feat(gpu): show VRAM in use as a bar with a percentage

Sum drm-resident-vram0 from DRM fdinfo, the same per-client counters
gputop reads, counting each drm-client-id once since a client repeats
its totals on every fd it holds. One grep over /proc/*/fdinfo, about
10ms warm, sampled through new_gate every ~2s.

The 12 GiB total is hardcoded: xe exposes no VRAM capacity in sysfs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

1 files changed, 33 insertions, 7 deletions

diff --git a/widgets/gpu.lua b/widgets/gpu.lua
index 7d5f2ed..abac856 100644
--- a/widgets/gpu.lua
+++ b/widgets/gpu.lua
@@ -1,10 +1,9 @@
--- GPU: Arc B580 temperatures and fan speed.
+-- GPU: Arc B580 temperatures, VRAM in use and fan speed.
 --
 -- No utilisation figure, deliberately. The xe driver exposes no
--- gpu_busy_percent, intel_gpu_top refuses the device outright ("Detected Xe
--- device which is not supported"), and gputop prints per-process rows with
--- ANSI escapes, which is not an interface to build a widget on. The card shows
--- what the hardware actually reports rather than inventing a number.
+-- gpu_busy_percent and intel_gpu_top refuses the device outright ("Detected Xe
+-- device which is not supported"). VRAM comes from DRM fdinfo, the per-client
+-- counters gputop itself reads (see data.drm_vram).
 --
 -- The integrated AMD GPU does expose gpu_busy_percent and is deliberately not
 -- shown: the discrete card is the one in use.
@@ -14,6 +13,15 @@ local data = require 'lib.data'
 
 local M = {}
 
+-- The B580's VRAM. fdinfo reports what clients hold, not the card's size, and
+-- xe puts no capacity in sysfs. ponytail: hardcoded like the 'ARC B580' label.
+local VRAM_TOTAL = 12 * 1024 ^ 3
+
+-- fdinfo is a popen over every fd on the system, so sample it every ~2s, not
+-- every frame.
+local gate = data.new_gate(1.9)
+local vram
+
 -- chip, file, label, warn, crit. Package (temp2) is the headline above, so it
 -- is not repeated here; these are the rows drawn under it.
 local TEMPS = {
@@ -24,6 +32,8 @@ function M.draw(cr, rect, colors)
   local inner = card.card(cr, rect, colors)
   local function clamp(v, lo, hi) return math.max(lo, math.min(hi, v)) end
 
+  if gate() then vram = data.drm_vram() end
+
   local pkg = data.sensor('xe', 'temp2_input')
   local big_txt = pkg and string.format('%d\u{00B0}', pkg) or '--'
 
@@ -32,8 +42,8 @@ function M.draw(cr, rect, colors)
   -- only where the header or a row would overrun the card. So a tall cell
   -- gets larger type instead of empty space, and nothing ever collides.
   local LABEL_F, ROW_F = 0.32, 0.34
-  local row_n = #TEMPS + 1                            -- VRAM + FAN
-  local S_h = inner.h * 0.94 / (2.088 + row_n * 0.646)
+  local row_n = #TEMPS + 2                            -- VRAM + MEM + FAN
+  local S_h = inner.h * 0.94 / (2.088 + row_n * 0.646 + 0.3)  -- + MEM bar
   local groups = {
     { { card.title_str(card.ICON.gpu, 'GPU'), card.FONT_MONO, LABEL_F },
       { big_txt, card.FONT_HEAVY, 1.0 } },
@@ -43,6 +53,8 @@ function M.draw(cr, rect, colors)
     groups[#groups + 1] = { { t[3], card.FONT_MONO, ROW_F },
                             { '888\u{00B0}', card.FONT_MONO, ROW_F } }
   end
+  groups[#groups + 1] = { { 'MEM', card.FONT_MONO, ROW_F },
+                          { '100%', card.FONT_MONO, ROW_F } }
   groups[#groups + 1] = { { 'FAN', card.FONT_MONO, ROW_F },
                           { '8888 RPM', card.FONT_MONO, ROW_F } }
   local S = clamp(math.min(S_h, card.fit_unit(cr, inner.w * 0.96, groups, 100)), 10, 72)
@@ -80,6 +92,20 @@ function M.draw(cr, rect, colors)
     ey = ey + step
   end
 
+  -- VRAM as a fraction of the card's, a bar under the row like RAM in system.
+  local bar_h = S * 0.12
+  if ey + step + bar_h + S * 0.1 <= inner.y + inner.h then
+    local frac = vram and vram / VRAM_TOTAL
+    local colour = frac and card.threshold(frac * 100, 25, 75, colors) or colors.label
+    card.rgba(cr, colors.label)
+    card.text(cr, inner.x, ey, 'MEM')
+    card.rgba(cr, colour)
+    card.text_right(cr, inner.x + inner.w, ey,
+      frac and string.format('%d%%', math.floor(frac * 100 + 0.5)) or '--')
+    card.bar(cr, inner.x, ey + row_size * 0.6, inner.w, bar_h, frac, colour, colors)
+    ey = ey + step + bar_h + S * 0.1
+  end
+
   -- Fan RPM is not a temperature, so it carries no threshold colour: a fast
   -- fan is the cooling working, not a fault.
   if ey + step <= inner.y + inner.h then