+	if ((reg < rdev->rmmio_size || reg < RADEON_MIN_MMIO_SIZE) && !always_indirect)
+		return readl(((void __iomem *)rdev->rmmio) + reg);
+	else {
+		unsigned long flags;
+		uint32_t ret;
+
+		spin_lock_irqsave(&rdev->mmio_idx_lock, flags);
+		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
+		ret = readl(((void __iomem *)rdev->rmmio) + RADEON_MM_DATA);
+		spin_unlock_irqrestore(&rdev->mmio_idx_lock, flags);
+
+		return ret;
+	}
+}
+
+static inline void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v,
+				bool always_indirect)
+{
+	if ((reg < rdev->rmmio_size || reg < RADEON_MIN_MMIO_SIZE) && !always_indirect)
+		writel(v, ((void __iomem *)rdev->rmmio) + reg);
+	else {
+		unsigned long flags;
+
+		spin_lock_irqsave(&rdev->mmio_idx_lock, flags);
+		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
+		writel(v, ((void __iomem *)rdev->rmmio) + RADEON_MM_DATA);
+		spin_unlock_irqrestore(&rdev->mmio_idx_lock, flags);
+	}
+}
+
 u32 r100_io_rreg(struct radeon_device *rdev, u32 reg);
 void r100_io_wreg(struct radeon_device *rdev, u32 reg, u32 v);
Signature · 1 line
1.8.3.1
Reviewed-by on v1 1/1
#1 lore raw

Am 20.04.2014 19:29, schrieb Lauri Kasanen:

Quoted 12 lines

This was originally un-inlined by Andi Kleen in 2011 citing size concerns. Indeed, a first attempt at inlining it grew radeon.ko by 7%.

However, 2% of cpu is spent in this function. Simply inlining it gave 1% more fps in Urban Terror.

v2: We know the minimum MMIO size. Adding it to the if allows the compiler to optimize the branch out, improving both performance and size.

The v2 patch decreases radeon.ko size by 2%. I didn't re-benchmark, but common sense says perf is now more than 1% better.

Quoted 6 lines

v3: Also change _wreg, make the threshold a define.

Inlining _wreg increased the size a bit compared to v2, so now radeon.ko is only 1% smaller.

Signed-off-by Lauri Kasanen <[email protected]>
Reviewed-by Christian König <[email protected]>
Quoted 100 lines
---
  drivers/gpu/drm/radeon/r100.c   | 33 ---------------------------------
  drivers/gpu/drm/radeon/radeon.h | 40 ++++++++++++++++++++++++++++++++++++----
  2 files changed, 36 insertions(+), 37 deletions(-)
drivers/gpu/drm/radeon/r100.c +0 -33
index b6c3264..a4e7871 100644
@@ -4086,39 +4086,6 @@ int r100_init(struct radeon_device *rdev)
  	return 0;
  }
 
-uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg,
-		      bool always_indirect)
-{
-	if (reg < rdev->rmmio_size && !always_indirect)
-		return readl(((void __iomem *)rdev->rmmio) + reg);
-	else {
-		unsigned long flags;
-		uint32_t ret;
-
-		spin_lock_irqsave(&rdev->mmio_idx_lock, flags);
-		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
-		ret = readl(((void __iomem *)rdev->rmmio) + RADEON_MM_DATA);
-		spin_unlock_irqrestore(&rdev->mmio_idx_lock, flags);
-
-		return ret;
-	}
-}
-
-void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v,
-		  bool always_indirect)
-{
-	if (reg < rdev->rmmio_size && !always_indirect)
-		writel(v, ((void __iomem *)rdev->rmmio) + reg);
-	else {
-		unsigned long flags;
-
-		spin_lock_irqsave(&rdev->mmio_idx_lock, flags);
-		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
-		writel(v, ((void __iomem *)rdev->rmmio) + RADEON_MM_DATA);
-		spin_unlock_irqrestore(&rdev->mmio_idx_lock, flags);
-	}
-}
-
  u32 r100_io_rreg(struct radeon_device *rdev, u32 reg)
  {
  	if (reg < rdev->rio_mem_size)
drivers/gpu/drm/radeon/radeon.h +36 -4
index f21db7a..a749b6c 100644
@@ -2328,10 +2328,42 @@ int radeon_device_init(struct radeon_device *rdev,
  void radeon_device_fini(struct radeon_device *rdev);
  int radeon_gpu_wait_for_idle(struct radeon_device *rdev);
 
-uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg,
-		      bool always_indirect);
-void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v,
-		  bool always_indirect);
+#define RADEON_MIN_MMIO_SIZE 0x10000
+
+static inline uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg,
+				    bool always_indirect)
+{
+	/* The mmio size is 64kb at minimum. Allows the if to be optimized out. */
+	if ((reg < rdev->rmmio_size || reg < RADEON_MIN_MMIO_SIZE) && !always_indirect)
+		return readl(((void __iomem *)rdev->rmmio) + reg);
+	else {
+		unsigned long flags;
+		uint32_t ret;
+
+		spin_lock_irqsave(&rdev->mmio_idx_lock, flags);
+		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
+		ret = readl(((void __iomem *)rdev->rmmio) + RADEON_MM_DATA);
+		spin_unlock_irqrestore(&rdev->mmio_idx_lock, flags);
+
+		return ret;
+	}
+}
+
+static inline void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v,
+				bool always_indirect)
+{
+	if ((reg < rdev->rmmio_size || reg < RADEON_MIN_MMIO_SIZE) && !always_indirect)
+		writel(v, ((void __iomem *)rdev->rmmio) + reg);
+	else {
+		unsigned long flags;
+
+		spin_lock_irqsave(&rdev->mmio_idx_lock, flags);
+		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
+		writel(v, ((void __iomem *)rdev->rmmio) + RADEON_MM_DATA);
+		spin_unlock_irqrestore(&rdev->mmio_idx_lock, flags);
+	}
+}
+
  u32 r100_io_rreg(struct radeon_device *rdev, u32 reg);
  void r100_io_wreg(struct radeon_device *rdev, u32 reg, u32 v);
author
#2 lore raw
Quoted 13 lines · On Sun, 20 Apr 2014 19:41:11 +0200 Christian König <[email protected]>…

Am 20.04.2014 19:29, schrieb Lauri Kasanen:

Quoted 12 lines

This was originally un-inlined by Andi Kleen in 2011 citing size concerns. Indeed, a first attempt at inlining it grew radeon.ko by 7%.

However, 2% of cpu is spent in this function. Simply inlining it gave 1% more fps in Urban Terror.

v2: We know the minimum MMIO size. Adding it to the if allows the compiler to optimize the branch out, improving both performance and size.

The v2 patch decreases radeon.ko size by 2%. I didn't re-benchmark, but common sense says perf is now more than 1% better.

Quoted 8 lines
Quoted 6 lines

v3: Also change _wreg, make the threshold a define.

Inlining _wreg increased the size a bit compared to v2, so now radeon.ko is only 1% smaller.

Signed-off-by Lauri Kasanen <[email protected]>
Reviewed-by Christian König <[email protected]>

Ping. Although reviewed in April, seems this wasn't applied to any tree?

- Lauri

#3 lore raw

Am 10.07.2014 10:48, schrieb Lauri Kasanen:

Quoted 16 lines
Quoted 13 lines · On Sun, 20 Apr 2014 19:41:11 +0200 Christian König <[email protected]>…

Am 20.04.2014 19:29, schrieb Lauri Kasanen:

Quoted 12 lines

This was originally un-inlined by Andi Kleen in 2011 citing size concerns. Indeed, a first attempt at inlining it grew radeon.ko by 7%.

However, 2% of cpu is spent in this function. Simply inlining it gave 1% more fps in Urban Terror.

v2: We know the minimum MMIO size. Adding it to the if allows the compiler to optimize the branch out, improving both performance and size.

The v2 patch decreases radeon.ko size by 2%. I didn't re-benchmark, but common sense says perf is now more than 1% better.

Quoted 8 lines
Quoted 7 lines
Quoted 6 lines

v3: Also change _wreg, make the threshold a define.

Inlining _wreg increased the size a bit compared to v2, so now radeon.ko is only 1% smaller.

Signed-off-by Lauri Kasanen <[email protected]>
Reviewed-by Christian König <[email protected]>

Ping. Although reviewed in April, seems this wasn't applied to any tree?

Sorry looks like I missed it. Alex can you pull that in your 3.17 branch?

Thanks,
Christian.

Quoted 2 lines

- Lauri

Applied
#4 lore raw
Quoted 28 lines · Christian König

Am 10.07.2014 10:48, schrieb Lauri Kasanen:

Quoted 26 lines
Quoted 23 lines · On Sun, 20 Apr 2014 19:41:11 +0200 Christian König <[email protected]>…

Am 20.04.2014 19:29, schrieb Lauri Kasanen:

Quoted 22 lines

This was originally un-inlined by Andi Kleen in 2011 citing size concerns.
Indeed, a first attempt at inlining it grew radeon.ko by 7%.

However, 2% of cpu is spent in this function. Simply inlining it gave 1% more fps
in Urban Terror.

v2: We know the minimum MMIO size. Adding it to the if allows the compiler to
optimize the branch out, improving both performance and size.

The v2 patch decreases radeon.ko size by 2%. I didn't re-benchmark, but common sense
says perf is now more than 1% better.

v3: Also change _wreg, make the threshold a define.

Inlining _wreg increased the size a bit compared to v2, so now radeon.ko is only 1% smaller.

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Quoted 1 line
Signed-off-by Lauri Kasanen <[email protected]>
Reviewed-by Christian König <[email protected]>

Ping. Although reviewed in April, seems this wasn't applied to any tree?

Sorry looks like I missed it. Alex can you pull that in your 3.17 branch?

Applied. thanks.

Alex
_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel